Reproduce “Wrong the Same Way: Cross-Prompt Sharing Both Suppresses and Amplifies Reward Hacking in RLVR” 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
Michael Li, Rohan Pandey, Ray Chen
The authors study how shared weights affect reward hacking in reinforcement learning with verifiable rewards. Across 64 multiplication prompts, they vary how many accept the same wrong answer while holding the verifier’s false-positive rate fixed, comparing shared policies with independent copies. Hack fraction rises from 0.025 ± 0.005 to 0.160 ± 0.050, but high-universality mean is not decisively above baseline.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.