AI RESEARCH
Beyond Correctness: Harmonizing Process and Outcome Rewards through RL Training
arXiv CS.LG
•
ArXi:2509.03403v2 Announce Type: replace Reinforcement Learning with Verifiable Rewards (RLVR) improves final-answer accuracy on reasoning tasks, but it does not reliably improve reasoning quality. Because outcome rewards only assess final answers, they also reward spurious successes: flawed reasoning can still receive maximal reward when it accidentally reaches the correct outcome. This outcome reward hacking creates biased gradients, making current RLVR insufficient for learning faithful reasoning.