Learning to Reason from Feedback at Test-Time

Yanyang Li, Michael Lyu, Liwei Wang

Solving complex tasks in a single attempt is challenging for large language models (LLMs). Iterative interaction with the environment and feedback is often required to achieve succ…