Sanjiban Choudhury

All Roads Lead to Likelihood: The Value of Reinforcement Learning in Fine-Tuning

A Smooth Sea Never Made a Skilled SAILOR: Robust Imitation via Learning to Search

Efficient Imitation under Misspecification

Imitation Learning from a Single Temporally Misaligned Video

Multi-Turn Code Generation Through Single-Step Rewards

Non-Adversarial Inverse Reinforcement Learning via Successor Feature Matching

Robotouille: An Asynchronous Planning Benchmark for LLM Agents

Better than Your Teacher: LLM Agents that learn from Privileged AI Feedback

Query-Efficient Planning with Language Models

Aligning Large Language Models with Domain Adaptation