id
string | submitter
string | authors
string | title
string | comments
string | journal-ref
string | doi
string | report-no
string | categories
string | license
string | abstract
string | versions
list | update_date
timestamp[s] | authors_parsed
sequence |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
2506.03513 | Wei Kou | Wei Kou, Xiaoxuan Lin, Bing'ang Guo, Xurong Chen | Physics-Informed Neural Network Approach to Quark-Antiquark Color Flux Tube | 14 pages, 9 figures | null | null | null | hep-ph hep-th | http://creativecommons.org/licenses/by/4.0/ | We introduce a physics-informed neural network (PINNs) framework for modelling the spatial distribution of chromodynamic fields induced by quark-antiquark pairs, based on lattice Monte Carlo simulations. In contrast to conventional neural networks, PINNs incorporate physical laws--expressed here as differential equations governing type-II superconductivity--directly into the training objective. By embedding these equations into the loss function, we guide the network to learn physically consistent solutions. Adopting an inverse problem approach, we extract the parameters of the superconducting equations from lattice QCD data and subsequently solve them. To accommodate physical boundary conditions, we recast the system into an integro-differential form and extend the analysis within the fractional PINNs framework. The accuracy of the reconstructed field distribution is assessed via relative $L_2$-error norms. We further extract physical observables such as the string tension and the mean width of the flux tube, offering quantitative insight into the confinement mechanism. This method enables the reconstruction of colour field profiles as functions of quark-antiquark separation without recourse to predefined parametric models. Our results illuminate aspects of the dual Meissner effect and highlight the promise of data-driven strategies in addressing non-perturbative challenges in quantum chromodynamics. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 02:57:11 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Kou",
"Wei",
""
],
[
"Lin",
"Xiaoxuan",
""
],
[
"Guo",
"Bing'ang",
""
],
[
"Chen",
"Xurong",
""
]
] |
2506.03514 | Martin Beseda | Paulina Lewandowska, Martin Beseda | Benchmarking gate-based quantum devices via certification of qubit von Neumann measurements | This manuscript is currently under review in Quantum Science and Technology | null | null | null | quant-ph | http://creativecommons.org/licenses/by-sa/4.0/ | We present an updated version of PyQBench, an open-source Python library designed for benchmarking gate-based quantum computers, with a focus on certifying qubit von Neumann measurements. This version extends PyQBench's capabilities by incorporating a certification scheme of quantum measurements that evaluates the accuracy on Noisy Intermediate-Scale Quantum devices, alongside its original functionality of von Neumann measurements' discrimination. PyQBench offers a user-friendly command-line interface and Python library integration, allowing users to implement custom measurement schemes and error models for more advanced benchmarking tasks. The new version is specifically designed to support IBM Q devices through integration with the Qiskit library, enabling seamless benchmarking on real quantum hardware. By providing flexible benchmarking tools, PyQBench addresses the critical need for reliable performance metrics in the NISQ era, contributing to the development of error mitigation techniques and the verification of quantum measurement fidelity. The source code is available on GitHub under an open-source license, encouraging community collaboration and further advancements in quantum hardware benchmarking. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 02:57:43 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Lewandowska",
"Paulina",
""
],
[
"Beseda",
"Martin",
""
]
] |
2506.03515 | Masaya Kawamura | Masaya Kawamura, Takuya Hasumi, Yuma Shirahata, Ryuichi Yamamoto | BitTTS: Highly Compact Text-to-Speech Using 1.58-bit Quantization and Weight Indexing | Accepted to INTERSPEECH 2025 | null | null | null | eess.AS cs.LG cs.SD eess.SP | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | This paper proposes a highly compact, lightweight text-to-speech (TTS) model for on-device applications. To reduce the model size, the proposed model introduces two techniques. First, we introduce quantization-aware training (QAT), which quantizes model parameters during training to as low as 1.58-bit. In this case, most of 32-bit model parameters are quantized to ternary values {-1, 0, 1}. Second, we propose a method named weight indexing. In this method, we save a group of 1.58-bit weights as a single int8 index. This allows for efficient storage of model parameters, even on hardware that treats values in units of 8-bit. Experimental results demonstrate that the proposed method achieved 83 % reduction in model size, while outperforming the baseline of similar model size without quantization in synthesis quality. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 03:02:18 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Kawamura",
"Masaya",
""
],
[
"Hasumi",
"Takuya",
""
],
[
"Shirahata",
"Yuma",
""
],
[
"Yamamoto",
"Ryuichi",
""
]
] |
2506.03516 | Arnab Debnath | Arnab Debnath, Gregory J. Stein, Jana Kosecka | SemNav: A Model-Based Planner for Zero-Shot Object Goal Navigation Using Vision-Foundation Models | Accepted at CVPR 2025 workshop - Foundation Models Meet Embodied Agents | null | null | null | cs.RO cs.AI | http://creativecommons.org/licenses/by/4.0/ | Object goal navigation is a fundamental task in embodied AI, where an agent is instructed to locate a target object in an unexplored environment. Traditional learning-based methods rely heavily on large-scale annotated data or require extensive interaction with the environment in a reinforcement learning setting, often failing to generalize to novel environments and limiting scalability. To overcome these challenges, we explore a zero-shot setting where the agent operates without task-specific training, enabling more scalable and adaptable solution. Recent advances in Vision Foundation Models (VFMs) offer powerful capabilities for visual understanding and reasoning, making them ideal for agents to comprehend scenes, identify relevant regions, and infer the likely locations of objects. In this work, we present a zero-shot object goal navigation framework that integrates the perceptual strength of VFMs with a model-based planner that is capable of long-horizon decision making through frontier exploration. We evaluate our approach on the HM3D dataset using the Habitat simulator and demonstrate that our method achieves state-of-the-art performance in terms of success weighted by path length for zero-shot object goal navigation. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 03:04:54 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Debnath",
"Arnab",
""
],
[
"Stein",
"Gregory J.",
""
],
[
"Kosecka",
"Jana",
""
]
] |
2506.03517 | Ziyi Wu | Ziyi Wu, Anil Kag, Ivan Skorokhodov, Willi Menapace, Ashkan Mirzaei, Igor Gilitschenski, Sergey Tulyakov, Aliaksandr Siarohin | DenseDPO: Fine-Grained Temporal Preference Optimization for Video Diffusion Models | Project page: https://snap-research.github.io/DenseDPO/ | null | null | null | cs.CV | http://creativecommons.org/licenses/by/4.0/ | Direct Preference Optimization (DPO) has recently been applied as a post-training technique for text-to-video diffusion models. To obtain training data, annotators are asked to provide preferences between two videos generated from independent noise. However, this approach prohibits fine-grained comparisons, and we point out that it biases the annotators towards low-motion clips as they often contain fewer visual artifacts. In this work, we introduce DenseDPO, a method that addresses these shortcomings by making three contributions. First, we create each video pair for DPO by denoising corrupted copies of a ground truth video. This results in aligned pairs with similar motion structures while differing in local details, effectively neutralizing the motion bias. Second, we leverage the resulting temporal alignment to label preferences on short segments rather than entire clips, yielding a denser and more precise learning signal. With only one-third of the labeled data, DenseDPO greatly improves motion generation over vanilla DPO, while matching it in text alignment, visual quality, and temporal consistency. Finally, we show that DenseDPO unlocks automatic preference annotation using off-the-shelf Vision Language Models (VLMs): GPT accurately predicts segment-level preferences similar to task-specifically fine-tuned video reward models, and DenseDPO trained on these labels achieves performance close to using human labels. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 03:06:08 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Wu",
"Ziyi",
""
],
[
"Kag",
"Anil",
""
],
[
"Skorokhodov",
"Ivan",
""
],
[
"Menapace",
"Willi",
""
],
[
"Mirzaei",
"Ashkan",
""
],
[
"Gilitschenski",
"Igor",
""
],
[
"Tulyakov",
"Sergey",
""
],
[
"Siarohin",
"Aliaksandr",
""
]
] |
2506.03518 | Jinze Li | Liu Yaokun, Li Jinze, Yu Kaiping | Two self-starting single-solve third-order explicit integration algorithms for second-order nonlinear dynamics | 34 pages, 24 figures | null | null | null | math.NA cs.NA physics.comp-ph | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The single-step explicit time integration methods have long been valuable for solving large-scale nonlinear structural dynamic problems, classified into single-solve and multi-sub-step approaches. However, no existing explicit single-solve methods achieve third-order accuracy. The paper addresses this gap by proposing two new third-order explicit algorithms developed within the framework of self-starting single-solve time integration algorithms, which incorporates 11 algorithmic parameters. The study reveals that fully explicit methods with single-solve cannot reach third-order accuracy for general dynamic problems. Consequently, two novel algorithms are proposed: Algorithm 1 is a fully explicit scheme that achieves third-order accuracy in displacement and velocity for undamped problems; Algorithm 2, which employs implicit treatment of velocity and achieves third-order accuracy for general dynamic problems. Across a suite of both linear and nonlinear benchmarks, the new algorithms consistently outperform existing single-solve explicit methods in accuracy. Their built-in numerical dissipation effectively filters out spurious high-frequency components, as demonstrated by two wave propagation problems. Finally, when applied to the realistic engineering problem, both of them deliver superior numerical precision at minimal computational cost. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 03:06:24 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Yaokun",
"Liu",
""
],
[
"Jinze",
"Li",
""
],
[
"Kaiping",
"Yu",
""
]
] |
2506.03520 | Han Zhang | Han Zhang, KaWing Tsang, Zhenhui Peng | VChatter: Exploring Generative Conversational Agents for Simulating Exposure Therapy to Reduce Social Anxiety | null | null | null | null | cs.HC | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Many people struggle with social anxiety, feeling fear, or even physically uncomfortable in social situations like talking to strangers. Exposure therapy, a clinical method that gradually and repeatedly exposes individuals to the source of their fear and helps them build coping mechanisms, can reduce social anxiety but traditionally requires human therapists' guidance and constructions of situations. In this paper, we developed a multi-agent system VChatter to explore large language models(LLMs)-based conversational agents for simulating exposure therapy with users. Based on a survey study (N=36) and an expert interview, VChatter includes an Agent-P, which acts as a psychotherapist to design the exposure therapy plans for users, and two Agent-Hs, which can take on different interactive roles in low, medium, and high exposure scenarios. A six-day qualitative study (N=10) showcases VChatter's usefulness in reducing users' social anxiety, feelings of isolation, and avoidance of social interactions. We demonstrated the feasibility of using LLMs-based conversational agents to simulate exposure therapy for addressing social anxiety and discussed future concerns for designing agents tailored to social anxiety. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 03:11:25 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Zhang",
"Han",
""
],
[
"Tsang",
"KaWing",
""
],
[
"Peng",
"Zhenhui",
""
]
] |
2506.03521 | Weinan He | Weinan He, Zilei Wang, Yixin Zhang | Target Semantics Clustering via Text Representations for Robust Universal Domain Adaptation | Camera-ready version for AAAI 2025 | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Universal Domain Adaptation (UniDA) focuses on transferring source domain knowledge to the target domain under both domain shift and unknown category shift. Its main challenge lies in identifying common class samples and aligning them. Current methods typically obtain target domain semantics centers from an unconstrained continuous image representation space. Due to domain shift and the unknown number of clusters, these centers often result in complex and less robust alignment algorithm. In this paper, based on vision-language models, we search for semantic centers in a semantically meaningful and discrete text representation space. The constrained space ensures almost no domain bias and appropriate semantic granularity for these centers, enabling a simple and robust adaptation algorithm. Specifically, we propose TArget Semantics Clustering (TASC) via Text Representations, which leverages information maximization as a unified objective and involves two stages. First, with the frozen encoders, a greedy search-based framework is used to search for an optimal set of text embeddings to represent target semantics. Second, with the search results fixed, encoders are refined based on gradient descent, simultaneously achieving robust domain alignment and private class clustering. Additionally, we propose Universal Maximum Similarity (UniMS), a scoring function tailored for detecting open-set samples in UniDA. Experimentally, we evaluate the universality of UniDA algorithms under four category shift scenarios. Extensive experiments on four benchmarks demonstrate the effectiveness and robustness of our method, which has achieved state-of-the-art performance. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 03:11:53 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"He",
"Weinan",
""
],
[
"Wang",
"Zilei",
""
],
[
"Zhang",
"Yixin",
""
]
] |
2506.03522 | Daniel Campa | Daniel Campa, Mehdi Saeedi, Ian Colbert, Srinjoy Das | Path Generation and Evaluation in Video Games: A Nonparametric Statistical Approach | 8 pages, 9 figures, Accepted at the IEEE Conference on Games 2025 (IEEE CoG) | null | null | null | cs.LG stat.ML | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Navigation path traces play a crucial role in video game design, serving as a vital resource for both enhancing player engagement and fine-tuning non-playable character behavior. Generating such paths with human-like realism can enrich the overall gaming experience, and evaluating path traces can provide game designers insights into player interactions. Despite the impressive recent advancements in deep learning-based generative modeling, the video game industry hesitates to adopt such models for path generation, often citing their complex training requirements and interpretability challenges. To address these problems, we propose a novel path generation and evaluation approach that is grounded in principled nonparametric statistics and provides precise control while offering interpretable insights. Our path generation method fuses two statistical techniques: (1) nonparametric model-free transformations that capture statistical characteristics of path traces through time; and (2) copula models that capture statistical dependencies in space. For path evaluation, we adapt a nonparametric three-sample hypothesis test designed to determine if the generated paths are overfit (mimicking the original data too closely) or underfit (diverging too far from it). We demonstrate the precision and reliability of our proposed methods with empirical analysis on two existing gaming benchmarks to showcase controlled generation of diverse navigation paths. Notably, our novel path generator can be fine-tuned with user controllable parameters to create navigation paths that exhibit varying levels of human-likeness in contrast to those produced by neural network-based agents. The code is available at https://github.com/daniel-campa/mf-copula. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 03:14:30 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Campa",
"Daniel",
""
],
[
"Saeedi",
"Mehdi",
""
],
[
"Colbert",
"Ian",
""
],
[
"Das",
"Srinjoy",
""
]
] |
2506.03523 | Chong Li | Chong Li, Jiajun Zhang, Chengqing Zong | TokAlign: Efficient Vocabulary Adaptation via Token Alignment | ACL 2025, our codes and models are available at https://github.com/ZNLP/TokAlign | null | null | null | cs.CL | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Tokenization serves as a foundational step for Large Language Models (LLMs) to process text. In new domains or languages, the inefficiency of the tokenizer will slow down the training and generation of LLM. The mismatch in vocabulary also hinders deep knowledge transfer between LLMs like token-level distillation. To mitigate this gap, we propose an efficient method named TokAlign to replace the vocabulary of LLM from the token co-occurrences view, and further transfer the token-level knowledge between models. It first aligns the source vocabulary to the target one by learning a one-to-one mapping matrix for token IDs. Model parameters, including embeddings, are rearranged and progressively fine-tuned for the new vocabulary. Our method significantly improves multilingual text compression rates and vocabulary initialization for LLMs, decreasing the perplexity from 3.4$\text{e}^2$ of strong baseline methods to 1.2$\text{e}^2$ after initialization. Experimental results on models across multiple parameter scales demonstrate the effectiveness and generalization of TokAlign, which costs as few as 5k steps to restore the performance of the vanilla model. After unifying vocabularies between LLMs, token-level distillation can remarkably boost (+4.4% than sentence-level distillation) the base model, costing only 235M tokens. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 03:15:57 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Li",
"Chong",
""
],
[
"Zhang",
"Jiajun",
""
],
[
"Zong",
"Chengqing",
""
]
] |
2506.03525 | Daeun Lee | Daeun Lee, Jaehong Yoon, Jaemin Cho, Mohit Bansal | Video-Skill-CoT: Skill-based Chain-of-Thoughts for Domain-Adaptive Video Reasoning | Project website: https://video-skill-cot.github.io/ | null | null | null | cs.CV cs.AI cs.CL | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Recent advances in Chain-of-Thought (CoT) reasoning have improved complex video understanding, but existing methods often struggle to adapt to domain-specific skills (e.g., event detection, spatial relation understanding, emotion understanding) over various video content. To address this, we propose Video-Skill-CoT (a.k.a. Video-SKoT), a framework that automatically constructs and leverages skill-aware CoT supervisions for domain-adaptive video reasoning. First, we construct skill-based CoT annotations: we extract domain-relevant reasoning skills from training questions, cluster them into a shared skill taxonomy, and create detailed multi-step CoT rationale tailored to each video-question pair for training. Second, we introduce a skill-specific expert learning framework. Each expert module specializes in a subset of reasoning skills and is trained with lightweight adapters using the collected CoT supervision. We demonstrate the effectiveness of the proposed approach on three video understanding benchmarks, where Video-SKoT consistently outperforms strong baselines. We also provide in-depth analyses on comparing different CoT annotation pipelines and learned skills over multiple video domains. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 03:18:01 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Lee",
"Daeun",
""
],
[
"Yoon",
"Jaehong",
""
],
[
"Cho",
"Jaemin",
""
],
[
"Bansal",
"Mohit",
""
]
] |
2506.03526 | Dakang Cen | Dakang Cen, Wenlong Zhang, and Junbin Zhong | A randomized progressive iterative regularization method for data fitting problems | 28 pages,31 figures | null | null | null | math.NA cs.NA | http://creativecommons.org/licenses/by/4.0/ | In this work, we investigate data fitting problems with random noises. A randomized progressive iterative regularization method is proposed. It works well for large-scale matrix computations and converges in expectation to the least-squares solution. Furthermore, we present an optimal estimation for the regularization parameter, which inspires the construction of self-consistent algorithms without prior information. The numerical results confirm the theoretical analysis and show the performance in curve and surface fittings. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 03:18:38 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Cen",
"Dakang",
""
],
[
"Zhang",
"Wenlong",
""
],
[
"Zhong",
"Junbin",
""
]
] |
2506.03527 | Feng Xiao | Lu Li, Yun Wan, Feng Xiao | Distinguishing True Influence from Hyperprolificity with Citation Distance | null | null | null | null | cs.DL | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Accurately evaluating scholarly influence is essential for fair academic assessment, yet traditional bibliometric indicators - dominated by publication and citation counts - often favor hyperprolific authors over those with deeper, long-term impact. We propose the x-index, a novel citation-based metric that conceptualizes citation as a process of knowledge diffusion and incorporates citation distance to reflect the structural reach of scholarly work. By weighting citations according to the collaborative proximity between citing and cited authors, the x-index captures both the depth and breadth of influence within evolving academic networks. Empirical analyses show that the x-index significantly improves the rankings of Turing Award recipients while reducing those of hyperprolific authors, better aligning rankings with recognized academic merit. It also demonstrates superior discriminatory power among early-career researchers and reveals stronger sensitivity to institutional research quality. These results suggest that the x-index offers a more equitable and forward-looking alternative to existing metrics, with practical applications in talent identification, funding decisions, and academic recommendation systems. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 03:19:11 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Li",
"Lu",
""
],
[
"Wan",
"Yun",
""
],
[
"Xiao",
"Feng",
""
]
] |
2506.03528 | Soumen Banerjee | Soumen Banerjee and Yi-Chun Chen and Yifei Sun | Correlated equilibrium implementation: Navigating toward social optima with learning dynamics | null | null | null | null | econ.TH | http://creativecommons.org/licenses/by/4.0/ | Implementation theory has made significant advances in characterizing which social choice functions can be implemented in Nash equilibrium, but these results typically assume sophisticated strategic reasoning by agents. However, evidence exists to show that agents frequently cannot perform such reasoning. In this paper, we present a finite mechanism which fully implements Maskin-monotonic social choice functions as the outcome of the unique correlated equilibrium of the induced game. Due to the results in Hart and MasColell (2000), this yields that even when agents use a simple adaptive heuristic like regret minimization rather than computing equilibrium strategies, the designer can expect to implement the SCF correctly. We demonstrate the mechanism's effectiveness through simulations in a bilateral trade environment, where agents using regret matching converge to the desired outcomes despite having no knowledge of others' preferences or the equilibrium structure. The mechanism does not use integer games or modulo games. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 03:19:49 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Banerjee",
"Soumen",
""
],
[
"Chen",
"Yi-Chun",
""
],
[
"Sun",
"Yifei",
""
]
] |
2506.03529 | Lei Sun | Zhecheng Sun, Weibin Ni, Denan Li, Xiya Du, Shi Liu, Lei Sun | Ultralong Room-Temperature Qubit Lifetimes of Covalent Organic Frameworks | 21 pages, 5 figures | null | null | null | quant-ph cond-mat.mes-hall physics.chem-ph | http://creativecommons.org/licenses/by-nc-nd/4.0/ | Molecular electron spin qubits offer atomic-level tunability and room-temperature quantum coherence. Their integration into engineered solid-state matrices can enhance performance towards ambient quantum information technologies. Herein, we demonstrate covalent organic frameworks (COFs) as programmable matrices of stable organic radical qubits allowing strategic optimization of spin-phonon and spin-spin interactions. Using two classic boronate-ester frameworks, COF-5 and COF-108, to host semiquinone-like radical qubits, we achieve ultralong spin relaxation time (T1 > 300 {\mu}s) at 298 K, which outperforms most molecular qubits and rivals inorganic spin defects. The suppression of spin relaxation is attributed to rigid and neutral structures as well as carbon-centered spin distributions that effectively weaken spin-phonon coupling. Employing dynamical decoupling methods to both COFs improves their quantum coherence and enables room-temperature detection of nuclear spins including 1H, 11B, and 13C. Our work establishes COFs as designer quantum materials, opening new avenues for quantum sensing of nuclear spins at room temperature. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 03:20:30 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Sun",
"Zhecheng",
""
],
[
"Ni",
"Weibin",
""
],
[
"Li",
"Denan",
""
],
[
"Du",
"Xiya",
""
],
[
"Liu",
"Shi",
""
],
[
"Sun",
"Lei",
""
]
] |
2506.03530 | Guanzhou Ke | Guanzhou Ke, Yi Xie, Xiaoli Wang, Guoqing Chao, Bo Wang, Shengfeng He | How Far Are We from Predicting Missing Modalities with Foundation Models? | null | null | null | null | cs.MM cs.CL cs.CV | http://creativecommons.org/licenses/by-nc-sa/4.0/ | Multimodal foundation models have demonstrated impressive capabilities across diverse tasks. However, their potential as plug-and-play solutions for missing modality prediction remains underexplored. To investigate this, we categorize existing approaches into three representative paradigms, encompassing a total of 42 model variants, and conduct a comprehensive evaluation in terms of prediction accuracy and adaptability to downstream tasks. Our analysis reveals that current foundation models often fall short in two critical aspects: (i) fine-grained semantic extraction from the available modalities, and (ii) robust validation of generated modalities. These limitations lead to suboptimal and, at times, misaligned predictions. To address these challenges, we propose an agentic framework tailored for missing modality prediction. This framework dynamically formulates modality-aware mining strategies based on the input context, facilitating the extraction of richer and more discriminative semantic features. In addition, we introduce a \textit{self-refinement mechanism}, which iteratively verifies and enhances the quality of generated modalities through internal feedback. Experimental results show that our method reduces FID for missing image prediction by at least 14% and MER for missing text prediction by at least 10% compared to baselines. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 03:22:44 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Ke",
"Guanzhou",
""
],
[
"Xie",
"Yi",
""
],
[
"Wang",
"Xiaoli",
""
],
[
"Chao",
"Guoqing",
""
],
[
"Wang",
"Bo",
""
],
[
"He",
"Shengfeng",
""
]
] |
2506.03531 | Daniel Ovalle | Daniel Ovalle, Lorenz T. Biegler, Ignacio E. Grossmann, Carl D. Laird, Mateo Dulce Rubio | Conformal Mixed-Integer Constraint Learning with Feasibility Guarantees | null | null | null | null | cs.LG math.OC | http://creativecommons.org/licenses/by/4.0/ | We propose Conformal Mixed-Integer Constraint Learning (C-MICL), a novel framework that provides probabilistic feasibility guarantees for data-driven constraints in optimization problems. While standard Mixed-Integer Constraint Learning methods often violate the true constraints due to model error or data limitations, our C-MICL approach leverages conformal prediction to ensure feasible solutions are ground-truth feasible. This guarantee holds with probability at least $1{-}\alpha$, under a conditional independence assumption. The proposed framework supports both regression and classification tasks without requiring access to the true constraint function, while avoiding the scalability issues associated with ensemble-based heuristics. Experiments on real-world applications demonstrate that C-MICL consistently achieves target feasibility rates, maintains competitive objective performance, and significantly reduces computational cost compared to existing methods. Our work bridges mathematical optimization and machine learning, offering a principled approach to incorporate uncertainty-aware constraints into decision-making with rigorous statistical guarantees. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 03:26:31 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Ovalle",
"Daniel",
""
],
[
"Biegler",
"Lorenz T.",
""
],
[
"Grossmann",
"Ignacio E.",
""
],
[
"Laird",
"Carl D.",
""
],
[
"Rubio",
"Mateo Dulce",
""
]
] |
2506.03532 | YunYao Zhang | Yunyao Zhang, Zikai Song, Hang Zhou, Wenfeng Ren, Yi-Ping Phoebe Chen, Junqing Yu, Wei Yang | GA-S$^3$: Comprehensive Social Network Simulation with Group Agents | Accepted by Findings of ACL 2025 | null | null | null | cs.SI cs.CY | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Social network simulation is developed to provide a comprehensive understanding of social networks in the real world, which can be leveraged for a wide range of applications such as group behavior emergence, policy optimization, and business strategy development. However, billions of individuals and their evolving interactions involved in social networks pose challenges in accurately reflecting real-world complexities. In this study, we propose a comprehensive Social Network Simulation System (GA-S3) that leverages newly designed Group Agents to make intelligent decisions regarding various online events. Unlike other intelligent agents that represent an individual entity, our group agents model a collection of individuals exhibiting similar behaviors, facilitating the simulation of large-scale network phenomena with complex interactions at a manageable computational cost. Additionally, we have constructed a social network benchmark from 2024 popular online events that contains fine-grained information on Internet traffic variations. The experiment demonstrates that our approach is capable of achieving accurate and highly realistic prediction results. Code is open at https://github.com/AI4SS/GAS-3. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 03:27:05 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Zhang",
"Yunyao",
""
],
[
"Song",
"Zikai",
""
],
[
"Zhou",
"Hang",
""
],
[
"Ren",
"Wenfeng",
""
],
[
"Chen",
"Yi-Ping Phoebe",
""
],
[
"Yu",
"Junqing",
""
],
[
"Yang",
"Wei",
""
]
] |
2506.03533 | Apurva Gandhi | Apurva Gandhi and Graham Neubig | Go-Browse: Training Web Agents with Structured Exploration | null | null | null | null | cs.CL | http://creativecommons.org/licenses/by/4.0/ | One of the fundamental problems in digital agents is their lack of understanding of their environment. For instance, a web browsing agent may get lost in unfamiliar websites, uncertain what pages must be visited to achieve its goals. To address this, we propose Go-Browse, a method for automatically collecting diverse and realistic web agent data at scale through structured exploration of web environments. Go-Browse achieves efficient exploration by framing data collection as a graph search, enabling reuse of information across exploration episodes. We instantiate our method on the WebArena benchmark, collecting a dataset of 10K successful task-solving trajectories and 40K interaction steps across 100 URLs. Fine-tuning a 7B parameter language model on this dataset achieves a success rate of 21.7% on the WebArena benchmark, beating GPT-4o mini by 2.4% and exceeding current state-of-the-art results for sub-10B parameter models by 2.9%. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 03:27:56 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Gandhi",
"Apurva",
""
],
[
"Neubig",
"Graham",
""
]
] |
2506.03534 | Zheng Xiao | Julie Tzu-Yueh Wang, Zheng Xiao | Hyperbolicity and GCD for n+1 divisors with non-empty intersection | 21 pages, any comments are welcome | null | null | null | math.CV math.NT | http://creativecommons.org/licenses/by/4.0/ | We study hyperbolicity for quasi-projective varieties where the boundary divisor consists of n+1 numerically parallel effective divisors on a complex projective variety of dimension n, allowing non-empty intersection. Under explicit local conditions on beta constants or intersection multiplicities, we prove that all entire curves are algebraically degenerate. Our approach extends the method of Levin-Huang-Xiao to higher dimensions, establishing a second main theorem for regular sequences of closed subschemes. This also yields a GCD-type estimate in the same geometric setting. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 03:29:52 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Wang",
"Julie Tzu-Yueh",
""
],
[
"Xiao",
"Zheng",
""
]
] |
2506.03535 | Qiming Zhu | Qiming Zhu, Jialun Cao, Xuanang Chen, Yaojie Lu, Hongyu Lin, Xianpei Han, Le Sun, Shing-Chi Cheung | Across Programming Language Silos: A Study on Cross-Lingual Retrieval-augmented Code Generation | null | null | null | null | cs.SE | http://creativecommons.org/licenses/by/4.0/ | Current research on large language models (LLMs) with retrieval-augmented code generation (RACG) mainly focuses on single-language settings, leaving cross-lingual effectiveness and security unexplored. Multi-lingual RACG systems are valuable for migrating code-bases across programming languages (PLs), yet face risks from error (e.g. adversarial data corruption) propagation in cross-lingual transfer. We construct a dataset spanning 13 PLs with nearly 14k instances to explore utility and robustness of multi-lingual RACG systems. Our investigation reveals four key insights: (1) Effectiveness: multi-lingual RACG significantly enhances multi-lingual code LLMs generation; (2) Inequality: Java demonstrate superior cross-lingual utility over Python in RACG; (3) Robustness: Adversarial attacks degrade performance significantly in mono-lingual RACG but show mitigated impacts in cross-lingual scenarios; Counterintuitively, perturbed code may improve RACG in cross-lingual scenarios; (4) Specialization: Domain-specific code retrievers outperform significantly general text retrievers. These findings establish foundation for developing effective and secure multi-lingual code assistants. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 03:31:00 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Zhu",
"Qiming",
""
],
[
"Cao",
"Jialun",
""
],
[
"Chen",
"Xuanang",
""
],
[
"Lu",
"Yaojie",
""
],
[
"Lin",
"Hongyu",
""
],
[
"Han",
"Xianpei",
""
],
[
"Sun",
"Le",
""
],
[
"Cheung",
"Shing-Chi",
""
]
] |
2506.03536 | Yang Liu | Yang Liu, Yuchen Zhang, Hongwei Yu, Puxun Wu | Constraining the Baryon Fraction in the Intergalactic Medium with 92 localized Fast Radio Bursts | 13 pages, 2 figures, 1 table | null | null | null | astro-ph.CO gr-qc | http://creativecommons.org/licenses/by/4.0/ | Fast radio bursts (FRBs) are emerging as powerful cosmological probes for constraining the baryon fraction in the intergalactic medium (IGM), offering a promising avenue to address the missing baryon problem. In this paper, we analyze constraints on the IGM baryon fraction ($f_\mathrm{IGM}$) using 92 localized FRBs, incorporating a corrected probability distribution function for the IGM dispersion measure within three different cosmological models. We find that variations in the underlying cosmological model have a negligible impact on the inferred values of $f_\mathrm{IGM}$. While the NE2001 Galactic electron density model yields slightly higher $f_\mathrm{IGM}$ values compared to the YMW16 model, the results are consistent within the 1$\sigma$ confidence level. Additionally, there is no statistically significant evidence for redshift evolution in $f_\mathrm{IGM}$. Our analysis constrains $f_\mathrm{IGM}$ to the range $0.8 \sim 0.9$, providing strong support for the idea in which the majority of the missing baryons reside in the diffuse IGM. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 03:34:29 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Liu",
"Yang",
""
],
[
"Zhang",
"Yuchen",
""
],
[
"Yu",
"Hongwei",
""
],
[
"Wu",
"Puxun",
""
]
] |
2506.03537 | Daiki Niimi | Daiki Niimi, An Fujino, Taro Suzuki, and Junichi Meguro | Robust Position Estimation by Rao-Blackwellized Particle Filter without Integer Ambiguity Resolution in Urban Environments | Accepted to the 2025 IEEE/ION Position, Location and Navigation Symposium (PLANS) | null | null | null | cs.RO | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | This study proposes a centimeter-accurate positioning method that utilizes a Rao-Blackwellized particle filter (RBPF) without requiring integer ambiguity resolution in global navigation satellite system (GNSS) carrier phase measurements. The conventional positioning method employing a particle filter (PF) eliminates the necessity for ambiguity resolution by calculating the likelihood from the residuals of the carrier phase based on the particle position. However, this method encounters challenges, particularly in urban environments characterized by non-line-of-sight (NLOS) multipath errors. In such scenarios, PF tracking may fail due to the degradation of velocity estimation accuracy used for state transitions, thereby complicating subsequent position estimation. To address this issue, we apply Rao-Blackwellization to the conventional PF framework, treating position and velocity as distinct states and employing the Kalman filter for velocity estimation. This approach enhances the accuracy of velocity estimation and, consequently, the precision of position estimation. Moreover, the proposed method rejects NLOS multipath signals based on the pseudorange residuals at each particle position during the velocity estimation step. This process not only enhances velocity accuracy, but also preserves particle diversity by allowing particles to transition to unique states with varying velocities. Consequently, particles are more likely to cluster around the true position, thereby enabling more accurate position estimation. Vehicular experiments in urban environments demonstrated the effectiveness of proposed method in achieving a higher positioning accuracy than conventional PF-based and conventional GNSS positioning methods. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 03:34:38 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Niimi",
"Daiki",
""
],
[
"Fujino",
"An",
""
],
[
"Suzuki",
"Taro",
""
],
[
"Meguro",
"Junichi",
""
]
] |
2506.03538 | Chengqi Li | Chengqi Li, Zhihao Shi, Yangdi Lu, Wenbo He, Xiangyu Xu | Robust Neural Rendering in the Wild with Asymmetric Dual 3D Gaussian Splatting | null | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | 3D reconstruction from in-the-wild images remains a challenging task due to inconsistent lighting conditions and transient distractors. Existing methods typically rely on heuristic strategies to handle the low-quality training data, which often struggle to produce stable and consistent reconstructions, frequently resulting in visual artifacts. In this work, we propose Asymmetric Dual 3DGS, a novel framework that leverages the stochastic nature of these artifacts: they tend to vary across different training runs due to minor randomness. Specifically, our method trains two 3D Gaussian Splatting (3DGS) models in parallel, enforcing a consistency constraint that encourages convergence on reliable scene geometry while suppressing inconsistent artifacts. To prevent the two models from collapsing into similar failure modes due to confirmation bias, we introduce a divergent masking strategy that applies two complementary masks: a multi-cue adaptive mask and a self-supervised soft mask, which leads to an asymmetric training process of the two models, reducing shared error modes. In addition, to improve the efficiency of model training, we introduce a lightweight variant called Dynamic EMA Proxy, which replaces one of the two models with a dynamically updated Exponential Moving Average (EMA) proxy, and employs an alternating masking strategy to preserve divergence. Extensive experiments on challenging real-world datasets demonstrate that our method consistently outperforms existing approaches while achieving high efficiency. Codes and trained models will be released. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 03:40:33 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Li",
"Chengqi",
""
],
[
"Shi",
"Zhihao",
""
],
[
"Lu",
"Yangdi",
""
],
[
"He",
"Wenbo",
""
],
[
"Xu",
"Xiangyu",
""
]
] |
2506.03539 | Shin-Ted Lin | Chang-Hao Fang, Yi-Ke Shu, Shin-Ted Lin, Shu-Kui Liu, Hao-Yang Xing, Jing-Jun Zhu, Hsin-Chang Chi, Muhammed Deniz, Hai-Tao Jia, Han-Yu Li, Qian-Yun Li, Ren-Ming-Jie Li, Yu Liu, Xiao-Yu Peng, Hao-Yu Shi, Qin Wang, Henry Tsz-King Wong, Yu-Lu Yan, Li-Tao Yang and Qian Yue | Germanium Atomic Compton Scattering Measurements and ${ab}$ ${initio}$ Many-Body Calculations: Implications for Electronic recoil Dark Matter Detection | 16 pages, 12 figures | null | null | null | astro-ph.IM | http://creativecommons.org/publicdomain/zero/1.0/ | Diverse searches for direct dark matter (DM) in effective electromagnetic and leptophilic interactions resulting from new physics, as well as Weakly Interacting Massive Particles (WIMPs) with unconventional electronic recoils, are intensively pursued. Low-energy backgrounds from radioactive $\gamma$ rays via Compton scattering and photon coherent scattering are unavoidable in terrestrial detectors. The interpretation of dark matter experimental data is dependent on a better knowledge of the background in the low-energy region. We provide a 2.3% measurement of atomic Compton scattering in the low momentum transfer range of 180 eV/c to 25 keV/c, using a 10-g germanium detector bombarded by a $^{137}\mathrm{Cs}$ source with a 7.2 m-Curie radioactivity and the scatter photon collected by a cylindrical NaI[Tl] detector. The ability to detect Compton scattering's doubly differential cross section (DDCS) gives a special test for clearly identifying the kinematic restraints in atomic many-body systems, notably the Livermore model. Additionally, a low-energy-background comparison is made between coherent photon scattering and Compton scattering replacing the scattering function of ${GEANT4}$@software, which uses a completely relativistic impulse approximation (RIA) together with Multi-Configuration Dirac-Fock (MCDF) wavefunctions. For the purpose of investigating sub-GeV mass and electronic-recoil dark matter theories, signatures including low energy backgrounds via high energy $\gamma$ rays in germanium targets are discussed. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 03:43:50 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Fang",
"Chang-Hao",
""
],
[
"Shu",
"Yi-Ke",
""
],
[
"Lin",
"Shin-Ted",
""
],
[
"Liu",
"Shu-Kui",
""
],
[
"Xing",
"Hao-Yang",
""
],
[
"Zhu",
"Jing-Jun",
""
],
[
"Chi",
"Hsin-Chang",
""
],
[
"Deniz",
"Muhammed",
""
],
[
"Jia",
"Hai-Tao",
""
],
[
"Li",
"Han-Yu",
""
],
[
"Li",
"Qian-Yun",
""
],
[
"Li",
"Ren-Ming-Jie",
""
],
[
"Liu",
"Yu",
""
],
[
"Peng",
"Xiao-Yu",
""
],
[
"Shi",
"Hao-Yu",
""
],
[
"Wang",
"Qin",
""
],
[
"Wong",
"Henry Tsz-King",
""
],
[
"Yan",
"Yu-Lu",
""
],
[
"Yang",
"Li-Tao",
""
],
[
"Yue",
"Qian",
""
]
] |
2506.03540 | Tek Prasad Adhikari Dr. | Tek P. Adhikari, Santanu Mondal, Zhicheng He, Agata Rozanska and Krzysztof Hryniewicz | Changing Look AGN: A study of Optical/UV and the Highly Ionized Fe K$\alpha$ X-ray Line Flux Variations Using Photo-Ionization Simulations | 22 pages, 23 figures, Accepted on 24-May-2025 for publication in The Astrophysical Journal | null | null | null | astro-ph.GA astro-ph.HE | http://creativecommons.org/licenses/by/4.0/ | Significant variability in broad emission line strengths of active galactic nuclei (AGN) over months to years has been observed, often accompanied by intrinsic continuum changes. Such spectral variability challenges the traditional AGN classification scheme, which attributes differences between Type 1 and Type 2 to geometrical effects, as transitions between these types occur on timescales shorter than viscous ones. In this work, using the {\sc cloudy} photo-ionization simulations, we investigated the response of the major emission line fluxes, in the optical/UV and hard X-ray bands, to changes in the intensity and shape of the continuum emission of the AGN under two scenarios: (i) changes in the X-ray power-law while keeping disc emission fixed, and (ii) broadband continuum variations. We demonstrate that BLR line fluxes are insensitive to X-ray power-law changes alone. Considering a well-studied case of the changing-look (CL) AGN Mrk 1018, which exhibits variations in the intrinsic disc emission, as well as the X-ray power-law, our simulations reproduce observed brightening and dimming trends of the BLR emission. Moreover, we show that the highly ionized Fe K$\alpha$ X-ray flux, primarily produced by the H-like and He-like ions of Fe, strongly depends on the X-ray strength of the intrinsic SED. These findings suggest that the origin of highly ionized Fe K$\alpha$ emission is in the coronal part of the accretion disk and that the CL phenomenon can be triggered by intrinsic changes in the accretion properties of AGN. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 03:49:43 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Adhikari",
"Tek P.",
""
],
[
"Mondal",
"Santanu",
""
],
[
"He",
"Zhicheng",
""
],
[
"Rozanska",
"Agata",
""
],
[
"Hryniewicz",
"Krzysztof",
""
]
] |
2506.03541 | Xiaofeng Zhou | Xiaofeng Zhou, Heyan Huang, Lizi Liao | Debate, Reflect, and Distill: Multi-Agent Feedback with Tree-Structured Preference Optimization for Efficient Language Model Enhancement | 16 pages, 10 figures. The camera-ready paper for Findings of ACL 2025 | null | null | null | cs.CL cs.AI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Large Language Models (LLMs) continue to set new standards in knowledge-intensive and complex reasoning tasks, yet their high computational demands limit widespread adoption. While distilling large models into smaller ones offers a sustainable solution, current techniques--such as static knowledge distillation, resource-intensive reinforcement learning from human feedback, or limited self-reflection--struggle to yield substantial and lasting performance gains. In this paper, we present a novel Debate and Reflect (D&R) framework that orchestrates multi-turn debates between smaller models and stronger teacher models, eliciting actionable feedback (e.g., error analysis, corrective strategies) to guide student models. Further, we introduce Tree-structured Direct Preference Optimization (T-DPO) to efficiently leverage these debate logs, organizing interactions into a hierarchical format for effective training. Empirical evaluations across diverse NLP benchmarks demonstrate that our approach significantly improves smaller-model accuracy, robustness, and generalization, outperforming conventional baselines by a large margin. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 03:52:20 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Zhou",
"Xiaofeng",
""
],
[
"Huang",
"Heyan",
""
],
[
"Liao",
"Lizi",
""
]
] |
2506.03542 | Yongxiang Tang | Yongxiang Tang, Yanhua Cheng, Xiaocheng Liu, Chenchen Jiao, Yanxiang Zeng, Ning Luo, Pengjia Yuan, Xialong Liu, Peng Jiang | Learning Monotonic Probabilities with a Generative Cost Model | null | null | null | null | cs.LG | http://creativecommons.org/licenses/by-sa/4.0/ | In many machine learning tasks, it is often necessary for the relationship between input and output variables to be monotonic, including both strictly monotonic and implicitly monotonic relationships. Traditional methods for maintaining monotonicity mainly rely on construction or regularization techniques, whereas this paper shows that the issue of strict monotonic probability can be viewed as a partial order between an observable revenue variable and a latent cost variable. This perspective enables us to reformulate the monotonicity challenge into modeling the latent cost variable. To tackle this, we introduce a generative network for the latent cost variable, termed the Generative Cost Model (GCM), which inherently addresses the strict monotonic problem, and propose the Implicit Generative Cost Model (IGCM) to address the implicit monotonic problem. We further validate our approach with a numerical simulation of quantile regression and conduct multiple experiments on public datasets, showing that our method significantly outperforms existing monotonic modeling techniques. The code for our experiments can be found at https://github.com/tyxaaron/GCM. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 03:54:26 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Tang",
"Yongxiang",
""
],
[
"Cheng",
"Yanhua",
""
],
[
"Liu",
"Xiaocheng",
""
],
[
"Jiao",
"Chenchen",
""
],
[
"Zeng",
"Yanxiang",
""
],
[
"Luo",
"Ning",
""
],
[
"Yuan",
"Pengjia",
""
],
[
"Liu",
"Xialong",
""
],
[
"Jiang",
"Peng",
""
]
] |
2506.03543 | Wanghao (Wanghao) Ye | Wanghao Ye, Sihan Chen, Yiting Wang, Shwai He, Bowei Tian, Guoheng Sun, Ziyi Wang, Ziyao Wang, Yexiao He, Zheyu Shen, Meng Liu, Yuning Zhang, Meng Feng, Yang Wang, Siyuan Peng, Yilong Dai, Zhenle Duan, Hanzhang Qin, and Ang Li | CogniPair: From LLM Chatbots to Conscious AI Agents -- GNWT-Based Multi-Agent Digital Twins for Social Pairing -- Dating & Hiring Applications | null | null | null | null | cs.AI cs.CY cs.MA | http://creativecommons.org/licenses/by/4.0/ | Current large language model (LLM) agents lack authentic human psychological processes necessary for genuine digital twins and social AI applications. To address this limitation, we present a computational implementation of Global Workspace Theory (GNWT) that integrates human cognitive architecture principles into LLM agents, creating specialized sub-agents for emotion, memory, social norms, planning, and goal-tracking coordinated through a global workspace mechanism. However, authentic digital twins require accurate personality initialization. We therefore develop a novel adventure-based personality test that evaluates true personality through behavioral choices within interactive scenarios, bypassing self-presentation bias found in traditional assessments. Building on these innovations, our CogniPair platform enables digital twins to engage in realistic simulated dating interactions and job interviews before real encounters, providing bidirectional cultural fit assessment for both romantic compatibility and workplace matching. Validation using 551 GNWT-Agents and Columbia University Speed Dating dataset demonstrates 72% correlation with human attraction patterns, 77.8% match prediction accuracy, and 74% agreement in human validation studies. This work advances psychological authenticity in LLM agents and establishes a foundation for intelligent dating platforms and HR technology solutions. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 03:54:30 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Ye",
"Wanghao",
""
],
[
"Chen",
"Sihan",
""
],
[
"Wang",
"Yiting",
""
],
[
"He",
"Shwai",
""
],
[
"Tian",
"Bowei",
""
],
[
"Sun",
"Guoheng",
""
],
[
"Wang",
"Ziyi",
""
],
[
"Wang",
"Ziyao",
""
],
[
"He",
"Yexiao",
""
],
[
"Shen",
"Zheyu",
""
],
[
"Liu",
"Meng",
""
],
[
"Zhang",
"Yuning",
""
],
[
"Feng",
"Meng",
""
],
[
"Wang",
"Yang",
""
],
[
"Peng",
"Siyuan",
""
],
[
"Dai",
"Yilong",
""
],
[
"Duan",
"Zhenle",
""
],
[
"Qin",
"Hanzhang",
""
],
[
"Li",
"Ang",
""
]
] |
2506.03544 | Bruce Reed | Bruce Reed | The Global Structure of a Typical Graph Without $H$ as an Induced Subgraph when $H$ is a Cycle | null | null | null | null | math.CO | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | One way to certify that a graph does not contain an induced cycle of length six is to provide a partition of its vertex set into (i) a stable set, and (ii) a graph containing no stable set of size three and no induced matching of size two. We show that almost every graph which does not contain a cycle of length six as an induced subgraph has such a certificate. We obtain similar characterizations of the structure of almost all graphs which contain no induced cycle of length $k$ for all even $k$ exceeding six. (Similar results were obtained for $k=3$ by Erdos, Kleitman, and Rothschild in 1976, for $k =4,5$ by Promel and Steger in 1991 and for odd $k$ exceeding 5 by Balogh and Butterfield in 2009.) We prove that a simiiar theorem for all $H$ holds up to the deletion of a set of $o(|V(G)|)$ vertices and ask for which $H$ the characterization holds fully. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 04:03:05 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Reed",
"Bruce",
""
]
] |
2506.03545 | Hung Tran | Hung Tran and Kazuo Yamazaki | On Ricci Solitons with Isoparametric Potential Functions | comments welcome | null | null | null | math.DG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | This paper studies a complete gradient Ricci soliton with an isoparametric potential function. Our first theorem asserts that, for the steady case, there is a critical level set of codimension greater than one. This is consistent with construction of cohomogeneity one models with singular orbits. There is a partial result for the shrinking case. We also study a particular ansatz of popular interest and obtain asymptotic behaviors. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 04:04:33 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Tran",
"Hung",
""
],
[
"Yamazaki",
"Kazuo",
""
]
] |
2506.03546 | Yuanchen Bai | Yuanchen Bai, Zijian Ding, Angelique Taylor | From Virtual Agents to Robot Teams: A Multi-Robot Framework Evaluation in High-Stakes Healthcare Context | null | null | null | null | cs.RO cs.AI cs.MA | http://creativecommons.org/licenses/by-nc-sa/4.0/ | Advancements in generative models have enabled multi-agent systems (MAS) to perform complex virtual tasks such as writing and code generation, which do not generalize well to physical multi-agent robotic teams. Current frameworks often treat agents as conceptual task executors rather than physically embodied entities, and overlook critical real-world constraints such as spatial context, robotic capabilities (e.g., sensing and navigation). To probe this gap, we reconfigure and stress-test a hierarchical multi-agent robotic team built on the CrewAI framework in a simulated emergency department onboarding scenario. We identify five persistent failure modes: role misalignment; tool access violations; lack of in-time handling of failure reports; noncompliance with prescribed workflows; bypassing or false reporting of task completion. Based on this analysis, we propose three design guidelines emphasizing process transparency, proactive failure recovery, and contextual grounding. Our work informs the development of more resilient and robust multi-agent robotic systems (MARS), including opportunities to extend virtual multi-agent frameworks to the real world. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 04:05:38 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Bai",
"Yuanchen",
""
],
[
"Ding",
"Zijian",
""
],
[
"Taylor",
"Angelique",
""
]
] |
2506.03547 | Giulia Macario Dr. | Giulia Macario (1 and 6), Giuseppe Pupillo (2), Gianni Bernardi (2 and 3 and 4), Paola Di Ninni (1), Giovanni Comoretto (1), Pietro Bolli (1), Andrea Mattana (2), Jader Monari (2), Federico Perini (2), Marco Schiaffino (2), Marcin Sokolowski (5), Randall Wayth (5 and 6), Jishnu N. Thekkeppattu (5) ((1) INAF, Osservatorio Astrofisico di Arcetri, Florence, Italy, (2) INAF, Istituto di Radioastronomia, Bologna, Italy, (3) Department of Physics and Electronics, Rhodes University, Makhanda, South Africa, (4) SARAO, Cape Town, South Africa, (5) ICRAR, Curtin University, Bentley, WA, Australia, (6) SKAO, Kenisington, WA, Australia) | Intrinsic Cross Polarization Ratio maps from all-sky observations with the SKA-Low prototype station Aperture Array Verification System 2 | 16 pages, 4 figures, Accepted for publication in Journal of Astronomical Telescopes, Instruments, and Systems on June 4th, 2025 | null | null | null | astro-ph.IM | http://creativecommons.org/licenses/by/4.0/ | The low-frequency radio telescope of the Square Kilometre Array (SKA-Low), currently under construction in the remote Murchison shire in the Western Australia's outback, will observe the sky between 50 MHz and 350 MHz with unprecedented sensitivity and stringent requirements for polarization accuracy. In this work, we investigate the instrumental polarization purity of a SKA-Low prototype station by means of the Intrinsic Cross-Polarization Ratio (IXR) figure of merit. We derive all-sky experimental IXR maps using data from the Aperture Array Verification System 2 (AAVS2). The results are presented at three frequencies within the SKA-Low bandwidth (110, 160, and 230 MHz) with a quantitative comparison between observed and simulated all-sky IXR maps. Our findings show good agreement in IXR map distributions and promising consistency in their radial profiles, meeting SKA-Low's IXR specification overall. This study offers an empirical approach to verifying SKA-Low's polarization performance using all-sky observations from individual stations and will potentially support the telescope's early science commissioning phase. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 04:06:08 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Macario",
"Giulia",
"",
"1 and 6"
],
[
"Pupillo",
"Giuseppe",
"",
"INAF, Istituto di Radioastronomia, Bologna, Italy"
],
[
"Bernardi",
"Gianni",
"",
"2 and 3 and 4"
],
[
"Di Ninni",
"Paola",
"",
"INAF, Osservatorio Astrofisico di Arcetri, Florence, Italy"
],
[
"Comoretto",
"Giovanni",
"",
"INAF, Osservatorio Astrofisico di Arcetri, Florence, Italy"
],
[
"Bolli",
"Pietro",
"",
"INAF, Osservatorio Astrofisico di Arcetri, Florence, Italy"
],
[
"Mattana",
"Andrea",
"",
"INAF, Istituto di Radioastronomia, Bologna, Italy"
],
[
"Monari",
"Jader",
"",
"INAF, Istituto di Radioastronomia, Bologna, Italy"
],
[
"Perini",
"Federico",
"",
"INAF, Istituto di Radioastronomia, Bologna, Italy"
],
[
"Schiaffino",
"Marco",
"",
"INAF, Istituto di Radioastronomia, Bologna, Italy"
],
[
"Sokolowski",
"Marcin",
"",
"ICRAR, Curtin University, Bentley, WA, Australia"
],
[
"Wayth",
"Randall",
"",
"5 and 6"
],
[
"Thekkeppattu",
"Jishnu N.",
"",
"ICRAR, Curtin University, Bentley, WA, Australia"
]
] |
2506.03548 | Chenglong Ye | Chenglong Ye, Gang Xiong, Junyou Shang, Xingyuan Dai, Xiaoyan Gong, Yisheng Lv | SUMO-MCP: Leveraging the Model Context Protocol for Autonomous Traffic Simulation and Optimization | null | null | null | null | cs.AI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Traffic simulation tools, such as SUMO, are essential for urban mobility research. However, such tools remain challenging for users due to complex manual workflows involving network download, demand generation, simulation setup, and result analysis. In this paper, we introduce SUMO-MCP, a novel platform that not only wraps SUMO' s core utilities into a unified tool suite but also provides additional auxiliary utilities for common preprocessing and postprocessing tasks. Using SUMO-MCP, users can issue simple natural-language prompts to generate traffic scenarios from OpenStreetMap data, create demand from origin-destination matrices or random patterns, run batch simulations with multiple signal-control strategies, perform comparative analyses with automated reporting, and detect congestion for signal-timing optimization. Furthermore, the platform allows flexible custom workflows by dynamically combining exposed SUMO tools without additional coding. Experiments demonstrate that SUMO-MCP significantly makes traffic simulation more accessible and reliable for researchers. We will release code for SUMO-MCP at https://github.com/ycycycl/SUMO-MCP in the future. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 04:08:11 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Ye",
"Chenglong",
""
],
[
"Xiong",
"Gang",
""
],
[
"Shang",
"Junyou",
""
],
[
"Dai",
"Xingyuan",
""
],
[
"Gong",
"Xiaoyan",
""
],
[
"Lv",
"Yisheng",
""
]
] |
2506.03549 | Wen Yu Kon | Wen Yu Kon, Ignatius William Primaatmaja, Kaushik Chakraborty, and Charles Lim | Quantum Secure Key Exchange with Position-based Credentials | 27 pages, 4 figures | null | null | null | quant-ph cs.CR | http://creativecommons.org/licenses/by/4.0/ | Quantum key distribution (QKD) provides an information-theoretic way of securely exchanging secret keys, and typically relies on pre-shared keys or public keys for message authentication. To lift the requirement of pre-shared or public keys, Buhrman et. al. [SIAM J. Comput. 43, 150 (2014)] proposed utilizing the location of a party as a credential. Here, we extend upon the proposal, develop a QKD protocol with location credentials using quantum position verification (QPV) based message and identity authentication. By using QKD with delayed authentication as a base, and later simplifying QPV-based message authentication, we significantly reduce the number of QPV runs, which currently acts as a bottleneck. Besides demonstrating security for the proposed protocol, we also provide improvements to QPV security analysis, including generalization of the QPV adversary model, tightening a trace distance bound using semidefinite programming, and propose a multi-basis QPV requiring only BB84 state preparation but with multiple measurement basis. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 04:08:44 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Kon",
"Wen Yu",
""
],
[
"Primaatmaja",
"Ignatius William",
""
],
[
"Chakraborty",
"Kaushik",
""
],
[
"Lim",
"Charles",
""
]
] |
2506.03550 | Kanami Imamura | Kanami Imamura, Tomohiko Nakamura, Norihiro Takamune, Kohei Yatabe, Hiroshi Saruwatari | Local Equivariance Error-Based Metrics for Evaluating Sampling-Frequency-Independent Property of Neural Network | 5 pages, 4 figures, accepted for European Signal Processing Conference 2025 (EUSIPCO 2025) | null | null | null | cs.SD eess.AS | http://creativecommons.org/licenses/by-sa/4.0/ | Audio signal processing methods based on deep neural networks (DNNs) are typically trained only at a single sampling frequency (SF) and therefore require signal resampling to handle untrained SFs. However, recent studies have shown that signal resampling can degrade performance with untrained SFs. This problem has been overlooked because most studies evaluate only the performance at trained SFs. In this paper, to assess the robustness of DNNs to SF changes, which we refer to as the SF-independent (SFI) property, we propose three metrics to quantify the SFI property on the basis of local equivariance error (LEE). LEE measures the robustness of DNNs to input transformations. By using signal resampling as input transformation, we extend LEE to measure the robustness of audio source separation methods to signal resampling. The proposed metrics are constructed to quantify the SFI property in specific network components responsible for predicting time-frequency masks. Experiments on music source separation demonstrated a strong correlation between the proposed metrics and performance degradation at untrained SFs. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 04:08:59 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Imamura",
"Kanami",
""
],
[
"Nakamura",
"Tomohiko",
""
],
[
"Takamune",
"Norihiro",
""
],
[
"Yatabe",
"Kohei",
""
],
[
"Saruwatari",
"Hiroshi",
""
]
] |
2506.03551 | Jamal Al-Yasiri Mr. | Jamal H. Al-Yasiri, Mohamad Fadli Bin Zolkipli, Nik Fatinah N Mohd Farid, Mohammed Alsamman, Zainab Ali Mohammed | A Threat Intelligence Event Extraction Conceptual Model for Cyber Threat Intelligence Feeds | IEEE conference paper | 2024 7th International Conference on Internet Applications, Protocols, and Services (NETAPPS), Kuala Lumpur: IEEE, 11,2024 | 10.1109/NETAPPS63333.2024.10823639 | null | cs.CR | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In response to the escalating cyber threats, the efficiency of Cyber Threat Intelligence (CTI) data collection has become paramount in ensuring robust cybersecurity. However, existing works encounter significant challenges in preprocessing large volumes of multilingual threat data, leading to inefficiencies in real-time threat analysis. This paper presents a systematic review of current techniques aimed at enhancing CTI data collection efficiency. Additionally, it proposes a conceptual model to further advance the effectiveness of threat intelligence feeds. Following the PRISMA guidelines, the review examines relevant studies from the Scopus database, highlighting the critical role of artificial intelligence (AI) and machine learning models in optimizing CTI data preprocessing. The findings underscore the importance of AI-driven methods, particularly supervised and unsupervised learning, in significantly improving the accuracy of threat detection and event extraction, thereby strengthening cybersecurity. Furthermore, the study identifies a gap in the existing research and introduces XBC conceptual model integrating XLM-RoBERTa, BiGRU, and CRF, specifically developed to address this gap. This paper contributes conceptually to the field by providing a detailed analysis of current CTI data collection techniques and introducing an innovative conceptual model to enhance future threat intelligence capabilities. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 04:09:01 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Al-Yasiri",
"Jamal H.",
""
],
[
"Zolkipli",
"Mohamad Fadli Bin",
""
],
[
"Farid",
"Nik Fatinah N Mohd",
""
],
[
"Alsamman",
"Mohammed",
""
],
[
"Mohammed",
"Zainab Ali",
""
]
] |
2506.03552 | Kyong Sei Lee | Kyong Sei Lee, Giuseppe Iaselli, Youngmin Jo, Minho Kang, Tae Jeong Kim, Dayron Ramos Lopez, and Gabriella Pugliese | Develoment of thin high-pressure-laminate RPC electrodes for future high-energy experiments | 6 pages | null | null | null | physics.ins-det hep-ex | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In this R&D, an innovative method for producing thin high-pressure laminate (HPL) electrodes for resistive plate chambers (RPC) for future high-energy experiments is introduced. Instead of using thick phenolic HPL (2-mm thick Bakelite), which has been used for conventional RPC triggers, the RPC electrodes in the present study are constructed by bonding 500 {\mu}m-thick melamine-based HPL to a graphite-coated polycarbonate plate. A double-gap RPC prototype to demostrate the present technology has been constructed and tested for cosmic muons. Furthermore, the uniform detector characteristrics shown in the test result allows us to explore the present technology in future high-energy experiments. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 04:09:05 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Lee",
"Kyong Sei",
""
],
[
"Iaselli",
"Giuseppe",
""
],
[
"Jo",
"Youngmin",
""
],
[
"Kang",
"Minho",
""
],
[
"Kim",
"Tae Jeong",
""
],
[
"Lopez",
"Dayron Ramos",
""
],
[
"Pugliese",
"Gabriella",
""
]
] |
2506.03553 | Zhen Chen | Zhen Chen, Yijia Wu, X. C. Xie | Three-Majorana Cotunneling Interferometer for Non-Abelian Braiding and Topological Quantum Gate Implementation | 16 pages, 8 figures | null | null | null | cond-mat.mes-hall | http://creativecommons.org/licenses/by/4.0/ | We propose a novel scheme for performing Majorana zero mode (MZM) braiding utilizing cotunneling processes in a three-MZM system incorporating reference arms. This approach relies on the interference between cotunneling paths through the MZMs and reference arms, establishing an effective, tunable coupling between the MZMs. The strength and sign of this coupling can be manipulated via the reference arms and applied magnetic flux. Notably, the introduction of a half quantum flux reverses the coupling sign, enabling an echo-like protocol to eliminate dynamic phases during braiding. Our setup, requiring only three MZMs, represents a minimal platform for demonstrating non-Abelian braiding statistics. We demonstrate that this system facilitates the implementation of Clifford gates via braiding and, significantly, permits the realization of non-Clifford gates, such as the $T$ gate, by geometric phase, thereby offering a potential pathway towards universal topological quantum computation. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 04:12:28 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Chen",
"Zhen",
""
],
[
"Wu",
"Yijia",
""
],
[
"Xie",
"X. C.",
""
]
] |
2506.03554 | Reo Yoneyama | Reo Yoneyama, Masaya Kawamura, Ryo Terashima, Ryuichi Yamamoto, Tomoki Toda | Comparative Analysis of Fast and High-Fidelity Neural Vocoders for Low-Latency Streaming Synthesis in Resource-Constrained Environments | Accepted to Interspeech 2025 | null | null | null | cs.SD eess.AS | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In real-time speech synthesis, neural vocoders often require low-latency synthesis through causal processing and streaming. However, streaming introduces inefficiencies absent in batch synthesis, such as limited parallelism, inter-frame dependency management, and parameter loading overhead. This paper proposes multi-stream Wavehax (MS-Wavehax), an efficient neural vocoder for low-latency streaming, by extending the aliasing-free neural vocoder Wavehax with multi-stream decomposition. We analyze the latency-throughput trade-off in a CPU-only environment and identify key bottlenecks in streaming neural vocoders. Our findings provide practical insights for optimizing chunk sizes and designing vocoders tailored to specific application demands and hardware constraints. Furthermore, our subjective evaluations show that MS-Wavehax delivers high speech quality under causal and non-causal conditions while being remarkably compact and easily deployable in resource-constrained environments. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 04:18:15 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Yoneyama",
"Reo",
""
],
[
"Kawamura",
"Masaya",
""
],
[
"Terashima",
"Ryo",
""
],
[
"Yamamoto",
"Ryuichi",
""
],
[
"Toda",
"Tomoki",
""
]
] |
2506.03555 | Tianpei Zhang | Tianpei Zhang and Jufeng Zhao and Yiming Zhu and Guangmang Cui | WIFE-Fusion:Wavelet-aware Intra-inter Frequency Enhancement for Multi-model Image Fusion | null | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Multimodal image fusion effectively aggregates information from diverse modalities, with fused images playing a crucial role in vision systems. However, existing methods often neglect frequency-domain feature exploration and interactive relationships. In this paper, we propose wavelet-aware Intra-inter Frequency Enhancement Fusion (WIFE-Fusion), a multimodal image fusion framework based on frequency-domain components interactions. Its core innovations include: Intra-Frequency Self-Attention (IFSA) that leverages inherent cross-modal correlations and complementarity through interactive self-attention mechanisms to extract enriched frequency-domain features, and Inter-Frequency Interaction (IFI) that enhances enriched features and filters latent features via combinatorial interactions between heterogeneous frequency-domain components across modalities. These processes achieve precise source feature extraction and unified modeling of feature extraction-aggregation. Extensive experiments on five datasets across three multimodal fusion tasks demonstrate WIFE-Fusion's superiority over current specialized and unified fusion methods. Our code is available at https://github.com/Lmmh058/WIFE-Fusion. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 04:18:32 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Zhang",
"Tianpei",
""
],
[
"Zhao",
"Jufeng",
""
],
[
"Zhu",
"Yiming",
""
],
[
"Cui",
"Guangmang",
""
]
] |
2506.03556 | Riaz-Ul-Haque Mian | Wang WeiQuan and Riaz-ul-Haque Mian | Optimizing FPGA and Wafer Test Coverage with Spatial Sampling and Machine Learning | null | null | null | null | cs.LG | http://creativecommons.org/licenses/by-nc-nd/4.0/ | In semiconductor manufacturing, testing costs remain significantly high, especially during wafer and FPGA testing. To reduce the number of required tests while maintaining predictive accuracy, this study investigates three baseline sampling strategies: Random Sampling, Stratified Sampling, and k-means Clustering Sampling. To further enhance these methods, this study proposes a novel algorithm that improves the sampling quality of each approach. This research is conducted using real industrial production data from wafer-level tests and silicon measurements from various FPGAs. This study introduces two hybrid strategies: Stratified with Short Distance Elimination (S-SDE) and k-means with Short Distance Elimination (K-SDE). Their performance is evaluated within the framework of Gaussian Process Regression (GPR) for predicting wafer and FPGA test data. At the core of our proposed approach is the Short Distance Elimination (SDE) algorithm, which excludes spatially proximate candidate points during sampling, thereby ensuring a more uniform distribution of training data across the physical domain. A parameter sweep was conducted over the (alpha, beta) thresholds, where alpha and beta are in the range {0, 1, 2, 3, 4} and not both zero, to identify the optimal combination that minimizes RMSD. Experimental results on a randomly selected wafer file reveal that (alpha, beta) equal (2, 2) yields the lowest RMSD. Accordingly, all subsequent experiments adopt this parameter configuration. The results demonstrate that the proposed SDE-based strategies enhance predictive accuracy: K-SDE improves upon k-means sampling by 16.26 percent (wafer) and 13.07 percent (FPGA), while S-SDE improves upon stratified sampling by 16.49 percent (wafer) and 8.84 percent (FPGA). | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 04:19:06 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"WeiQuan",
"Wang",
""
],
[
"Mian",
"Riaz-ul-Haque",
""
]
] |
2506.03557 | Lin Sun | Lin Sun, Chuang Liu, Peng Liu, Bingyang Li, Weijia Lu, Ning Wu | BPO: Revisiting Preference Modeling in Direct Preference Optimization | null | null | null | null | cs.CL | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Direct Preference Optimization (DPO) have emerged as a popular method for aligning Large Language Models (LLMs) with human preferences. While DPO effectively preserves the relative ordering between chosen and rejected responses through pairwise ranking losses, it often neglects absolute reward magnitudes. This oversight can decrease the likelihood of chosen responses and increase the risk of generating out-of-distribution responses, leading to poor performance. We term this issue Degraded Chosen Responses (DCR).To address this issue, we propose Balanced Preference Optimization (BPO), a novel framework that dynamically balances the optimization of chosen and rejected responses through two key components: balanced reward margin and gap adaptor. Unlike previous methods, BPO can fundamentally resolve DPO's DCR issue, without introducing additional constraints to the loss function. Experimental results on multiple mathematical reasoning tasks show that BPO significantly outperforms DPO, improving accuracy by +10.1% with Llama-3.1-8B-Instruct (18.8% to 28.9%) and +11.7% with Qwen2.5-Math-7B (35.0% to 46.7%). It also surpasses DPO variants by +3.6% over IPO (43.1%), +5.0% over SLiC (41.7%), and +3.1% over Cal-DPO (43.6%) on the same model. Remarkably, our algorithm requires only a single line of code modification, making it simple to implement and fully compatible with existing DPO-based frameworks. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 04:21:01 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Sun",
"Lin",
""
],
[
"Liu",
"Chuang",
""
],
[
"Liu",
"Peng",
""
],
[
"Li",
"Bingyang",
""
],
[
"Lu",
"Weijia",
""
],
[
"Wu",
"Ning",
""
]
] |
2506.03558 | Jiawei Chen | Jiawei Chen, Xinyan Guan, Qianhao Yuan, Guozhao Mo, Weixiang Zhou, Yaojie Lu, Hongyu Lin, Ben He, Le Sun, Xianpei Han | ConsistentChat: Building Skeleton-Guided Consistent Dialogues for Large Language Models from Scratch | null | null | null | null | cs.CL | http://creativecommons.org/licenses/by/4.0/ | Current instruction data synthesis methods primarily focus on single-turn instructions and often neglect cross-turn coherence, resulting in context drift and reduced task completion rates in extended conversations. To address this limitation, we propose Skeleton-Guided Multi-Turn Dialogue Generation, a framework that constrains multi-turn instruction synthesis by explicitly modeling human conversational intent. It operates in two stages: (1) Intent Modeling, which captures the global structure of human dialogues by assigning each conversation to one of nine well-defined intent trajectories, ensuring a coherent and goal-oriented information flow; and (2) Skeleton Generation, which constructs a structurally grounded sequence of user queries aligned with the modeled intent, thereby serving as a scaffold that constrains and guides the downstream instruction synthesis process. Based on this process, we construct ConsistentChat, a multi-turn instruction dataset with approximately 15,000 multi-turn conversations and 224,392 utterances. Experiments on the Light, Topdial, and MT-Eval benchmarks show that models fine-tuned on ConsistentChat achieve a 20-30% improvement in chat consistency and up to a 15% increase in task success rate, significantly outperforming models trained on existing single-turn and multi-turn instruction datasets. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 04:21:48 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Chen",
"Jiawei",
""
],
[
"Guan",
"Xinyan",
""
],
[
"Yuan",
"Qianhao",
""
],
[
"Mo",
"Guozhao",
""
],
[
"Zhou",
"Weixiang",
""
],
[
"Lu",
"Yaojie",
""
],
[
"Lin",
"Hongyu",
""
],
[
"He",
"Ben",
""
],
[
"Sun",
"Le",
""
],
[
"Han",
"Xianpei",
""
]
] |
2506.03559 | Zeeshan Gul | M. Zeeshan Gul, W. Ahmad, M.M.M. Nasir, Faisal Javed, Orhan Donmez and Bander Almutairi | Study of Complexity Factor and Stability of Dynamical Systems in $f(G)$ Gravity | 28 pages | Eur. Phys. J. C 85(2025)595 | 10.1140/epjc/s10052-025-14163-4 | null | gr-qc | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In this paper, we evaluate the complexity of the non-static cylindrical geometry with anisotropic matter configuration in the framework of modified Gauss-Bonnet theory. In this perspective, we calculate modified field equations, the C energy formula, and the mass function that helps to understand the astrophysical structures in this modified gravity. Furthermore, we use the Weyl tensor and obtain different structure scalars by orthogonally splitting the Riemann tensor. One of these scalars, $YTF$ is referred to as the complexity factor. This parameter measures the system's complexity due to non-uniform energy density and non-isotropic pressure. We select the identical complexity factor for the structure as used in the non-static scenario while considering the analogous criterion for the most elementary pattern of development. This technique involves formulating structural scalars that illustrate the fundamental features of the system. A fluid distribution that satisfies the vanishing complexity requirement and evolves homologously is characterized as isotropic, geodesic, homogeneous, and shear-free. In the dissipative scenario, the fluid remains geodesic while exhibiting shear, resulting in an extensive array of solutions. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 04:23:21 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Gul",
"M. Zeeshan",
""
],
[
"Ahmad",
"W.",
""
],
[
"Nasir",
"M. M. M.",
""
],
[
"Javed",
"Faisal",
""
],
[
"Donmez",
"Orhan",
""
],
[
"Almutairi",
"Bander",
""
]
] |
2506.03560 | Yusuke Shimada | Kenji Fukushima, Yusuke Shimada | Imaginary Rotating Gluonic Matter at Strong Coupling | 17 pages, 6 figures | null | null | null | hep-ph hep-lat hep-th | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We write down an effective theory of the Polyakov loop to investigate the deconfinement phase transition of imaginary-rotating gluonic matter using the strong-coupling expansion. We find the strength of the nearest-neighbor Polyakov-loop interaction modified by the sum of contributions involving the chair-type loops along the temporal direction. Our results show that the deconfinement transition temperature increases with increasing imaginary angular velocity, which agrees with the predictions from the models and the high-temperature perturbative calculations. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 04:25:12 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Fukushima",
"Kenji",
""
],
[
"Shimada",
"Yusuke",
""
]
] |
2506.03561 | Kaiwen Fang | Kaiwen Fang, Inho Sin, Geunyeong Byeon | Disjunctive Benders Decomposition | null | null | null | null | math.OC | http://creativecommons.org/licenses/by/4.0/ | We propose a novel enhancement to Benders Decomposition (BD) that generates valid inequalities for the convex hull of the Benders reformulation, addressing a key limitation of conventional BD-its cuts are typically tight only for the continuous relaxation. Our method efficiently integrates disjunctive programming theory with BD, introducing a new routine that leverages existing cut-generating oracles for uncovering constraints required to construct valid inequalities for the convex hull. For mixed-binary linear programs, this approach eliminates the need to solve the master problem as a mixed-integer program. Additionally, we extend the a posteriori strengthening and lifting procedure for lift-and-project cuts into the BD framework, and present an approximate routine for generating lift-and-project cuts. Numerical results on large-scale instances show that our approach significantly reduces the number of branch-and-bound nodes required to reach the lower bound achieved by conventional BD, often by orders of magnitude. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 04:26:08 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Fang",
"Kaiwen",
""
],
[
"Sin",
"Inho",
""
],
[
"Byeon",
"Geunyeong",
""
]
] |
2506.03562 | Stefanos Theodorakopoulos | Antonis Papapantoleon, Alexandros Saplaouras, Stefanos Theodorakopoulos | Stability of backward propagation of chaos | arXiv admin note: text overlap with arXiv:2408.13758 | null | null | null | math.PR | http://creativecommons.org/licenses/by/4.0/ | The purpose of the present paper is to introduce and establish a notion of stability for the backward propagation of chaos, with respect to (initial) data sets. Consider, for example, a sequence of discrete-time martingales converging to a continuous-time limit, and a system of mean-field BSDEs that satisfies the backward propagation of chaos, i.e. converges to a sequence of i.i.d. McKean-Vlasov BSDEs. Then, we say that the backward propagation of chaos is stable if the system of mean-field BSDEs driven by the discrete-time martingales converges to the sequence of McKean-Vlasov BSDEs driven by the continuous-time limit. We consider the convergence scheme of the backward propagation of chaos as the image of the corresponding data set under which this scheme is established. Then, using an appropriate notion of convergence for data sets, we are able to show a variety of continuity properties for this functional point of view. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 04:26:34 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Papapantoleon",
"Antonis",
""
],
[
"Saplaouras",
"Alexandros",
""
],
[
"Theodorakopoulos",
"Stefanos",
""
]
] |
2506.03563 | Jie Shen | Guoan Li, Xiaofan Shi, Ziwei Dou, Guang Yang, Jiayu Shi, Marco Rossi, Ghada Badawy, Yuxiao Song, Ruixuan Zhang, Yupeng Li, Zhiyuan Zhang, Anqi Wang, Xingchen Guo, Xiao Deng, Bingbing Tong, Peiling Li, Zhaozheng Lyu, Guangtong Liu, Fanming Qu, Erik P. A. M. Bakkers, Micha{\l} P. Nowak, Pawe{\l} W\'ojcik, Li Lu, and Jie Shen | Enhanced and modulable induced superconducting gap and effective Land\'e g-factor in Pb-InSb hybrid devices | null | null | null | null | cond-mat.mes-hall cond-mat.supr-con quant-ph | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The hybrid system of a conventional superconductor (SC) on a semiconductor (SM) nanowire with strong spin-orbit coupling (SOC) represents a promising platform for achieving topological superconductivity and Majorana zero modes (MZMs) towards topological quantum computation. While aluminum (Al)-based hybrid nanowire devices have been widely utilized, their limited superconducting gap and intrinsic weak SOC as well as small Land\'e g-factor may hinder future experimental advancements. In contrast, we demonstrate that lead (Pb)-based hybrid quantum devices exhibit a remarkably large and hard proximity-induced superconducting gap, exceeding that of Al by an order of magnitude. By exploiting electrostatic gating to modulate wavefunction distribution and SC-SM interfacial coupling, this gap can be continuously tuned from its maximum value (~1.4 meV, matching the bulk Pb gap) down to nearly zero while maintaining the hardness. Furthermore, magnetic-field-dependent measurements reveal a radial evolution of the gap structure with anti-crossing feature, indicative of strong SOC and huge effective g-factors up to 76. These findings underscore the superior functionality of Pb-based hybrid systems, significantly advancing their potential for realizing and stabilizing MZMs and the further scalable topological quantum architectures. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 04:27:57 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Li",
"Guoan",
""
],
[
"Shi",
"Xiaofan",
""
],
[
"Dou",
"Ziwei",
""
],
[
"Yang",
"Guang",
""
],
[
"Shi",
"Jiayu",
""
],
[
"Rossi",
"Marco",
""
],
[
"Badawy",
"Ghada",
""
],
[
"Song",
"Yuxiao",
""
],
[
"Zhang",
"Ruixuan",
""
],
[
"Li",
"Yupeng",
""
],
[
"Zhang",
"Zhiyuan",
""
],
[
"Wang",
"Anqi",
""
],
[
"Guo",
"Xingchen",
""
],
[
"Deng",
"Xiao",
""
],
[
"Tong",
"Bingbing",
""
],
[
"Li",
"Peiling",
""
],
[
"Lyu",
"Zhaozheng",
""
],
[
"Liu",
"Guangtong",
""
],
[
"Qu",
"Fanming",
""
],
[
"Bakkers",
"Erik P. A. M.",
""
],
[
"Nowak",
"Michał P.",
""
],
[
"Wójcik",
"Paweł",
""
],
[
"Lu",
"Li",
""
],
[
"Shen",
"Jie",
""
]
] |
2506.03564 | Yuri Uno | Yuri Uno, Tetsuya Hashimoto, Tomotsugu Goto, Shinnosuke Hisano, Yi Hang Valerie Wong, Arthur Chen, Sujin Eie, Simon C.-C. Ho, James O. Chibueze, Yu-Wei Lin, Seong Jin Kim, Tzu-Yin Hsu, Poya Wang, Pei Wang, and Murthadza Aznam | Do they repeat? Monitoring 36 non-repeating FRBs with FAST | 16 pages, 11 figure, accepted for publication in MNRAS. A recording of the oral presentation at TPS 2024 is available at: https://www.youtube.com/watch?v=SSxDKeOedHc | null | null | null | astro-ph.HE | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The origin of fast radio bursts (FRBs), highly energetic, millisecond-duration radio pulses originating from beyond our galaxy, remains unknown. Observationally, FRBs are classified as non-repeating or repeating, however, this classification is complicated by limited observing time and sensitivity constraints, which may result in some repeating FRBs being misidentified as non-repeating. To address this issue, we adopt both empirical and machine-learning techniques from previous studies to identify candidates that may have been misclassified. We conducted follow-up observations of 36 such candidates, each observed for 10 minutes using the Five-hundred-meter Aperture Spherical Telescope (FAST). No radio bursts exceeding a signal-to-noise ratio of 7 were detected, with a typical 7 sigma fluence limit of ~0.013 Jy ms. We constrain the repetition rates of these sources using two statistical models of FRB occurrence. Combining our FAST non-detections with prior observations, we derive upper limits on the repetition rates of ~$10^{-2.6}$-$10^{-0.22}$ hr$^{-1}$ under a Poisson process, and ~$10^{-2.3}$-$10^{-0.25}$ hr$^{-1}$ under a Weibull process. This work presents one of the most stringent upper limits on FRB repetition rates to date, based on a sample size five times larger than those used in previous studies. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 04:28:58 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Uno",
"Yuri",
""
],
[
"Hashimoto",
"Tetsuya",
""
],
[
"Goto",
"Tomotsugu",
""
],
[
"Hisano",
"Shinnosuke",
""
],
[
"Wong",
"Yi Hang Valerie",
""
],
[
"Chen",
"Arthur",
""
],
[
"Eie",
"Sujin",
""
],
[
"Ho",
"Simon C. -C.",
""
],
[
"Chibueze",
"James O.",
""
],
[
"Lin",
"Yu-Wei",
""
],
[
"Kim",
"Seong Jin",
""
],
[
"Hsu",
"Tzu-Yin",
""
],
[
"Wang",
"Poya",
""
],
[
"Wang",
"Pei",
""
],
[
"Aznam",
"Murthadza",
""
]
] |
2506.03565 | Haotian Tang | Jiashan Zheng and Haotian Tang | Some progress in global existence of solutions to a higher-dimensional chemotaxis system modelling Alopecia Areata | null | null | null | null | math.AP | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | This paper is concerned with different logistic damping effects on the global existence in a chemotaxis system \begin{equation*} \left\{\aligned
& u_{t}=\Delta u-\chi_{1}\nabla\cdot(u\nabla w)+w-\mu_{1}u^{r_{1}},&&x\in\Omega,t>0,
& v_{t}=\Delta v-\chi_{2}\nabla\cdot(v\nabla w)+w+ruv-\mu_{2}v^{r_{2}},&&x\in\Omega,t>0,
& w_{t}=\Delta w+u+v-w,&&x\in\Omega,t>0,\\ \endaligned\right. \end{equation*} which was initially proposed by Dobreva \emph{et al.} (\cite{DP2020}) to describe the dynamics of hair loss in Alopecia Areata form. Here, $\Omega\subset\mathbb R^{N}$ $(N\geq3)$ is a bounded domain with smooth boundary, and the parameters fulfill $\chi_{i}>0$, $\mu_{i}>0$, $r_{i}\geq2$ $(i=1,2)$ and $r>0$. It is proved that if $r_{1}=r_{2}=2$ and $\min\{\mu_{1},\mu_{1}\}>\mu^{\star}$ or $r_{i}>2$ $(i=1,2)$, the Neumann type initial-boundary value problem admits a unique classical solution which is globally bounded in $\Omega\times(0,\infty)$ for all sufficiently smooth initial data. The lower bound $\mu^{\ast}=\frac{2(N-2)_{+}}{N}C_{\frac{N}{2}+1}^{\frac{1}{\frac{N}{2}+1}}\max\{\chi_{1},\chi_{2}\}+\left[(\frac{2}{N})^{\frac{2}{N+2}}\frac{N}{N+2}\right]r$, where $C_{\frac{N}{2}+1}$ is a positive constant corresponding to the maximal Sobolev regularity. Furthermore, the basic assumption $\mu_{i}>0$ $(i=1,2)$ can ensure the global existence of a weak solution. Notably, our findings not only first provide new insights into the weak solution theory of this system but also offer some novel quantized impact of the (generalized) logistic source on preventing blow-ups. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 04:29:55 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Zheng",
"Jiashan",
""
],
[
"Tang",
"Haotian",
""
]
] |
2506.03566 | Langlin Huang | Langlin Huang, Chengsong Huang, Jixuan Leng, Di Huang, Jiaxin Huang | POSS: Position Specialist Generates Better Draft for Speculative Decoding | null | null | null | null | cs.CL cs.AI cs.LG | http://creativecommons.org/licenses/by-nc-nd/4.0/ | Speculative decoding accelerates Large Language Model (LLM) inference by using a small draft model to predict multiple tokens, and a large target model to verify these tokens in parallel. Recent studies leverage the hidden state of the target model to enhance draft model prediction accuracy. However, existing methods suffer from the degrading quality of draft token predictions at later positions, due to error accumulation in draft model generated features. In this paper, we propose Position Specialists (PosS), which consist of multiple position-specialized draft layers to generate tokens at assigned position(s). Position specialists greatly improve token acceptance rate at later positions per drafting round, as each specialist only needs to focus on handling a certain level of draft model feature deviation. Experiment results on Llama-3-8B-Instruct and Llama-2-13B-chat across six datasets demonstrate that PosS effectively improves over baselines on average acceptance length and speed-up ratio. Our codebase is available at https://github.com/shrango/PosS. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 04:30:30 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Huang",
"Langlin",
""
],
[
"Huang",
"Chengsong",
""
],
[
"Leng",
"Jixuan",
""
],
[
"Huang",
"Di",
""
],
[
"Huang",
"Jiaxin",
""
]
] |
2506.03567 | Junliang Wang | Hermann Edlbauer, Junliang Wang, A. M. Saffat-Ee Huq, Ian Thorvaldson, Michael T. Jones, Saiful Haque Misha, William J. Pappas, Christian M. Moehle, Yu-Ling Hsueh, Henric Bornemann, Samuel K. Gorman, Yousun Chung, Joris G. Keizer, Ludwik Kranz, Michelle Y. Simmons | An 11-qubit atom processor in silicon | Main text with 10 pages and 4 figures. Supplementary information with 14 pages and 10 figures | null | null | null | quant-ph cond-mat.mes-hall | http://creativecommons.org/licenses/by/4.0/ | Phosphorus atoms in silicon are an outstanding platform for quantum computing as their nuclear spins exhibit coherence time over seconds. By placing multiple phosphorus atoms within a radius of a few nanometers, they couple via the hyperfine interaction to a single, shared electron. Such a nuclear spin register enables multi-qubit control above the fault-tolerant threshold and the execution of small-scale quantum algorithms. To achieve quantum error correction, fast and efficient interconnects have to be implemented between spin registers while maintaining high fidelity across all qubit metrics. Here, we demonstrate such integration with a fully controlled 11-qubit atom processor composed of two multi-nuclear spin registers which are linked via electron exchange interaction. Through the development of scalable calibration and control protocols, we achieve coherent coupling between nuclear spins using a combination of single- and multi-qubit gates with all fidelities ranging from 99.5% to 99.99%. We verify the efficient all-to-all connectivity by preparing both local and non-local Bell states with a record state fidelity beyond 99% and extend entanglement through the generation of Greenberger-Horne-Zeilinger (GHZ) states over all data qubits. By establishing high-fidelity operation across interconnected nuclear-spin registers, we realise a key milestone towards fault-tolerant quantum computation with atom processors. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 04:31:08 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Edlbauer",
"Hermann",
""
],
[
"Wang",
"Junliang",
""
],
[
"Huq",
"A. M. Saffat-Ee",
""
],
[
"Thorvaldson",
"Ian",
""
],
[
"Jones",
"Michael T.",
""
],
[
"Misha",
"Saiful Haque",
""
],
[
"Pappas",
"William J.",
""
],
[
"Moehle",
"Christian M.",
""
],
[
"Hsueh",
"Yu-Ling",
""
],
[
"Bornemann",
"Henric",
""
],
[
"Gorman",
"Samuel K.",
""
],
[
"Chung",
"Yousun",
""
],
[
"Keizer",
"Joris G.",
""
],
[
"Kranz",
"Ludwik",
""
],
[
"Simmons",
"Michelle Y.",
""
]
] |
2506.03569 | Chen Chen | Xiaomi LLM-Core Team: Zihao Yue, Zhenru Lin, Yifan Song, Weikun Wang, Shuhuai Ren, Shuhao Gu, Shicheng Li, Peidian Li, Liang Zhao, Lei Li, Kainan Bao, Hao Tian, Hailin Zhang, Gang Wang, Dawei Zhu, Cici, Chenhong He, Bowen Ye, Bowen Shen, Zihan Zhang, Zihan Jiang, Zhixian Zheng, Zhichao Song, Zhenbo Luo, Yue Yu, Yudong Wang, Yuanyuan Tian, Yu Tu, Yihan Yan, Yi Huang, Xu Wang, Xinzhe Xu, Xingchen Song, Xing Zhang, Xing Yong, Xin Zhang, Xiangwei Deng, Wenyu Yang, Wenhan Ma, Weiwei Lv, Weiji Zhuang, Wei Liu, Sirui Deng, Shuo Liu, Shimao Chen, Shihua Yu, Shaohui Liu, Shande Wang, Rui Ma, Qiantong Wang, Peng Wang, Nuo Chen, Menghang Zhu, Kangyang Zhou, Kang Zhou, Kai Fang, Jun Shi, Jinhao Dong, Jiebao Xiao, Jiaming Xu, Huaqiu Liu, Hongshen Xu, Heng Qu, Haochen Zhao, Hanglong Lv, Guoan Wang, Duo Zhang, Dong Zhang, Di Zhang, Chong Ma, Chang Liu, Can Cai, Bingquan Xia | MiMo-VL Technical Report | 32 pages | null | null | null | cs.CL | http://creativecommons.org/licenses/by/4.0/ | We open-source MiMo-VL-7B-SFT and MiMo-VL-7B-RL, two powerful vision-language models delivering state-of-the-art performance in both general visual understanding and multimodal reasoning. MiMo-VL-7B-RL outperforms Qwen2.5-VL-7B on 35 out of 40 evaluated tasks, and scores 59.4 on OlympiadBench, surpassing models with up to 78B parameters. For GUI grounding applications, it sets a new standard with 56.1 on OSWorld-G, even outperforming specialized models such as UI-TARS. Our training combines four-stage pre-training (2.4 trillion tokens) with Mixed On-policy Reinforcement Learning (MORL) integrating diverse reward signals. We identify the importance of incorporating high-quality reasoning data with long Chain-of-Thought into pre-training stages, and the benefits of mixed RL despite challenges in simultaneous multi-domain optimization. We also contribute a comprehensive evaluation suite covering 50+ tasks to promote reproducibility and advance the field. The model checkpoints and full evaluation suite are available at https://github.com/XiaomiMiMo/MiMo-VL. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 04:32:54 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Core Team",
"",
""
],
[
"Yue",
"Zihao",
""
],
[
"Lin",
"Zhenru",
""
],
[
"Song",
"Yifan",
""
],
[
"Wang",
"Weikun",
""
],
[
"Ren",
"Shuhuai",
""
],
[
"Gu",
"Shuhao",
""
],
[
"Li",
"Shicheng",
""
],
[
"Li",
"Peidian",
""
],
[
"Zhao",
"Liang",
""
],
[
"Li",
"Lei",
""
],
[
"Bao",
"Kainan",
""
],
[
"Tian",
"Hao",
""
],
[
"Zhang",
"Hailin",
""
],
[
"Wang",
"Gang",
""
],
[
"Zhu",
"Dawei",
""
],
[
"Cici",
"",
""
],
[
"He",
"Chenhong",
""
],
[
"Ye",
"Bowen",
""
],
[
"Shen",
"Bowen",
""
],
[
"Zhang",
"Zihan",
""
],
[
"Jiang",
"Zihan",
""
],
[
"Zheng",
"Zhixian",
""
],
[
"Song",
"Zhichao",
""
],
[
"Luo",
"Zhenbo",
""
],
[
"Yu",
"Yue",
""
],
[
"Wang",
"Yudong",
""
],
[
"Tian",
"Yuanyuan",
""
],
[
"Tu",
"Yu",
""
],
[
"Yan",
"Yihan",
""
],
[
"Huang",
"Yi",
""
],
[
"Wang",
"Xu",
""
],
[
"Xu",
"Xinzhe",
""
],
[
"Song",
"Xingchen",
""
],
[
"Zhang",
"Xing",
""
],
[
"Yong",
"Xing",
""
],
[
"Zhang",
"Xin",
""
],
[
"Deng",
"Xiangwei",
""
],
[
"Yang",
"Wenyu",
""
],
[
"Ma",
"Wenhan",
""
],
[
"Lv",
"Weiwei",
""
],
[
"Zhuang",
"Weiji",
""
],
[
"Liu",
"Wei",
""
],
[
"Deng",
"Sirui",
""
],
[
"Liu",
"Shuo",
""
],
[
"Chen",
"Shimao",
""
],
[
"Yu",
"Shihua",
""
],
[
"Liu",
"Shaohui",
""
],
[
"Wang",
"Shande",
""
],
[
"Ma",
"Rui",
""
],
[
"Wang",
"Qiantong",
""
],
[
"Wang",
"Peng",
""
],
[
"Chen",
"Nuo",
""
],
[
"Zhu",
"Menghang",
""
],
[
"Zhou",
"Kangyang",
""
],
[
"Zhou",
"Kang",
""
],
[
"Fang",
"Kai",
""
],
[
"Shi",
"Jun",
""
],
[
"Dong",
"Jinhao",
""
],
[
"Xiao",
"Jiebao",
""
],
[
"Xu",
"Jiaming",
""
],
[
"Liu",
"Huaqiu",
""
],
[
"Xu",
"Hongshen",
""
],
[
"Qu",
"Heng",
""
],
[
"Zhao",
"Haochen",
""
],
[
"Lv",
"Hanglong",
""
],
[
"Wang",
"Guoan",
""
],
[
"Zhang",
"Duo",
""
],
[
"Zhang",
"Dong",
""
],
[
"Zhang",
"Di",
""
],
[
"Ma",
"Chong",
""
],
[
"Liu",
"Chang",
""
],
[
"Cai",
"Can",
""
],
[
"Xia",
"Bingquan",
""
]
] |
2506.03570 | Lin Sun | Lin Sun, Chuang Liu, Xiaofeng Ma, Tao Yang, Weijia Lu, Ning Wu | FreePRM: Training Process Reward Models Without Ground Truth Process Labels | null | null | null | null | cs.CL | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Recent advancements in Large Language Models (LLMs) have demonstrated that Process Reward Models (PRMs) play a crucial role in enhancing model performance. However, training PRMs typically requires step-level labels, either manually annotated or automatically generated, which can be costly and difficult to obtain at scale. To address this challenge, we introduce FreePRM, a weakly supervised framework for training PRMs without access to ground-truth step-level labels. FreePRM first generates pseudo step-level labels based on the correctness of final outcome, and then employs Buffer Probability to eliminate impact of noise inherent in pseudo labeling. Experimental results show that FreePRM achieves an average F1 score of 53.0% on ProcessBench, outperforming fully supervised PRM trained on Math-Shepherd by +24.1%. Compared to other open-source PRMs, FreePRM outperforms upon RLHFlow-PRM-Mistral-8B (28.4%) by +24.6%, EurusPRM (31.3%) by +21.7%, and Skywork-PRM-7B (42.1%) by +10.9%. This work introduces a new paradigm in PRM training, significantly reducing reliance on costly step-level annotations while maintaining strong performance. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 04:33:53 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Sun",
"Lin",
""
],
[
"Liu",
"Chuang",
""
],
[
"Ma",
"Xiaofeng",
""
],
[
"Yang",
"Tao",
""
],
[
"Lu",
"Weijia",
""
],
[
"Wu",
"Ning",
""
]
] |
2506.03571 | Kibae Lee | Chong Hyun Lee, Kibae Lee | DiagNet: Detecting Objects using Diagonal Constraints on Adjacency Matrix of Graph Neural Network | null | null | null | null | cs.CV cs.AI | http://creativecommons.org/licenses/by/4.0/ | We propose DaigNet, a new approach to object detection with which we can detect an object bounding box using diagonal constraints on adjacency matrix of a graph convolutional network (GCN). We propose two diagonalization algorithms based on hard and soft constraints on adjacency matrix and two loss functions using diagonal constraint and complementary constraint. The DaigNet eliminates the need for designing a set of anchor boxes commonly used. To prove feasibility of our novel detector, we adopt detection head in YOLO models. Experiments show that the DiagNet achieves 7.5% higher mAP50 on Pascal VOC than YOLOv1. The DiagNet also shows 5.1% higher mAP on MS COCO than YOLOv3u, 3.7% higher mAP than YOLOv5u, and 2.9% higher mAP than YOLOv8. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 04:34:48 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Lee",
"Chong Hyun",
""
],
[
"Lee",
"Kibae",
""
]
] |
2506.03572 | Saka\'e Fuchino | Saka\'e Fuchino | Extendible cardinals, and Laver-generic large cardinal axioms for extendibility | null | null | null | null | math.LO | http://creativecommons.org/licenses/by/4.0/ | We introduce (super-$C^{(\infty)}$-)Laver-generic large cardinal axioms for extendibility ((super-$C^{(\infty)}$-)LgLCAs for extendible, for short), and show that most of the previously known consequences of the (super-$C^{(\infty)}$-)LgLCAs for ultrahuge, in particular, the strong and general forms of Resurrection Principles, Maximality Principles, and Absoluteness Theorems, lready follow from (super-$C^{(\infty)}$-)LgLCAs for extendible. The consistency of the LgLCAs for extendible (for transfinitely iterable $\Sigma_2$-definable classes of posets) follows from an extendible cardinal while the consistency of super-$C^{(\infty)}$-LgLCAs for extendible follows from a model with a super-$C^{(\infty)}$-extendible cardinal. If $\kappa$ is an almost-huge cardinal, there are cofinally many $\kappa_0<\kappa$ such that $V_\kappa\models{\kappa_0\mbox{ is super-}C^{(\infty)}\mbox{ extendible}}$. Most of the known reflection properties follow already from some of the LgLCAs for supercompact. We give a survey on the related results. We also show the separation between some of the LgLCAs as well as between LgLCAs and their consequences. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 04:39:33 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Fuchino",
"Sakaé",
""
]
] |
2506.03573 | Lin Sun | Lin Sun, Can Zhang | Exchange of Perspective Prompting Enhances Reasoning in Large Language Models | null | null | null | null | cs.CL | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Large language models (LLMs) have made significant advancements in addressing diverse natural language processing (NLP) tasks. However, their performance is often limited by inherent comprehension of problems. To address this limitation, we propose Exchange-of-Perspective (EoP), a novel framework designed to exchange perspectives across different definitions of problem, so that it can break the fixed mindset from any particular formulation of the question. We conducted extensive and comprehensive experiments on 8 benchmarks. The results show that EoP can significantly improve performance. For instance, compared to the non-commutative baseline PHP, with GPT-3.5-Turbo and EoP, we observe a 3.6% improvement on AQuA (60.6% to 64.2%), while GPT-4-powered EoP demonstrates a 7.7% overall accuracy enhancement on Math (53.9% to 61.6%) and a 3.5% improvement on OlympiadBench Maths (43.5% to 47.0%) when using Qwen-2.5-72b. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 04:43:15 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Sun",
"Lin",
""
],
[
"Zhang",
"Can",
""
]
] |
2506.03574 | Meng Li | Meng Li, Zhen Zhao, Zhengping Che, Fei Liao, Kun Wu, Zhiyuan Xu, Pei Ren, Zhao Jin, Ning Liu, Jian Tang | SwitchVLA: Execution-Aware Task Switching for Vision-Language-Action Models | Website: https://switchvla.github.io | null | null | null | cs.RO | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Robots deployed in dynamic environments must be able to not only follow diverse language instructions but flexibly adapt when user intent changes mid-execution. While recent Vision-Language-Action (VLA) models have advanced multi-task learning and instruction following, they typically assume static task intent, failing to respond when new instructions arrive during ongoing execution. This limitation hinders natural and robust interaction in dynamic settings, such as retail or household environments, where real-time intent changes are common. We propose SwitchVLA, a unified, execution-aware framework that enables smooth and reactive task switching without external planners or additional switch-specific data. We model task switching as a behavior modulation problem conditioned on execution state and instruction context. Expert demonstrations are segmented into temporally grounded contact phases, allowing the policy to infer task progress and adjust its behavior accordingly. A multi-behavior conditional policy is then trained to generate flexible action chunks under varying behavior modes through conditioned trajectory modeling. Experiments in both simulation and real-world robotic manipulation demonstrate that SwitchVLA enables robust instruction adherence, fluid task switching, and strong generalization-outperforming prior VLA baselines in both task success rate and interaction naturalness. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 04:45:24 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Li",
"Meng",
""
],
[
"Zhao",
"Zhen",
""
],
[
"Che",
"Zhengping",
""
],
[
"Liao",
"Fei",
""
],
[
"Wu",
"Kun",
""
],
[
"Xu",
"Zhiyuan",
""
],
[
"Ren",
"Pei",
""
],
[
"Jin",
"Zhao",
""
],
[
"Liu",
"Ning",
""
],
[
"Tang",
"Jian",
""
]
] |
2506.03575 | Ryan Russell | Ryan L. Russell, Moritz Merklein, Choon Kong Lai, Cong Tinh Bui, Alvaro Casas-Bedoya, Duk-Yong Choi, Stephen J. Madden, Benjamin J. Eggleton | Brillouin lasers in Bragg grating microresonators | null | null | null | null | physics.optics | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Chip-scale coherent light sources are required in applications spanning metrology and sensing to telecommunications. Brillouin lasers (BLs) offer a route to ultra-coherent optical sources in compact microresonators with free spectral range (FSR) matched to the Brillouin frequency shift (BFS). However, BFS - FSR matching typically facilitates cascaded Brillouin scattering, constraining achievable BL output power and coherence. Here, we demonstrate inhibition of cascading in a planar-integrated chalcogenide microresonator by exploiting the photonic bandgap (PBG) associated with a post-fabrication inscribed, reconfigurable intracavity Bragg grating. The PBG inhibits energy transfer within the target Brillouin scattering pathway, such as from pump to first-order Stokes wave. As a quantitative measure of Brillouin scattering inhibition, we report at least six-fold increase in threshold for onset of BL oscillation, which is ultimately limited by thermorefraction. For on-chip pump power of 399 mW, sufficient for a tenth-order Brillouin cascade, complete inhibition was achieved. Our work positions Bragg grating microresonators as an enabling platform for high performance on-chip BL sources, with reconfigurable modes of operation. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 04:45:44 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Russell",
"Ryan L.",
""
],
[
"Merklein",
"Moritz",
""
],
[
"Lai",
"Choon Kong",
""
],
[
"Bui",
"Cong Tinh",
""
],
[
"Casas-Bedoya",
"Alvaro",
""
],
[
"Choi",
"Duk-Yong",
""
],
[
"Madden",
"Stephen J.",
""
],
[
"Eggleton",
"Benjamin J.",
""
]
] |
2506.03576 | Zirui Chen | Zirui Chen, Xin Wang, Zhao Li, Wenbin Guo, Dongxiao He | KG-BiLM: Knowledge Graph Embedding via Bidirectional Language Models | null | null | null | null | cs.CL cs.AI | http://creativecommons.org/licenses/by-nc-nd/4.0/ | Recent advances in knowledge representation learning (KRL) highlight the urgent necessity to unify symbolic knowledge graphs (KGs) with language models (LMs) for richer semantic understanding. However, existing approaches typically prioritize either graph structure or textual semantics, leaving a gap: a unified framework that simultaneously captures global KG connectivity, nuanced linguistic context, and discriminative reasoning semantics. To bridge this gap, we introduce KG-BiLM, a bidirectional LM framework that fuses structural cues from KGs with the semantic expressiveness of generative transformers. KG-BiLM incorporates three key components: (i) Bidirectional Knowledge Attention, which removes the causal mask to enable full interaction among all tokens and entities; (ii) Knowledge-Masked Prediction, which encourages the model to leverage both local semantic contexts and global graph connectivity; and (iii) Contrastive Graph Semantic Aggregation, which preserves KG structure via contrastive alignment of sampled sub-graph representations. Extensive experiments on standard benchmarks demonstrate that KG-BiLM outperforms strong baselines in link prediction, especially on large-scale graphs with complex multi-hop relations - validating its effectiveness in unifying structural information and textual semantics. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 04:47:24 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Chen",
"Zirui",
""
],
[
"Wang",
"Xin",
""
],
[
"Li",
"Zhao",
""
],
[
"Guo",
"Wenbin",
""
],
[
"He",
"Dongxiao",
""
]
] |
2506.03577 | Yanhui Qu | Bernard Helffer, Qinghui Liu, Yanhui Qu, Qi Zhou | Cantor spectrum for multidimensional quasi-periodic Schr\"odinger operators | null | null | null | null | math.SP math-ph math.DS math.MG math.MP | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In this paper, we investigate the spectrum of a class of multidimensional quasi-periodic Schr\"odinger operators that exhibit a Cantor spectrum, which provides a resolution to a question posed by Damanik, Fillman, and Gorodetski \cite{DFG}. Additionally, we prove that for a dense set of irrational frequencies with positive Hausdorff dimension, the Hausdorff (and upper box) dimension of the spectrum of the critical almost Mathieu operator is positive, yet can be made arbitrarily small. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 04:53:00 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Helffer",
"Bernard",
""
],
[
"Liu",
"Qinghui",
""
],
[
"Qu",
"Yanhui",
""
],
[
"Zhou",
"Qi",
""
]
] |
2506.03578 | Chandra Shahi | Ashesh Giri, Chandra Shahi, and Adrienn Ruzsinszky | The isostructural alpha-gamma phase transition in cerium from the perspective of meta-generalized gradient approximations | 5 pages, 4 figures | null | null | null | cond-mat.mtrl-sci | http://creativecommons.org/licenses/by/4.0/ | Meta-generalized gradient approximations (meta-GGAs) on the third rung of the functional hierarchy are gaining increasing relevance for the electronic structure. Meta-GGAs are constructed from numerous ingredients including the orbital kinetic energy density that make them more flexible than generalized gradient approximations (GGAs) including the heavily used PBE-GGA. Still, most meta-GGAs cope with the expected limitations of a semilocal density functional when band gaps or localization of electrons are needed. On the other hand, meta-GGAs are implicit functionals of the orbitals. This feature resembles hybrid density functionals with exact exchange. Efforts in recent years demonstrate that some meta-GGAs can rise beyond the accuracy of semilocal approximation when band gaps are computed. Cerium is an ideal testbed to challenge some recent meta-GGAs. Cerium shows an isostructural alpha - gamma phase transition with delocalized and localized f electrons in each phase, respectively. Since the phonon entropy term was found negligible in the alpha - gamma phase transition of cerium by accurate experiments, all changes in the transition are driven by electronic correlation. The correlation of f electron systems is hardly captured by semilocal approximations but the recent LAK meta-GGA with ultranonlocality steps out of the framework of conventional semilocal density functionals and delivers spectacular accuracy for the phase transition of cerium. LAK and further meta-GGAs inspired by the success of LAK can open a forefront of meta-GGAs for quantum materials with localized electrons. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 04:53:32 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Giri",
"Ashesh",
""
],
[
"Shahi",
"Chandra",
""
],
[
"Ruzsinszky",
"Adrienn",
""
]
] |
2506.03579 | Tianshu Jiang | Tianshu Jiang and Jiayu Li | Kobayashi-Hitchin Correspondence for Saturated Reflexive Parabolic Sheaves on K\"ahler manifolds | null | null | null | null | math.DG math.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In this paper, we study the Kobayashi-Hitchin correspondence in the setting of parabolic sheaves with a simple normal crossing divisor over a compact K\"ahler manifold using the method of Hermitian-Yang-Mills flow. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 05:02:33 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Jiang",
"Tianshu",
""
],
[
"Li",
"Jiayu",
""
]
] |
2506.03580 | Florian Boudin | Enrico Benedetti, Akiko Aizawa, Florian Boudin | Automatically Suggesting Diverse Example Sentences for L2 Japanese Learners Using Pre-Trained Language Models | Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 4: Student Research Workshop) | null | null | null | cs.CL | http://creativecommons.org/licenses/by/4.0/ | Providing example sentences that are diverse and aligned with learners' proficiency levels is essential for fostering effective language acquisition. This study examines the use of Pre-trained Language Models (PLMs) to produce example sentences targeting L2 Japanese learners. We utilize PLMs in two ways: as quality scoring components in a retrieval system that draws from a newly curated corpus of Japanese sentences, and as direct sentence generators using zero-shot learning. We evaluate the quality of sentences by considering multiple aspects such as difficulty, diversity, and naturalness, with a panel of raters consisting of learners of Japanese, native speakers -- and GPT-4. Our findings suggest that there is inherent disagreement among participants on the ratings of sentence qualities, except for difficulty. Despite that, the retrieval approach was preferred by all evaluators, especially for beginner and advanced target proficiency, while the generative approaches received lower scores on average. Even so, our experiments highlight the potential for using PLMs to enhance the adaptability of sentence suggestion systems and therefore improve the language learning journey. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 05:13:05 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Benedetti",
"Enrico",
""
],
[
"Aizawa",
"Akiko",
""
],
[
"Boudin",
"Florian",
""
]
] |
2506.03581 | Jade Powell | Jade Powell, Bernhard M\"uller | Gravitational waves from core-collapse supernovae with no electromagnetic counterparts | Submitted to Classical and Quantum Gravity focus issue Gravitational Wave Physics and Astrophysics Ten Years After GW150914 | null | null | null | astro-ph.HE gr-qc | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Core-collapse supernovae (CCSNe) are regularly observed electromagnetically, prompting targetted searches for their gravitational-wave emission. However, there are scenarios where these powerful explosions may not have any observable electromagnetic signal, but would still have strong detectable emission in gravitational waves and neutrinos. A regular CCSN explosion may be obscured by matter in the Galaxy. A star may undergo a failed CCSN explosion, where the stalled shockwave is not revived, and would eventually form a black hole. Higher mass progenitor stars may revive the shock, but form a black hole too quickly for the shockwave to reach the surface of the star and produce an electromagnetic signal. Previous work has shown that we can determine if a black hole forms from the CCSN neutrino emission if there are long duration sinusoidal modulations in the neutrino signal caused by the standing accretion shock instability (SASI). The SASI also produces an observable signature in the gravitational-wave emission. In this paper, we investigate if we can distinguish between different scenarios for electromagnetically dark CCSNe using the gravitational-wave emission alone. We find, using a reconstruction of the SASI mode, abrupt end times of the gravitational-wave emission, and the rate of change of frequency of the dominant mode, that we are able to accurately distinguish between an obscured CCSN, a failed CCSN, and an explosion with fast black hole formation. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 05:22:41 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Powell",
"Jade",
""
],
[
"Müller",
"Bernhard",
""
]
] |
2506.03582 | Rui Yann | Rui Yann, Xianglei Xing | ViTSGMM: A Robust Semi-Supervised Image Recognition Network Using Sparse Labels | null | null | null | null | cs.CV cs.AI cs.LG | http://creativecommons.org/licenses/by/4.0/ | We present ViTSGMM, an image recognition network that leverages semi-supervised learning in a highly efficient manner. Existing works often rely on complex training techniques and architectures, while their generalization ability when dealing with extremely limited labeled data remains to be improved. To address these limitations, we construct a hierarchical mixture density classification decision mechanism by optimizing mutual information between feature representations and target classes, compressing redundant information while retaining crucial discriminative components. Experimental results demonstrate that our method achieves state-of-the-art performance on STL-10 and CIFAR-10/100 datasets when using negligible labeled samples. Notably, this paper also reveals a long-overlooked data leakage issue in the STL-10 dataset for semi-supervised learning tasks and removes duplicates to ensure the reliability of experimental results. Code available at https://github.com/Shu1L0n9/ViTSGMM. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 05:24:28 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Yann",
"Rui",
""
],
[
"Xing",
"Xianglei",
""
]
] |
2506.03583 | Zhigang Yang | Zhigang Yang and Huiguang Yao and Linmao Tian and Xuezhi Zhao and Qiang Li and Qi Wang | A Large-Scale Referring Remote Sensing Image Segmentation Dataset and Benchmark | null | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Referring Remote Sensing Image Segmentation is a complex and challenging task that integrates the paradigms of computer vision and natural language processing. Existing datasets for RRSIS suffer from critical limitations in resolution, scene diversity, and category coverage, which hinders the generalization and real-world applicability of refer segmentation models. To facilitate the development of this field, we introduce NWPU-Refer, the largest and most diverse RRSIS dataset to date, comprising 15,003 high-resolution images (1024-2048px) spanning 30+ countries with 49,745 annotated targets supporting single-object, multi-object, and non-object segmentation scenarios. Additionally, we propose the Multi-scale Referring Segmentation Network (MRSNet), a novel framework tailored for the unique demands of RRSIS. MRSNet introduces two key innovations: (1) an Intra-scale Feature Interaction Module (IFIM) that captures fine-grained details within each encoder stage, and (2) a Hierarchical Feature Interaction Module (HFIM) to enable seamless cross-scale feature fusion, preserving spatial integrity while enhancing discriminative power. Extensive experiments conducte on the proposed NWPU-Refer dataset demonstrate that MRSNet achieves state-of-the-art performance across multiple evaluation metrics, validating its effectiveness. The dataset and code are publicly available at https://github.com/CVer-Yang/NWPU-Refer. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 05:26:51 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Yang",
"Zhigang",
""
],
[
"Yao",
"Huiguang",
""
],
[
"Tian",
"Linmao",
""
],
[
"Zhao",
"Xuezhi",
""
],
[
"Li",
"Qiang",
""
],
[
"Wang",
"Qi",
""
]
] |
2506.03584 | Mike Ludkovski | Eduardo F. L. de Melo and Michael Ludkovski and Rodrigo S. Targino | Analyzing Pension Fund Mortality with Gaussian Processes in a Sub Population Framework | 34 pages, 15 figures | null | null | null | stat.AP | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Pension fund populations often have mortality experiences that are substantially different from the national benchmark. In a motivating case study of Brazilian corporate pension funds, pensioners are observed to have mortality that is 40-55% below the national average, due to the underlying socioeconomic disparities. Direct analysis of a pension fund population is challenging due to very sparse data, with age-specific annual death counts often in low single digits. We design and study a collection of stochastic sub-population frameworks that coherently capture and project pensioner mortality rates via deflator factors relative to a reference population. Superseding parametric approaches, we propose Gaussian process (GP) based models that flexibly estimate Age- and/or Year-specific deflators. We demonstrate that the GP models achieve better goodness of fit and uncertainty quantification. Our models are illustrated on two Brazilian pension funds in the context of exogenous national and insurance industry mortality tables. The GP models are implemented in R Stan using a fully Bayesian approach and take into account over-dispersion relative to the Poisson likelihood. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 05:33:00 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"de Melo",
"Eduardo F. L.",
""
],
[
"Ludkovski",
"Michael",
""
],
[
"Targino",
"Rodrigo S.",
""
]
] |
2506.03585 | Inseok Yeo | Inseok Yeo, Duksan Ryu, Jongmoon Baik | Improving LLM-Based Fault Localization with External Memory and Project Context | 12 Pages, 7 figures | null | null | null | cs.SE | http://creativecommons.org/licenses/by/4.0/ | Fault localization, the process of identifying the software components responsible for failures, is essential but often time-consuming. Recent advances in Large Language Models (LLMs) have enabled fault localization without extensive defect datasets or model fine-tuning. However, existing LLM-based methods rely only on general LLM capabilities and lack integration of project-specific knowledge, resulting in limited effectiveness, especially for complex software.
We introduce MemFL, a novel approach that enhances LLM-based fault localization by integrating project-specific knowledge via external memory. This memory includes static summaries of the project and dynamic, iterative debugging insights gathered from previous attempts. By leveraging external memory, MemFL simplifies debugging into three streamlined steps, significantly improving efficiency and accuracy. Iterative refinement through dynamic memory further enhances reasoning quality over time.
Evaluated on the Defects4J benchmark, MemFL using GPT-4o-mini localized 12.7% more bugs than current LLM-based methods, achieving this improvement with just 21% of the execution time (17.4 seconds per bug) and 33% of the API cost (0.0033 dollars per bug). On complex projects, MemFL's advantage increased to 27.6%. Additionally, MemFL with GPT-4.1-mini outperformed existing methods by 24.4%, requiring only 24.7 seconds and 0.0094 dollars per bug. MemFL thus demonstrates significant improvements by effectively incorporating project-specific knowledge into LLM-based fault localization, delivering high accuracy with reduced time and cost. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 05:33:32 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Yeo",
"Inseok",
""
],
[
"Ryu",
"Duksan",
""
],
[
"Baik",
"Jongmoon",
""
]
] |
2506.03586 | Yu Ma | Yu Ma, Chongtao Guo, Le Liang, Xiao Li, Shi Jin | Joint Beamforming and Resource Allocation for Delay Optimization in RIS-Assisted OFDM Systems: A DRL Approach | null | null | null | null | cs.AI cs.IT math.IT | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | This paper investigates a joint phase design and resource allocation problem in downlink reconfigurable intelligent surface (RIS)-assisted orthogonal frequency division multiplexing (OFDM) systems to optimize average delay, where data packets for each user arrive at the base station stochastically. The sequential optimization problem is inherently a Markov decision process (MDP), making it fall within the scope of reinforcement learning. To effectively handle the mixed action space and reduce the state space dimensionality, a hybrid deep reinforcement learning (DRL) approach is proposed. Specifically, proximal policy optimization (PPO)-$\Theta$ is employed to optimize RIS phase shift design, while PPO-N is responsible for subcarrier allocation decisions. To further mitigate the curse of dimensionality associated with subcarrier allocation, a multi-agent strategy is introduced to optimize subcarrier allocation indicater more efficiently. Moreover, to achieve more adaptive resource allocation and accurately capture network dynamics, key factors closely related to average delay, including the number of backlogged packets in buffers and the current packet arrivals, are incorporated into the state space. Furthermore, a transfer learning framework is introduced to enhance training efficiency and accelerate convergence. Simulation results demonstrate that the proposed algorithm significantly reduces average delay, enhances resource allocation efficiency, and achieves superior system robustness and fairness compared to baseline methods. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 05:33:33 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Ma",
"Yu",
""
],
[
"Guo",
"Chongtao",
""
],
[
"Liang",
"Le",
""
],
[
"Li",
"Xiao",
""
],
[
"Jin",
"Shi",
""
]
] |
2506.03587 | Florian Boudin | Florian Boudin, Akiko Aizawa | Preface to the Special Issue of the TAL Journal on Scholarly Document Processing | null | Traitement Automatique des Langues (TAL), volume 25, n{\deg}2/2024 | null | null | cs.DL cs.CL | http://creativecommons.org/licenses/by/4.0/ | The rapid growth of scholarly literature makes it increasingly difficult for researchers to keep up with new knowledge. Automated tools are now more essential than ever to help navigate and interpret this vast body of information. Scientific papers pose unique difficulties, with their complex language, specialized terminology, and diverse formats, requiring advanced methods to extract reliable and actionable insights. Large language models (LLMs) offer new opportunities, enabling tasks such as literature reviews, writing assistance, and interactive exploration of research. This special issue of the TAL journal highlights research addressing these challenges and, more broadly, research on natural language processing and information retrieval for scholarly and scientific documents. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 05:35:39 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Boudin",
"Florian",
""
],
[
"Aizawa",
"Akiko",
""
]
] |
2506.03588 | Hiroki Shiraishi | Hiroki Shiraishi, Hisao Ishibuchi, Masaya Nakata | A Class Inference Scheme With Dempster-Shafer Theory for Learning Fuzzy-Classifier Systems | null | ACM Transactions on Evolutionary Learning and Optimization (2025) | 10.1145/3717613 | null | cs.LG cs.AI cs.NE | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The decision-making process significantly influences the predictions of machine learning models. This is especially important in rule-based systems such as Learning Fuzzy-Classifier Systems (LFCSs) where the selection and application of rules directly determine prediction accuracy and reliability. LFCSs combine evolutionary algorithms with supervised learning to optimize fuzzy classification rules, offering enhanced interpretability and robustness. Despite these advantages, research on improving decision-making mechanisms (i.e., class inference schemes) in LFCSs remains limited. Most LFCSs use voting-based or single-winner-based inference schemes. These schemes rely on classification performance on training data and may not perform well on unseen data, risking overfitting. To address these limitations, this article introduces a novel class inference scheme for LFCSs based on the Dempster-Shafer Theory of Evidence (DS theory). The proposed scheme handles uncertainty well. By using the DS theory, the scheme calculates belief masses (i.e., measures of belief) for each specific class and the ``I don't know'' state from each fuzzy rule and infers a class from these belief masses. Unlike the conventional schemes, the proposed scheme also considers the ``I don't know'' state that reflects uncertainty, thereby improving the transparency and reliability of LFCSs. Applied to a variant of LFCS (i.e., Fuzzy-UCS), the proposed scheme demonstrates statistically significant improvements in terms of test macro F1 scores across 30 real-world datasets compared to conventional voting-based and single-winner-based fuzzy inference schemes. It forms smoother decision boundaries, provides reliable confidence measures, and enhances the robustness and generalizability of LFCSs in real-world applications. Our implementation is available at https://github.com/YNU-NakataLab/jUCS. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 05:38:49 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Shiraishi",
"Hiroki",
""
],
[
"Ishibuchi",
"Hisao",
""
],
[
"Nakata",
"Masaya",
""
]
] |
2506.03589 | Huy Le | Huy Le, Nhat Chung, Tung Kieu, Anh Nguyen, Ngan Le | BiMa: Towards Biases Mitigation for Text-Video Retrieval via Scene Element Guidance | 22 pages, 14 figures | null | null | null | cs.CV cs.AI cs.CL | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Text-video retrieval (TVR) systems often suffer from visual-linguistic biases present in datasets, which cause pre-trained vision-language models to overlook key details. To address this, we propose BiMa, a novel framework designed to mitigate biases in both visual and textual representations. Our approach begins by generating scene elements that characterize each video by identifying relevant entities/objects and activities. For visual debiasing, we integrate these scene elements into the video embeddings, enhancing them to emphasize fine-grained and salient details. For textual debiasing, we introduce a mechanism to disentangle text features into content and bias components, enabling the model to focus on meaningful content while separately handling biased information. Extensive experiments and ablation studies across five major TVR benchmarks (i.e., MSR-VTT, MSVD, LSMDC, ActivityNet, and DiDeMo) demonstrate the competitive performance of BiMa. Additionally, the model's bias mitigation capability is consistently validated by its strong results on out-of-distribution retrieval tasks. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 05:40:54 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Le",
"Huy",
""
],
[
"Chung",
"Nhat",
""
],
[
"Kieu",
"Tung",
""
],
[
"Nguyen",
"Anh",
""
],
[
"Le",
"Ngan",
""
]
] |
2506.03590 | Minh Luu | Minh Luu, Surya Jasper, Khoi Le, Evan Pan, Michael Quinn, Aakash Tyagi, Jiang Hu | VCDiag: Classifying Erroneous Waveforms for Failure Triage Acceleration | null | null | null | null | cs.LG | http://creativecommons.org/licenses/by/4.0/ | Failure triage in design functional verification is critical but time-intensive, relying on manual specification reviews, log inspections, and waveform analyses. While machine learning (ML) has improved areas like stimulus generation and coverage closure, its application to RTL-level simulation failure triage, particularly for large designs, remains limited. VCDiag offers an efficient, adaptable approach using VCD data to classify failing waveforms and pinpoint likely failure locations. In the largest experiment, VCDiag achieves over 94% accuracy in identifying the top three most likely modules. The framework introduces a novel signal selection and statistical compression approach, achieving over 120x reduction in raw data size while preserving features essential for classification. It can also be integrated into diverse Verilog/SystemVerilog designs and testbenches. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 05:44:03 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Luu",
"Minh",
""
],
[
"Jasper",
"Surya",
""
],
[
"Le",
"Khoi",
""
],
[
"Pan",
"Evan",
""
],
[
"Quinn",
"Michael",
""
],
[
"Tyagi",
"Aakash",
""
],
[
"Hu",
"Jiang",
""
]
] |
2506.03591 | Zhang Jia Xing | Jiaxing Zhang, Xinyi Zeng, Hao Tang | Resolving Task Objective Conflicts in Unified Multimodal Understanding and Generation via Task-Aware Mixture-of-Experts | null | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Unified multimodal large language models (MLLMs) based on end-to-end autoregressive (AR) transformers effectively integrate both understanding and generation tasks within a single framework. However, intrinsic Task Objective Conflicts between high-level semantic abstraction in understanding and fine-grained detail preservation in generation pose significant challenges, often leading to suboptimal trade-offs and task interference. Existing solutions, such as decoupling shared visual encoders, fall short of fundamentally resolving these conflicts due to inherent AR architecture. In this paper, we propose a novel approach that decouples internal components of AR to resolve task objective conflicts. Specifically, we design UTAMoE, a Unified Task-Aware Mixture-of-Experts (MoE) framework that decouples internal AR modules via a Task-Aware MoE Layer to create task-specific optimization subpaths. To enhance task differentiation while maintaining overall coordination, we introduce a novel Two-Stage Training Strategy. Extensive experiments on multimodal benchmarks demonstrate that UTAMoE mitigates task objective conflicts, achieving state-of-the-art performance across various tasks. Visualizations and ablation studies further validate the effectiveness of our approach. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 05:44:21 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Zhang",
"Jiaxing",
""
],
[
"Zeng",
"Xinyi",
""
],
[
"Tang",
"Hao",
""
]
] |
2506.03592 | Viktor Hangya | Viktor Hangya, Fabian K\"uch, Darina Gold | From Understanding to Generation: An Efficient Shortcut for Evaluating Language Models | null | null | null | null | cs.CL | http://creativecommons.org/licenses/by/4.0/ | Iterative evaluation of LLMs during training is essential to ensure expected capability development, but can be time- and compute-intensive. While NLU tasks, where the model selects from fixed answer choices, are cheap to evaluate, essential capabilities like reasoning and code generation rely on the more time-consuming NLG (token-by-token generation) format. In this work, our aim is to decrease the computational burden of NLG benchmarks in order to enable monitoring crucial LLM capabilities during model training. We reformulate generative tasks into computationally cheaper NLU alternatives. We test the performance correlation between the original and reformulated tasks using 8 LMs of various sizes and 4 capabilities: mathematical reasoning, code generation, factual knowledge and reading comprehension. Our results show a strong correlation between task formats, supporting capability assessment via cheaper alternatives and achieving over 35x average reduction in evaluation time. We plan to publish our benchmark adaptions. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 05:46:40 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Hangya",
"Viktor",
""
],
[
"Küch",
"Fabian",
""
],
[
"Gold",
"Darina",
""
]
] |
2506.03593 | Ray Groshan | Ray Groshan, Michael Ginn, Alexis Palmer | Is linguistically-motivated data augmentation worth it? | Accepted to ACL 2025 Main. First two authors contributed equally | null | null | null | cs.CL | http://creativecommons.org/licenses/by/4.0/ | Data augmentation, a widely-employed technique for addressing data scarcity, involves generating synthetic data examples which are then used to augment available training data. Researchers have seen surprising success from simple methods, such as random perturbations from natural examples, where models seem to benefit even from data with nonsense words, or data that doesn't conform to the rules of the language. A second line of research produces synthetic data that does in fact follow all linguistic constraints; these methods require some linguistic expertise and are generally more challenging to implement. No previous work has done a systematic, empirical comparison of both linguistically-naive and linguistically-motivated data augmentation strategies, leaving uncertainty about whether the additional time and effort of linguistically-motivated data augmentation work in fact yields better downstream performance.
In this work, we conduct a careful and comprehensive comparison of augmentation strategies (both linguistically-naive and linguistically-motivated) for two low-resource languages with different morphological properties, Uspanteko and Arapaho. We evaluate the effectiveness of many different strategies and their combinations across two important sequence-to-sequence tasks for low-resource languages: machine translation and interlinear glossing. We find that linguistically-motivated strategies can have benefits over naive approaches, but only when the new examples they produce are not significantly unlike the training data distribution. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 05:48:20 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Groshan",
"Ray",
""
],
[
"Ginn",
"Michael",
""
],
[
"Palmer",
"Alexis",
""
]
] |
2506.03594 | Shengjie Lin | Shengjie Lin, Jiading Fang, Muhammad Zubair Irshad, Vitor Campagnolo Guizilini, Rares Andrei Ambrus, Greg Shakhnarovich, Matthew R. Walter | SplArt: Articulation Estimation and Part-Level Reconstruction with 3D Gaussian Splatting | https://github.com/ripl/splart | null | null | null | cs.GR cs.CV cs.LG cs.MM cs.RO | http://creativecommons.org/licenses/by/4.0/ | Reconstructing articulated objects prevalent in daily environments is crucial for applications in augmented/virtual reality and robotics. However, existing methods face scalability limitations (requiring 3D supervision or costly annotations), robustness issues (being susceptible to local optima), and rendering shortcomings (lacking speed or photorealism). We introduce SplArt, a self-supervised, category-agnostic framework that leverages 3D Gaussian Splatting (3DGS) to reconstruct articulated objects and infer kinematics from two sets of posed RGB images captured at different articulation states, enabling real-time photorealistic rendering for novel viewpoints and articulations. SplArt augments 3DGS with a differentiable mobility parameter per Gaussian, achieving refined part segmentation. A multi-stage optimization strategy is employed to progressively handle reconstruction, part segmentation, and articulation estimation, significantly enhancing robustness and accuracy. SplArt exploits geometric self-supervision, effectively addressing challenging scenarios without requiring 3D annotations or category-specific priors. Evaluations on established and newly proposed benchmarks, along with applications to real-world scenarios using a handheld RGB camera, demonstrate SplArt's state-of-the-art performance and real-world practicality. Code is publicly available at https://github.com/ripl/splart. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 05:53:16 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Lin",
"Shengjie",
""
],
[
"Fang",
"Jiading",
""
],
[
"Irshad",
"Muhammad Zubair",
""
],
[
"Guizilini",
"Vitor Campagnolo",
""
],
[
"Ambrus",
"Rares Andrei",
""
],
[
"Shakhnarovich",
"Greg",
""
],
[
"Walter",
"Matthew R.",
""
]
] |
2506.03595 | Runa Eschenhagen | Runa Eschenhagen, Aaron Defazio, Tsung-Hsien Lee, Richard E. Turner, Hao-Jun Michael Shi | Purifying Shampoo: Investigating Shampoo's Heuristics by Decomposing its Preconditioner | null | null | null | null | cs.LG cs.AI stat.ML | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The recent success of Shampoo in the AlgoPerf contest has sparked renewed interest in Kronecker-factorization-based optimization algorithms for training neural networks. Despite its success, Shampoo relies heavily on several heuristics such as learning rate grafting and stale preconditioning to achieve performance at-scale. These heuristics increase algorithmic complexity, necessitate further hyperparameter tuning, and lack theoretical justification. This paper investigates these heuristics from the angle of Frobenius norm approximation to full-matrix Adam and decouples the preconditioner's eigenvalues and eigenbasis updates. We show that grafting from Adam mitigates the staleness and mis-scaling of the preconditioner's eigenvalues and how correcting the eigenvalues directly can eliminate the need for learning rate grafting. To manage the error induced by infrequent eigenbasis computations, we propose an adaptive criterion for determining the eigenbasis computation frequency motivated by terminating a warm-started QR algorithm. This criterion decouples the update frequency of different preconditioner matrices and enables us to investigate the impact of approximation error on convergence. These practical techniques offer a principled angle towards removing Shampoo's heuristics and developing improved Kronecker-factorization-based training algorithms. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 05:55:41 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Eschenhagen",
"Runa",
""
],
[
"Defazio",
"Aaron",
""
],
[
"Lee",
"Tsung-Hsien",
""
],
[
"Turner",
"Richard E.",
""
],
[
"Shi",
"Hao-Jun Michael",
""
]
] |
2506.03596 | Feng Han | Feng Han, Yang Jiao, Shaoxiang Chen, Junhao Xu, Jingjing Chen, Yu-Gang Jiang | ControlThinker: Unveiling Latent Semantics for Controllable Image Generation through Visual Reasoning | null | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The field of controllable image generation has seen significant advancements, with various architectures improving generation layout consistency with control signals. However, contemporary methods still face challenges in bridging the semantic gap between input text prompts with sparse semantics and the target images, often over-relying on low-level control signals to infer regional details. To address this challenge, we propose ControlThinker, a novel framework that employs a "comprehend-then-generate" paradigm. Firstly, by incentivizing the visual reasoning capability of a MLLM, latent semantics from control images are mined to enrich text prompts. This enriched semantic understanding then seamlessly aids in image generation without the need for additional complex modifications. To further tackle the uncertainty arising from the ambiguity of control images, we encourage broader exploration of reasoning trajectories and select the optimal one using a metric-based output reward model (ORM). Extensive experimental results demonstrate that ControlThinker effectively mitigates the semantic gap between raw text prompts and target images, resulting in improved visual quality and semantic consistency across a wide range of benchmarks. The code and models are available at https://github.com/Maplebb/ControlThinker. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 05:56:19 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Han",
"Feng",
""
],
[
"Jiao",
"Yang",
""
],
[
"Chen",
"Shaoxiang",
""
],
[
"Xu",
"Junhao",
""
],
[
"Chen",
"Jingjing",
""
],
[
"Jiang",
"Yu-Gang",
""
]
] |
2506.03597 | Michael Lam | Michael T. Lam, David L. Kaplan, Gabriella Agazie, Akash Anumarlapudi, Anne M. Archibald, Zaven Arzoumanian, Paul T. Baker, Paul R. Brook, H. Thankful Cromartie, Kathryn Crowter, Megan E. DeCesar, Paul B. Demorest, Timothy Dolch, Elizabeth C. Ferrara, William Fiore, Emmanuel Fonseca, Gabriel E. Freedman, Nate Garver-Daniels, Peter A. Gentile, Joseph Glaser, Deborah C. Good, Jeffrey S. Hazboun, Ross J. Jennings, Megan L. Jones, Matthew Kerr, Duncan R. Lorimer, Jing Luo, Ryan S. Lynch, Alexander McEwen, Maura A. McLaughlin, Natasha McMann, Bradley W. Meyers, Cherry Ng, David J. Nice, Timothy T. Pennucci, Benetge B. P. Perera, Nihan S. Pol, Henri A. Radovan, Scott M. Ransom, Paul S. Ray, Ann Schmiedekamp, Carl Schmiedekamp, Brent J. Shapiro-Albert, Joseph Simon, Ingrid H. Stairs, Kevin Stovall, Abhimanyu Susobhanan, Joseph K. Swiggum, Haley M. Wahl | The NANOGrav 15-Year Data Set: A Case Study for Simplified Dispersion Measure Modeling for PSR J1455-3330 and the Impact on Gravitational Wave Sensitivity | 11 pages, 6 figures, submitted to ApJ | null | null | null | astro-ph.HE | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Evidence for a low-frequency gravitational-wave background using pulsar timing arrays has generated recent interest into its underlying contributing sources. However, multiple investigations have seen that the significance of the evidence does not change with choice of pulsar modeling techniques but the resulting parameters from the gravitational wave searches do. PSR J1455-3330 is one of the longest-observed pulsars in the array monitored by the North American Nanohertz Observatory for Gravitational Waves (NANOGrav) but showed no evidence for long-timescale red noise, either intrinsic or the common signal found among many pulsars in the array. In this work, we argue that NANOGrav's piecewise-constant function used to model variations in radio-frequency-dependent dispersive delay should not be used for this pulsar, and a much simpler physical model of a fixed solar wind density plus a linear trend in dispersion measure is preferred. When the original model is replaced, (i) the pulsar's timing parallax signal changes from an upper limit to a significant detection, (ii) red noise becomes significant, and (iii) the red noise is consistent with the common signal found for the other pulsars. Neither of these signals are radio-frequency dependent. While the same physical motivation will not apply to many of the pulsars currently used in pulsar timing arrays, we argue for careful physically-motivated timing and noise modeling of pulsars used in precision timing experiments. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 05:59:53 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Lam",
"Michael T.",
""
],
[
"Kaplan",
"David L.",
""
],
[
"Agazie",
"Gabriella",
""
],
[
"Anumarlapudi",
"Akash",
""
],
[
"Archibald",
"Anne M.",
""
],
[
"Arzoumanian",
"Zaven",
""
],
[
"Baker",
"Paul T.",
""
],
[
"Brook",
"Paul R.",
""
],
[
"Cromartie",
"H. Thankful",
""
],
[
"Crowter",
"Kathryn",
""
],
[
"DeCesar",
"Megan E.",
""
],
[
"Demorest",
"Paul B.",
""
],
[
"Dolch",
"Timothy",
""
],
[
"Ferrara",
"Elizabeth C.",
""
],
[
"Fiore",
"William",
""
],
[
"Fonseca",
"Emmanuel",
""
],
[
"Freedman",
"Gabriel E.",
""
],
[
"Garver-Daniels",
"Nate",
""
],
[
"Gentile",
"Peter A.",
""
],
[
"Glaser",
"Joseph",
""
],
[
"Good",
"Deborah C.",
""
],
[
"Hazboun",
"Jeffrey S.",
""
],
[
"Jennings",
"Ross J.",
""
],
[
"Jones",
"Megan L.",
""
],
[
"Kerr",
"Matthew",
""
],
[
"Lorimer",
"Duncan R.",
""
],
[
"Luo",
"Jing",
""
],
[
"Lynch",
"Ryan S.",
""
],
[
"McEwen",
"Alexander",
""
],
[
"McLaughlin",
"Maura A.",
""
],
[
"McMann",
"Natasha",
""
],
[
"Meyers",
"Bradley W.",
""
],
[
"Ng",
"Cherry",
""
],
[
"Nice",
"David J.",
""
],
[
"Pennucci",
"Timothy T.",
""
],
[
"Perera",
"Benetge B. P.",
""
],
[
"Pol",
"Nihan S.",
""
],
[
"Radovan",
"Henri A.",
""
],
[
"Ransom",
"Scott M.",
""
],
[
"Ray",
"Paul S.",
""
],
[
"Schmiedekamp",
"Ann",
""
],
[
"Schmiedekamp",
"Carl",
""
],
[
"Shapiro-Albert",
"Brent J.",
""
],
[
"Simon",
"Joseph",
""
],
[
"Stairs",
"Ingrid H.",
""
],
[
"Stovall",
"Kevin",
""
],
[
"Susobhanan",
"Abhimanyu",
""
],
[
"Swiggum",
"Joseph K.",
""
],
[
"Wahl",
"Haley M.",
""
]
] |
2506.03598 | Ze Tong Tang | Zetong Tang, Qian Ma, Di Wu | Auto prompt sql: a resource-efficient architecture for text-to-sql translation in constrained environments | 4 pages,2 figures,EITCE 2025 | null | null | null | cs.CL cs.AI | http://creativecommons.org/licenses/by/4.0/ | Using the best Text-to-SQL methods in resource-constrained environments is challenging due to their reliance on resource-intensive open-source models. This paper introduces Auto Prompt SQL(AP-SQL), a novel architecture designed to bridge the gap between resource-efficient small open-source models and the powerful capabilities of large closed-source models for Text-to-SQL translation. Our method decomposes the task into schema filtering, retrieval-augmented text-to-SQL generation based on in-context examples, and prompt-driven schema linking and SQL generation. To improve schema selection accuracy, we fine-tune large language models. Crucially, we also explore the impact of prompt engineering throughout the process, leveraging Chain-of-Thought(CoT) and Graph-of-Thought(GoT) templates to significantly enhance the model's reasoning for accurate SQL generation. Comprehensive evaluations on the Spider benchmarks demonstrate the effectiveness of AP-SQL. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 06:04:46 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Tang",
"Zetong",
""
],
[
"Ma",
"Qian",
""
],
[
"Wu",
"Di",
""
]
] |
2506.03599 | Asher Spector | Asher Spector, Rina Foygel Barber, Emmanuel Cand\`es | Mosaic inference on panel data | 38 pages, 7 figures | null | null | null | stat.ME math.ST stat.TH | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Analysis of panel data via linear regression is widespread across disciplines. To perform statistical inference, such analyses typically assume that clusters of observations are jointly independent. For example, one might assume that observations in New York are independent of observations in New Jersey. Are such assumptions plausible? Might there be hidden dependencies between nearby clusters? This paper introduces a mosaic permutation test that can (i) test the cluster-independence assumption and (ii) produce confidence intervals for linear models without assuming the full cluster-independence assumption. The key idea behind our method is to apply a permutation test to carefully constructed residual estimates that obey the same invariances as the true errors. As a result, our method yields finite-sample valid inferences under a mild "local exchangeability" condition. This condition differs from the typical cluster-independence assumption, as neither assumption implies the other. Furthermore, our method is asymptotically valid under cluster-independence (with no exchangeability assumptions). Together, these results show our method is valid under assumptions that are arguably weaker than the assumptions underlying many classical methods. In experiments on well-studied datasets from the literature, we find that many existing methods produce variance estimates that are up to five times too small, whereas mosaic methods produce reliable results. We implement our methods in the python package mosaicperm. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 06:14:22 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Spector",
"Asher",
""
],
[
"Barber",
"Rina Foygel",
""
],
[
"Candès",
"Emmanuel",
""
]
] |
2506.03600 | S. Shankaranarayanan | Tharaka Alapati and S. Shankaranarayanan (IIT Bombay) | Axial quasi-normal modes of slowly rotating black holes in dynamical Chern-Simons gravity to second-order in spin and coupling | 19 Pages, 3 Figures, Comments welcome | null | null | null | gr-qc hep-th | http://creativecommons.org/licenses/by/4.0/ | We compute the quasi-normal mode (QNM) frequencies of slowly rotating black holes in dynamical Chern-Simons (dCS) gravity, including corrections up to second order in the black hole's dimensionless spin parameter ($\chi$) and second order in the dCS coupling parameter ($\alpha$). Due to the complexities of constructing a Newman-Penrose tetrad at this order, we employ a metric perturbation approach. We derive a system of coupled ordinary differential equations for the primary axial $l$-mode and the polar $l\pm 1$ modes, which is then solved numerically with appropriate ingoing and outgoing wave boundary conditions. Our numerical framework is validated in the General Relativistic limit against known Schwarzschild QNMs and highly accurate Kerr QNM results for $\chi \leq 0.15$. For the fundamental $n=0, l=m=2$ axial mode, we present detailed numerical results illustrating the dependence of QNM frequencies on both $\chi$ and $\alpha$. We observe that while rotation generally increases the damping time, increasing the dCS coupling parameter significantly reduces the damping time of the axial mode. This finding contrasts with previous analytical work on polar modes, which suggested an increase in damping time due to dCS effects, highlighting a crucial parity-dependent difference in how dCS gravity impacts black hole ringdowns. Furthermore, we provide an analytical fitting formula for this mode. These results, incorporating coupled spin and dCS effects at second order, provide more accurate theoretical predictions for testing dCS gravity with gravitational wave observations of black hole ringdowns. The refined QNM calculations are particularly relevant for lower-mass black hole merger events, such as GW230529, where dCS corrections may be more prominent and their distinct damping signatures could be observable. [Abridged Version] | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 06:14:27 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Alapati",
"Tharaka",
"",
"IIT Bombay"
],
[
"Shankaranarayanan",
"S.",
"",
"IIT Bombay"
]
] |
2506.03601 | Masatoshi Yamada | Nobuyoshi Ohta and Masatoshi Yamada | Essential Renormalization Group Equation for Gravity coupled to a Scalar field | 31 pages | null | null | NITEP 254 | hep-th gr-qc | http://creativecommons.org/licenses/by/4.0/ | We study the essential renormalization group equation, in which inessential couplings are removed via field redefinitions, for Einstein gravity coupled to a massive scalar field in the presence of a cosmological constant. Our results indicate that perturbatively nonrenormalizable terms can be eliminated due to the cosmological term, in contrast to the case of perturbation around flat spacetime. We find a nontrivial fixed point for the Newton coupling and the cosmological term. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 06:17:13 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Ohta",
"Nobuyoshi",
""
],
[
"Yamada",
"Masatoshi",
""
]
] |
2506.03602 | Hiroki Shiraishi | Hiroki Shiraishi, Yohei Hayamizu, Tomonori Hashiyama, Keiki Takadama, Hisao Ishibuchi, Masaya Nakata | Adapting Rule Representation With Four-Parameter Beta Distribution for Learning Classifier Systems | null | IEEE Transactions on Evolutionary Computation (2025) | 10.1109/TEVC.2025.3550915 | null | cs.LG cs.AI cs.NE | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Rule representations significantly influence the search capabilities and decision boundaries within the search space of Learning Classifier Systems (LCSs), a family of rule-based machine learning systems that evolve interpretable models through evolutionary processes. However, it is very difficult to choose an appropriate rule representation for each problem. Additionally, some problems benefit from using different representations for different subspaces within the input space. Thus, an adaptive mechanism is needed to choose an appropriate rule representation for each rule in LCSs. This article introduces a flexible rule representation using a four-parameter beta distribution and integrates it into a fuzzy-style LCS. The four-parameter beta distribution can form various function shapes, and this flexibility enables our LCS to automatically select appropriate representations for different subspaces. Our rule representation can represent crisp/fuzzy decision boundaries in various boundary shapes, such as rectangles and bells, by controlling four parameters, compared to the standard representations such as trapezoidal ones. Leveraging this flexibility, our LCS is designed to adapt the appropriate rule representation for each subspace. Moreover, our LCS incorporates a generalization bias favoring crisp rules where feasible, enhancing model interpretability without compromising accuracy. Experimental results on real-world classification tasks show that our LCS achieves significantly superior test accuracy and produces more compact rule sets. Our implementation is available at https://github.com/YNU-NakataLab/Beta4-UCS. An extended abstract related to this work is available at https://doi.org/10.36227/techrxiv.174900805.59801248/v1. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 06:19:49 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Shiraishi",
"Hiroki",
""
],
[
"Hayamizu",
"Yohei",
""
],
[
"Hashiyama",
"Tomonori",
""
],
[
"Takadama",
"Keiki",
""
],
[
"Ishibuchi",
"Hisao",
""
],
[
"Nakata",
"Masaya",
""
]
] |
2506.03603 | Alexander Scott | Maria Chudnovsky, Tung Nguyen, Alex Scott, Paul Seymour | The vertex sets of subtrees of a tree | null | null | null | null | math.CO | http://creativecommons.org/licenses/by/4.0/ | Let $\mathcal{F}$ be a set of subsets of a set $W$. When is there a tree $T$ with vertex set $W$ such that each member of $\mathcal{F}$ is the set of vertices of a subtree of $T$? It is necessary that $\mathcal{F}$ has the Helly property and the intersection graph of $\mathcal{F}$ is chordal. We will show that these two necessary conditions are together sufficient in the finite case, and more generally, they are sufficient if no element of $W$ belongs to infinitely many infinite sets in $\mathcal{F}$. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 06:24:53 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Chudnovsky",
"Maria",
""
],
[
"Nguyen",
"Tung",
""
],
[
"Scott",
"Alex",
""
],
[
"Seymour",
"Paul",
""
]
] |
2506.03604 | Luka Andren\v{s}ek | Luka Andrensek | A complete classification of endomorphisms of Kiselman's semigroup | 6 pages | null | null | null | math.GR | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Let $K_n$ denote Kiselman's semigroup. It was thoroughly studied in Kudryavtseva and Mazorchuk's paper On Kiselman's semigroup. The authors in that paper pose a question, whether it is possible to classify all endomorphisms of $K_n$ (Question 21). In this paper we give a complete classification and provide a boolean matrix semigroup that is isomorphic to End$(K_n)$. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 06:25:12 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Andrensek",
"Luka",
""
]
] |
2506.03605 | Tomoya Yoshida | Tomoya Yoshida, Shuhei Kurita, Taichi Nishimura, Shinsuke Mori | Generating 6DoF Object Manipulation Trajectories from Action Description in Egocentric Vision | CVPR 2025 | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Learning to use tools or objects in common scenes, particularly handling them in various ways as instructed, is a key challenge for developing interactive robots. Training models to generate such manipulation trajectories requires a large and diverse collection of detailed manipulation demonstrations for various objects, which is nearly unfeasible to gather at scale. In this paper, we propose a framework that leverages large-scale ego- and exo-centric video datasets -- constructed globally with substantial effort -- of Exo-Ego4D to extract diverse manipulation trajectories at scale. From these extracted trajectories with the associated textual action description, we develop trajectory generation models based on visual and point cloud-based language models. In the recently proposed egocentric vision-based in-a-quality trajectory dataset of HOT3D, we confirmed that our models successfully generate valid object trajectories, establishing a training dataset and baseline models for the novel task of generating 6DoF manipulation trajectories from action descriptions in egocentric vision. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 06:28:16 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Yoshida",
"Tomoya",
""
],
[
"Kurita",
"Shuhei",
""
],
[
"Nishimura",
"Taichi",
""
],
[
"Mori",
"Shinsuke",
""
]
] |
2506.03606 | Parismita Gogoi | Parismita Gogoi and Sishir Kalita and Wendy Lalhminghlui and Viyazonuo Terhiija and Moakala Tzudir and Priyankoo Sarmah and S. R. M. Prasanna | Tone recognition in low-resource languages of North-East India: peeling the layers of SSL-based speech models | Accepted in Interspeech2025 | null | null | null | eess.AS cs.AI cs.CL eess.SP | http://creativecommons.org/licenses/by/4.0/ | This study explores the use of self-supervised learning (SSL) models for tone recognition in three low-resource languages from North Eastern India: Angami, Ao, and Mizo. We evaluate four Wav2vec2.0 base models that were pre-trained on both tonal and non-tonal languages. We analyze tone-wise performance across the layers for all three languages and compare the different models. Our results show that tone recognition works best for Mizo and worst for Angami. The middle layers of the SSL models are the most important for tone recognition, regardless of the pre-training language, i.e. tonal or non-tonal. We have also found that the tone inventory, tone types, and dialectal variations affect tone recognition. These findings provide useful insights into the strengths and weaknesses of SSL-based embeddings for tonal languages and highlight the potential for improving tone recognition in low-resource settings. The source code is available at GitHub 1 . | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 06:32:12 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Gogoi",
"Parismita",
""
],
[
"Kalita",
"Sishir",
""
],
[
"Lalhminghlui",
"Wendy",
""
],
[
"Terhiija",
"Viyazonuo",
""
],
[
"Tzudir",
"Moakala",
""
],
[
"Sarmah",
"Priyankoo",
""
],
[
"Prasanna",
"S. R. M.",
""
]
] |
2506.03607 | Wing Man Casca Kwok | Wing Man Casca Kwok, Yip Chiu Tung, Kunal Bhagchandani | Analyzing Transformer Models and Knowledge Distillation Approaches for Image Captioning on Edge AI | null | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Edge computing decentralizes processing power to network edge, enabling real-time AI-driven decision-making in IoT applications. In industrial automation such as robotics and rugged edge AI, real-time perception and intelligence are critical for autonomous operations. Deploying transformer-based image captioning models at the edge can enhance machine perception, improve scene understanding for autonomous robots, and aid in industrial inspection.
However, these edge or IoT devices are often constrained in computational resources for physical agility, yet they have strict response time requirements. Traditional deep learning models can be too large and computationally demanding for these devices. In this research, we present findings of transformer-based models for image captioning that operate effectively on edge devices. By evaluating resource-effective transformer models and applying knowledge distillation techniques, we demonstrate inference can be accelerated on resource-constrained devices while maintaining model performance using these techniques. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 06:37:04 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Kwok",
"Wing Man Casca",
""
],
[
"Tung",
"Yip Chiu",
""
],
[
"Bhagchandani",
"Kunal",
""
]
] |
2506.03608 | Heng Yu | Di Fan, Heng Yu, Zhiyuan Xu | PDSE: A Multiple Lesion Detector for CT Images using PANet and Deformable Squeeze-and-Excitation Block | MIUA 2024 | null | null | null | cs.CV | http://creativecommons.org/licenses/by/4.0/ | Detecting lesions in Computed Tomography (CT) scans is a challenging task in medical image processing due to the diverse types, sizes, and locations of lesions. Recently, various one-stage and two-stage framework networks have been developed to focus on lesion localization. We introduce a one-stage lesion detection framework, PDSE, by redesigning Retinanet to achieve higher accuracy and efficiency for detecting lesions in multimodal CT images. Specifically, we enhance the path aggregation flow by incorporating a low-level feature map. Additionally, to improve model representation, we utilize the adaptive Squeeze-and-Excitation (SE) block and integrate channel feature map attention. This approach has resulted in achieving new state-of-the-art performance. Our method significantly improves the detection of small and multiscaled objects. When evaluated against other advanced algorithms on the public DeepLesion benchmark, our algorithm achieved an mAP of over 0.20. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 06:38:31 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Fan",
"Di",
""
],
[
"Yu",
"Heng",
""
],
[
"Xu",
"Zhiyuan",
""
]
] |
2506.03609 | William Campbell | William M. Campbell, Leonardo Mariani, Michael E. Tobar, Maxim Goryachev | Experimental Exclusion of Planetary Mass Primordial Black Hole Mergers | null | null | null | null | gr-qc | http://creativecommons.org/licenses/by/4.0/ | The multi-mode acoustic gravitational wave experiment (MAGE) is a high-frequency gravitational wave detection experiment featuring cryogenic quartz bulk acoustic wave resonators operating as sensitive strain antennas in the MHz regime. After 61 days of non-continuous data collection, we present bounds on the observable merger rate density of primordial black hole binary systems of chirp mass $1.2\times10^{-4}M_\odot<\mathcal{M}<1.7\times10^{-9}M_\odot$. The maximum achieved limit on the merger rate density is $\mathcal{R}>1.3\times10^{18}~\mathrm{kpc}^{-3}\mathrm{yr}^{-1}$ which corresponds to constraining yearly mergers to a distance of reach on the order of the solar system, or $1.0\times10^{-6}$ kpc during the observational period. In addition, we exclude significantly rare and strong events similar to those observed in previous predecessor experiments as non-gravitational background signals, utilising coincident analysis between multiple detectors. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 06:39:00 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Campbell",
"William M.",
""
],
[
"Mariani",
"Leonardo",
""
],
[
"Tobar",
"Michael E.",
""
],
[
"Goryachev",
"Maxim",
""
]
] |
2506.03610 | Dongmin Park | Dongmin Park, Minkyu Kim, Beongjun Choi, Junhyuck Kim, Keon Lee, Jonghyun Lee, Inkyu Park, Byeong-Uk Lee, Jaeyoung Hwang, Jaewoo Ahn, Ameya S. Mahabaleshwarkar, Bilal Kartal, Pritam Biswas, Yoshi Suhara, Kangwook Lee, Jaewoong Cho | Orak: A Foundational Benchmark for Training and Evaluating LLM Agents on Diverse Video Games | null | null | null | null | cs.AI | http://creativecommons.org/licenses/by/4.0/ | Large Language Model (LLM) agents are reshaping the game industry, particularly with more intelligent and human-preferable game characters. However, existing game benchmarks fall short of practical needs: they lack evaluations of diverse LLM capabilities across various game genres, studies of agentic modules crucial for complex gameplay, and fine-tuning datasets for aligning pre-trained LLMs into gaming agents. To fill these gaps, we present \textbf{\benchname{}}, a foundational benchmark designed to train and evaluate LLM agents across diverse real-world video games. Unlike existing benchmarks, Orak includes 12 popular video games spanning all major genres, enabling comprehensive studies of LLM capabilities and agentic modules essential for intricate game scenarios. To support consistent evaluation of LLMs, we introduce a plug-and-play interface based on Model Context Protocol (MCP) that enables LLMs to seamlessly connect with games and manipulate agentic modules. Additionally, we propose a fine-tuning dataset, consisting of LLM gameplay trajectories across diverse game genres. Orak offers a comprehensive evaluation framework, encompassing general game score leaderboards, LLM battle arenas, and in-depth analyses of visual input state, agentic strategies, and fine-tuning effects, establishing a foundation towards building generic gaming agents. Code is available at https://github.com/krafton-ai/Orak. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 06:40:33 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Park",
"Dongmin",
""
],
[
"Kim",
"Minkyu",
""
],
[
"Choi",
"Beongjun",
""
],
[
"Kim",
"Junhyuck",
""
],
[
"Lee",
"Keon",
""
],
[
"Lee",
"Jonghyun",
""
],
[
"Park",
"Inkyu",
""
],
[
"Lee",
"Byeong-Uk",
""
],
[
"Hwang",
"Jaeyoung",
""
],
[
"Ahn",
"Jaewoo",
""
],
[
"Mahabaleshwarkar",
"Ameya S.",
""
],
[
"Kartal",
"Bilal",
""
],
[
"Biswas",
"Pritam",
""
],
[
"Suhara",
"Yoshi",
""
],
[
"Lee",
"Kangwook",
""
],
[
"Cho",
"Jaewoong",
""
]
] |
2506.03611 | Xianhu Wu | Xianhu Wu, Guanglei Cui, Jieyu Bi, Gaojie Xia, Zewen Zuo, Min Gu | Gefitinib-Induced Interface Engineering Enhances the Defect Formation Energy for Highly Efficient and Stable Perovskite Solar Cells | null | null | null | null | cond-mat.mtrl-sci physics.app-ph | http://creativecommons.org/licenses/by/4.0/ | Poly(3,4-ethylenedioxythiophene):polystyrenesulfonate (PEDOT:PSS) has been widely used as a hole transport layer in perovskite solar cells (PSCs). However, the high interface defect density and energy level mismatch between PEDOT:PSS and perovskite can lead to significant open-circuit voltage loss. Additionally, the free PSS chains on the surface of PEDOT:PSS can absorb water molecules, promoting the degradation of perovskite at the PEDOT:PSS/perovskite interface. Here, gefitinib is used to modify the surface of PEDOT:PSS, removing a portion of the free PSS chains from the surface, reducing the PSS/PEDOT ratio, and enhancing the conductivity of PEDOT:PSS. Gefitinib has altered the energy level structure of PEDOT:PSS, facilitating hole transport at the interface. The Cl, F, and NH groups in gefitinib also passivated defects in the perovskite, reducing the defect density at the interface and significantly enhancing the stability of PSCs. This modification increased the open-circuit voltage from 1.077 to 1.110 V and the power conversion efficiency (PCE) from 17.01% to 19.63%. When gefitinib was used to modify the interface between SnO2 and perovskite, the PCE of PSCs (ITO/SnO2/perovskite/Spiro-OMETAD/Au) increased from 22.46% to 23.89%. This approach provides new perspectives and strategies for improving the efficiency and stability of PSCs. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 06:44:44 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Wu",
"Xianhu",
""
],
[
"Cui",
"Guanglei",
""
],
[
"Bi",
"Jieyu",
""
],
[
"Xia",
"Gaojie",
""
],
[
"Zuo",
"Zewen",
""
],
[
"Gu",
"Min",
""
]
] |
2506.03613 | Shaoshan Liu | Shaoshan Liu, Fan Wang, Hongjun Zhou, Yuanfeng Wang | Training Cross-Morphology Embodied AI Agents: From Practical Challenges to Theoretical Foundations | null | null | null | null | cs.AI cs.CC | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | While theory and practice are often seen as separate domains, this article shows that theoretical insight is essential for overcoming real-world engineering barriers. We begin with a practical challenge: training a cross-morphology embodied AI policy that generalizes across diverse robot morphologies. We formalize this as the Heterogeneous Embodied Agent Training (HEAT) problem and prove it reduces to a structured Partially Observable Markov Decision Process (POMDP) that is PSPACE-complete. This result explains why current reinforcement learning pipelines break down under morphological diversity, due to sequential training constraints, memory-policy coupling, and data incompatibility. We further explore Collective Adaptation, a distributed learning alternative inspired by biological systems. Though NEXP-complete in theory, it offers meaningful scalability and deployment benefits in practice. This work illustrates how computational theory can illuminate system design trade-offs and guide the development of more robust, scalable embodied AI. For practitioners and researchers to explore this problem, the implementation code of this work has been made publicly available at https://github.com/airs-admin/HEAT | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 06:44:49 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Liu",
"Shaoshan",
""
],
[
"Wang",
"Fan",
""
],
[
"Zhou",
"Hongjun",
""
],
[
"Wang",
"Yuanfeng",
""
]
] |
2506.03615 | Sarah Alyami | Sarah Alyami, Hamzah Luqman, Sadam Al-Azani, Maad Alowaifeer, Yazeed Alharbi, and Yaser Alonaizan | Isharah: A Large-Scale Multi-Scene Dataset for Continuous Sign Language Recognition | null | null | null | null | cs.CV | http://creativecommons.org/licenses/by-nc-nd/4.0/ | Current benchmarks for sign language recognition (SLR) focus mainly on isolated SLR, while there are limited datasets for continuous SLR (CSLR), which recognizes sequences of signs in a video. Additionally, existing CSLR datasets are collected in controlled settings, which restricts their effectiveness in building robust real-world CSLR systems. To address these limitations, we present Isharah, a large multi-scene dataset for CSLR. It is the first dataset of its type and size that has been collected in an unconstrained environment using signers' smartphone cameras. This setup resulted in high variations of recording settings, camera distances, angles, and resolutions. This variation helps with developing sign language understanding models capable of handling the variability and complexity of real-world scenarios. The dataset consists of 30,000 video clips performed by 18 deaf and professional signers. Additionally, the dataset is linguistically rich as it provides a gloss-level annotation for all dataset's videos, making it useful for developing CSLR and sign language translation (SLT) systems. This paper also introduces multiple sign language understanding benchmarks, including signer-independent and unseen-sentence CSLR, along with gloss-based and gloss-free SLT. The Isharah dataset is available on https://snalyami.github.io/Isharah_CSLR/. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 06:48:21 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Alyami",
"Sarah",
""
],
[
"Luqman",
"Hamzah",
""
],
[
"Al-Azani",
"Sadam",
""
],
[
"Alowaifeer",
"Maad",
""
],
[
"Alharbi",
"Yazeed",
""
],
[
"Alonaizan",
"Yaser",
""
]
] |
2506.03616 | Eunki Kim | Eunki Kim, Sangryul Kim, James Thorne | Learning to Insert [PAUSE] Tokens for Better Reasoning | 18 pages, 5 figures, ACL findings | null | null | null | cs.CL | http://creativecommons.org/licenses/by-nc-nd/4.0/ | To enhance reasoning capabilities, previous works have explored incorporating special-purpose tokens into the training process. These strategies strengthen the learning mechanism of transformer-based large language models (LLMs). Building on prior research, in which inserting dummy tokens consecutively just before reasoning steps can enhance effectiveness, we introduce a novel approach termed Dynamic Inserting Tokens Training (DIT). Our method identifies positions within sequences where model confidence is lowest according to token log-likelihood. Strategically inserting [PAUSE] tokens on these positions bolsters the model's predictive capabilities for subsequent tokens. Experimental results across diverse datasets and models, from the 2.7B model to the 8B model, demonstrate that DIT consistently outperforms traditional fine-tuning and previous token insertion methods. With this simple yet effective method, we achieve accuracy gains of up to 4.7%p on GSM8K, 3.23%p on AQUA-RAT, and pass@1 improvements of up to 3.4%p on MBPP datasets. Our work shows a model-based, dynamic approach rather than a heuristic one, thereby broadening the scope of research in reasoning. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 06:48:41 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Kim",
"Eunki",
""
],
[
"Kim",
"Sangryul",
""
],
[
"Thorne",
"James",
""
]
] |
2506.03617 | Changqing Ye | Shubin Fu, Qing Huo Liu, Qiwei Zhan, Eric T. Chung, Changqing Ye | MatExPre: A matrix exponential preconditioner for the high-frequency Helmholtz equation | null | null | null | null | math.NA cs.NA | http://creativecommons.org/licenses/by/4.0/ | In this article, we present a new preconditioner, MatExPre, for the high-frequency Helmholtz equation by leveraging the properties of matrix exponentials. Our approach begins by reformulating the Helmholtz equation into a Schr\"{o}dinger-like equation and constructing a time-domain solver based on a fixed-point iteration. We then establish a rigorous connection between the time-domain solver and matrix exponential integrators, which enables us to derive algebraic preconditioners that rely solely on sparse matrix-vector products. Spectral analysis and a detailed numerical implementation strategy, including performance improvements achieved through complex shifting, are discussed. Finally, numerical experiments on 2D and large-scale 3D homogeneous and inhomogeneous models, including benchmark seismic examples, substantiate the effectiveness and scalability of the proposed methods. | [
{
"version": "v1",
"created": "Wed, 04 Jun 2025 06:52:25 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Fu",
"Shubin",
""
],
[
"Liu",
"Qing Huo",
""
],
[
"Zhan",
"Qiwei",
""
],
[
"Chung",
"Eric T.",
""
],
[
"Ye",
"Changqing",
""
]
] |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.