Ethan Perez

Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training

arXiv 2024

Latent Adversarial Training Improves Robustness to Persistent Harmful Behaviors in LLMs

arXiv 2024

Sycophancy to Subterfuge: Investigating Reward-Tampering in Large Language Models

arXiv 2024

Language Models Learn to Mislead Humans via RLHF

arXiv 2024

Bias-Augmented Consistency Training Reduces Biased Reasoning in Chain-of-Thought

arXiv 2024

Debating with More Persuasive LLMs Leads to More Truthful Answers

arXiv 2024

Alignment faking in large language models

arXiv 2024

Rapid Response: Mitigating LLM Jailbreaks with a Few Examples

arXiv 2024

Looking Inward: Language Models Can Learn About Themselves by Introspection

arXiv 2024

Pretraining Language Models with Human Preferences

arXiv 2023

Improving Code Generation by Training with Natural Language Feedback

arXiv 2023

Vision-Language Models are Zero-Shot Reward Models for Reinforcement Learning

arXiv 2023

Language Models Don't Always Say What They Think: Unfaithful Explanations in Chain-of-Thought Prompting

language-models-don-t-always-say-what-they

Question Decomposition Improves the Faithfulness of Model-Generated Reasoning

arXiv 2023

Training Language Models with Language Feedback at Scale

arXiv 2023

Red Teaming Language Models to Reduce Harms: Methods, Scaling Behaviors, and Lessons Learned

arXiv 2022

Discovering Language Model Behaviors with Model-Written Evaluations

arXiv 2022

Constitutional AI: Harmlessness from AI Feedback

arXiv 2022

Few-shot Adaptation Works with UnpredicTable Data

arXiv 2022