PC-Dr.DPO improves reliability of majority-vote pseudo-labels in test-time RL
Prompt-Conditional Distributionally Robust DPO makes updates from sampling-sensitive prompts more conservative.
Stories per day · 30 days
2 RSS
3 stories, newest first · one card per story
Stories tagged #test-time-reinforcement-learning
Label-free steering: Compressing test-time reinforcement learning into bias-only subspaces
Bias-only label-free test-time RL lifts Qwen2.5-7B to 76.67% on MATH-500 using about 100K parameters.
Entropy-Regularized Rank-Masked Policy Optimization for Test-Time Reinforcement Learning in Code Generation
Researchers propose ERPO, enabling test-time reinforcement learning for code generation via probe-executed consensus rewards, rank masking, and entropy regularization.