Tree-Ising repairs Qwen reasoning loops with sparse BF16 edits
On two layers it scores 54.6% on MATH-500 versus 51.7% for a GPT-6 Astra edit and 7.9% for the base model.
Stories per day · 30 days
Relatedcomputer-algebra1formal-verification1ising1language-models1lean1proof-evaluation1quantum-computing1weight-editing1
4 stories, newest first · one card per story
Stories tagged #math-benchmarks
Ramanujan Challenge paper reports an AI pipeline for proof judging
The pipeline cut false-positive verdicts on natural-language proofs from 23% to 6% versus one-shot LLM judgment.
Which Language Should a Skeleton Speak? Language Choices in Multilingual Reasoning
LASEF finds English reasoning skeletons help smaller models slightly, but language choice is context-dependent.
What Does Privileged Information Add to On-Policy Self-Distillation?
Study of on-policy self-distillation on 5,319 AMPLE-Math problems finds reference-free distillation drives most gains, with modest extra benefit from privileged references.