Update README.md

README.md

@@ -28,12 +28,14 @@ Paris Noah's Ark Lab consists of 3 research teams that cover the following topic
 
    - [TAG: A Decentralized Framework for Multi-Agent Hierarchical Reinforcement Learning](https://huggingface.co/papers/2502.15425): distributed multi-agent hierarchical reinforcement learning framework.
    - [AdaPTS: Adapting Univariate Foundation Models to Probabilistic Multivariate Time Series Forecasting](https://arxiv.org/abs/2502.10235): simple yet powerful tricks to extend foundation models.
+   - [Clustering Head: A Visual Case Study of the Training Dynamics in Transformers](https://arxiv.org/abs/2410.24050): visual and theoretical understanding of training dynamics in transformers.
    - [Large Language Models as Markov Chains](https://huggingface.co/papers/2410.02724):  theoretical insights on their generalization and convergence properties.
    - [A Systematic Study Comparing Hyperparameter Optimization Engines on Tabular Data](https://balazskegl.medium.com/navigating-the-maze-of-hyperparameter-optimization-insights-from-a-systematic-study-6019675ea96c): insights to navigate the maze of hyperopt techniques. 
 
 ### 2025
 
    - *(ICLR'25)* - [Zero-shot Model-based Reinforcement Learning using Large Language Models](https://huggingface.co/papers/2410.11711): disentangled in-context learning for multivariate time series forecasting and model-based RL.
+   - *(ICASSP'25)* - [Easing Optimization Paths: A Circuit Perspective](https://arxiv.org/abs/2501.02362): mechanistic interpretability of training dynamics in transformers.
    - *(Neurocomputing)* - [Self-training: A survey](https://www.sciencedirect.com/science/article/pii/S0925231224016758): know more about pseudo-labeling strategies.
 
 ### 2024