Ding et al.
Entropy signals guide efficient reasoning repair in LLMs
Ding et al. introduce REWIND, a training-free method that uses token entropy dynamics to select optimal points for regenerating suffixes, improving reasoning accuracy across multiple models and benchmarks.
Chinese Tech
Yi Ding · Mengyu Wang · Min Zhou · Menglin Yang
The Hong Kong University of Science and Technology (Guangzhou) · University of Edinburgh · Huawei
Research Digest··3 min read
The authors find that incorrect reasoning traces in large language models have mean token entropy roughly 112% higher than correct ones, a signal that can be used to localize errors.
Why this paper
From Huawei and 2 others
In one line
REWIND uses token entropy to guide targeted suffix regeneration, improving reasoning accuracy across models and benchmarks.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ·No stated limitations found
- ✓Reports numbers on named benchmarks (2 benchmarks)
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§