Debate as Reward: A Multi-Agent Reward System for Scientific Ideation via RL Post-Training

A novel RL framework utilizing multi-agent debate to curb reward hacking and elevate scientific ideation quality.

Large Language Models (LLMs) have demonstrated potential in automating scientific ideation, yet current approaches relying on iterative prompting or complex multi-agent architectures often suffer from hallucination or computational inefficiency. A critical bottleneck in applying ...

ArXiv CS.AI2026-04-21