Paper summary
Paper summary: Does Reinforcement Learning Really Incentivize Reasoning Capacity in LLMs Beyond the Base Model?
Measuring pass@k at large k, this study finds RL with verifiable rewards makes models better at k = 1 but doesn't expand what the base model can solve at all; distillation, by contrast, can.
- RL-trained models win at pass@1, but base models solve more problems when allowed many attempts (large k).
- The reasoning paths after RLVR were already present in the base model; RL mostly makes them more likely.