Scoring Higher, Answering Worse: Mitigating Reward Hacking in Rubric-Based RL via Protocol-Level Rubrics
Paper • 2609.38847 • Published • 61
Florian Weber
flweber0225
AI & ML interests
lost in latent space
Recent Activity
upvoted a paper 5 days ago
CheckerBench: Can Long-Horizon Agents Synthesize Static-Analysis Checkers?Organizations
None yet
upvoted a paper 1 day ago