Reproduce “Stabilizing On-Policy Targeted Guidance in Agentic Reinforcement Learning” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Kevin Frans, Alexander M Rush, Sergey Levine, Pieter Abbeel, Ashvin Nair
The authors present targeted guidance, an on-policy self-distillation method for teaching coding agents behaviors missing from their usual attempts. After a failed turn, textual feedback conditions the same model’s token predictions, and distillation applies only to that turn while reinforcement learning rewards remain in the objective. In the semantic-search demonstration, it learned a new format without reducing coding pass rate; generalization beyond this experiment is unestablished.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.