5B to 14B parameters) across math, coding, and science QA tasks.
Projecting wayward reasoning steps back into a focused latent region cuts token use and boosts accuracy
Hou et al. show that efficient reasoning steps cluster in latent space; a quadratic programming correction reduces overthinking without causing underthinking.
Top University
Zhinan Hou · XingChen Li · Keyou You
Tsinghua University
Research Digest··3 min read
The authors observe that in large reasoning models, efficient reasoning steps occupy a concentrated region in the latent space while verbose steps deviate.
Why this paper
From Tsinghua University
In one line
Projecting hidden states into a concentrated latent region reduces reasoning tokens by up to 52.8% while improving accuracy.
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 (6 benchmarks)
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§