Jiaxin Wen

Unsupervised Elicitation of Language Models

Measuring Harmfulness of Computer-Using Agents

Efficient Hyperparameter Tuning via Trajectory Invariance Principle

Predicting Empirical AI Research Outcomes with Language Models

Adaptive Deployment of Untrusted LLMs Reduces Distributed Threats

Language Models Learn to Mislead Humans via RLHF

CodePlan: Unlocking Reasoning Potential in Large Language Models by Scaling Code-form Planning

When Bigger is Better: Revisiting Large-Batch Optimization in Language Model Pretraining

SmartBackdoor: Malicious Language Model Agents that Avoid Being Caught