Sparse but Critical: A Token-Level Analysis of Distributional Shifts in RLVR Fine-Tuning of LLMs
Open paper
Benchmark
Model
Rank
Results
mathematical-reasoning-on-aime-2025
Qwen2.5-Math-7B with DAPO (High-KL boost)
–
Accuracy: 20.0
mathematical-reasoning-on-aime24
Qwen2.5-Math-7B with DAPO (High-KL boost)
–
Acc: 36.74
Rank counts only results with a code link.