Scoring Higher, Answering Worse: Mitigating Reward Hacking in Rubric-Based RL via Protocol-Level Rubrics
Paper • 2609.38847 • Published • 68
Daniel Schmidt
danielsds
AI & ML interests
anomaly detection, responsible ai, model compression, autonomous driving
Recent Activity
liked a model 4 days ago
autotrust/JEV-35B upvoted a paper 4 days ago
AgentSWE: Can Coding Agents Build the Agent You Actually Want?Organizations
None yet
