Attention entropy splits RL training tokens into stable anchors and volatile explorers, and entropy-aware reweighting improves held-out reasoning performance.
For each generated response, we extract the content inside \boxed{} and compare it with the reference answer using string matching
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Not All Tokens Learn Alike: Attention Entropy Reveals Heterogeneous Signals in RL Reasoning
Attention entropy splits RL training tokens into stable anchors and volatile explorers, and entropy-aware reweighting improves held-out reasoning performance.