Almost Surely Safe Alignment of Large Language Models at Inference-Time
Paper
•
2502.01208
•
Published
•
11
Haitham Bou Ammar
hba123
AI & ML interests
LLMs, VLMs, Robotics, Reinforcement Learning, Bayesian Optimisation
Recent Activity
authored
a paper
7 days ago
Almost Surely Safe Alignment of Large Language Models at Inference-Time
reacted
to
their
post
with 😎
8 days ago
We developed a method that ensures almost-sure safety (i.e., safety with probability approaching 1). We proved this result. We then, present a practical implementation which we call InferenceGuard. InferenceGuard has impressive practical results: 91.04% on Alpaca-7B and 100% safety results on Beaver 7B-v3.
Now, it is easy to get high safety results like those if we want a dumb model, e.g., just don't answer or answer with EOS and so on. However, our goal is not to only have safe results, but also to make sure that the rewards are high - we want a good trade-off between safety and rewards! That's exactly, what we show. InferenceGuard achieves that!
Check it out: https://huggingface.co/papers/2502.01208
reacted
to
their
post
with 🔥
8 days ago
We developed a method that ensures almost-sure safety (i.e., safety with probability approaching 1). We proved this result. We then, present a practical implementation which we call InferenceGuard. InferenceGuard has impressive practical results: 91.04% on Alpaca-7B and 100% safety results on Beaver 7B-v3.
Now, it is easy to get high safety results like those if we want a dumb model, e.g., just don't answer or answer with EOS and so on. However, our goal is not to only have safe results, but also to make sure that the rewards are high - we want a good trade-off between safety and rewards! That's exactly, what we show. InferenceGuard achieves that!
Check it out: https://huggingface.co/papers/2502.01208
Organizations
None yet
hba123's activity
upvoted
a
paper
8 days ago
upvoted
an
article
about 1 month ago
Article
Accelerating Language Model Inference with Mixture of Attentions
By
and 1 other
•
•
24
upvoted
an
article
about 2 months ago
Article
Deriving DPO's Loss
By
•
•
26
upvoted
a
paper
3 months ago
upvoted
2
papers
6 months ago
upvoted
a
paper
7 months ago
upvoted
3
collections
7 months ago
upvoted
7
papers
7 months ago
Human-like Episodic Memory for Infinite Context LLMs
Paper
•
2407.09450
•
Published
•
60
Towards Robust Speech Representation Learning for Thousands of Languages
Paper
•
2407.00837
•
Published
•
10
Show Less, Instruct More: Enriching Prompts with Definitions and Guidelines for Zero-Shot NER
Paper
•
2407.01272
•
Published
•
8
DogeRM: Equipping Reward Models with Domain Knowledge through Model Merging
Paper
•
2407.01470
•
Published
•
5
ROS-LLM: A ROS framework for embodied AI with task feedback and structured reasoning
Paper
•
2406.19741
•
Published
•
60
Accurate Prediction of Ligand-Protein Interaction Affinities with Fine-Tuned Small Language Models
Paper
•
2407.00111
•
Published
•
5
Auto Cherry-Picker: Learning from High-quality Generative Data Driven by Language
Paper
•
2406.20085
•
Published
•
11
upvoted
a
paper
12 months ago