Paper
When Does External Guidance Help LLM Reasoning?
An academic paper proposing the GA-GRPO unified framework. It systematically analyzes the effects of external guidance on LLM reasoning training and provides closed-form theoretical solutions for convergence rates, bias bounds, and optimal guidance weighting.
Timeline (last 90 days)
Oct 7
The paper "When Does External Guidance Help LLM Reasoning?" proposes a unified framework called Guidance-Augmented GRPO (GA-GRPO), providing convergence rates, bias bounds, and optimal weighting rules for external guidance methods.
Unverified50%