RL Generative Scaling
Investigates how reinforcement learning post-trained generative models behave under test-time scaling, including quality degradation with more sampled candidates.
3 papers
Where this stands
The written synthesis of this thread is for subscribers. Subscribe.