An entropy-based early stopping rule cuts token usage by roughly 50 percent on reasoning benchmarks without sacrificing accuracy.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Think or Not? Exploring Thinking Efficiency in Large Reasoning Models via an Information-Theoretic Lens
An entropy-based early stopping rule cuts token usage by roughly 50 percent on reasoning benchmarks without sacrificing accuracy.