Ignore the KL Penalty! Boosting Exploration on Critical Tokens to Enhance RL Fine-Tuning
Paper
•
2502.06533
•
Published
•
18
•
2
Roman Plaud
lecraquito
AI & ML interests
None yet
Recent Activity
authored
a paper
15 days ago
To Each Metric Its Decoding: Post-Hoc Optimal Decision Rules of
Probabilistic Hierarchical Classifiers
upvoted
a
paper
about 2 months ago
Absolute Zero: Reinforced Self-play Reasoning with Zero Data
upvoted
a
paper
about 2 months ago
Beyond the 80/20 Rule: High-Entropy Minority Tokens Drive Effective
Reinforcement Learning for LLM Reasoning
Organizations
commented
a
paper
6 months ago