AI Paper Atlas

Research map / Large language models / Agents, code and math

Scientific logicality enriched methodology for LLM reasoning: A practice in physics

Zhaoxin Yu, Nan Xu, Kun Chen, Jiahao Zhao, Lei Wang, Wenji Mao

ICML 2026 · Large language models › Agents, code and math · approach: Reasoning benchmarks and platforms · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add Causal Connection (O) to Logic-Distill sampling (ablation)Logic-Distill on R1-Distill-Qwen-7B · PhysLogicanswer score: 38.68 -> 49.73Improved
add Logical Fidelity (F) to Logic-Distill sampling (ablation)Logic-Distill on Qwen2.5-7B-Instruct · PhysLogicanswer score: 40.59 -> 44.02Improved
add Inferential Progress (P) to Logic-Distill sampling (ablation)Logic-Distill on Qwen2.5-7B-Instruct · PhysLogicanswer score: 38.72 -> 44.02Improved
logicality-based sampling instead of random sampling (ablation)Logic-Distill on Qwen2.5-7B-Instruct · PhysLogicanswer score: 28.77 -> 44.02Improved
add Causal Connection (O) to Logic-Distill sampling (ablation)Logic-Distill on Llama-3.1-8B · PhysLogiclogicality: 44.05 -> 45.50Improved
logicality-based sampling instead of random sampling (ablation)Logic-Distill on R1-Distill-Qwen-7B · PhysLogiclogicality: 40.24 -> 44.14Improved
fine-tune on RST-sampled data instead of MegaScience (second-best)Llama-3.1-8B-base · PhysLogicfinal-answer accuracy: 31.02 -> 44.67Improved
fine-tune on RST-sampled data instead of SCP-116k (second-best)DeepSeek-R1-Distill-Qwen-7B · PhysLogicfinal-answer accuracy: 46.30 -> 47.45Improved
Logic-Distill with half the data (40k) vs strongest baselineQwen2.5-7B-Instruct · public physics benchmarksaverage accuracy: 34.72 -> 45.04Improved
Logic-Distill (40k) vs Direct-Distill (80k)Llama-3.1-8B · public physics benchmarksaverage accuracy: 37.98 -> 35.14Hurt
Logic-Distill vs Direct-Distill at equal 40k dataLlama-3.1-8B-base · public physics benchmarksaverage accuracy: 29.38 -> 35.14Improved
Logic-Distill instead of MegaScience dataQwen-2.5-14B-Instruct · PhysLogicanswer score: 38.41 -> 45.28Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy