AI Paper Atlas

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

LeanForPhysics: Comprehensive Reasoning Framework for University-level Physics in Lean4

Yuxin Li, Minghao LIU, Ruida WANG, JI WenZhao, Zhitao He, Rui Pan, Junming Huang, Tong Zhang, Yi R. Fung

ICLR 2026 · Large language models › Agents, code and math · approach: Inductive and abstract reasoning evaluation

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 PhysLib to contextGemini-2.5-pro · LeanPhysBench Overallpass@16: 6.67% -> 40.00%Improved
add PhysLib to contextClaude-Sonnet-4 · LeanPhysBench Overallpass@16: 3.50% -> 34.50%Improved
add PhysLib to contextGPT-4o · LeanPhysBench Overallpass@16: 2.50% -> 14.17%Improved
add PhysLib to contextGoedel-Prover-V2-8B · LeanPhysBench Overallpass@16: 10.00% -> 14.67%Improved
add PhysLib to contextDeepSeek-Prover-V2-7B · LeanPhysBench Overallpass@16: 11.17% -> 14.83%Improved
add PhysLib to contextKimina-Prover-8B · LeanPhysBench Overallpass@16: 9.33% -> 13.50%Improved
add PhysLib to contextQwen3-8B · LeanPhysBench Overallpass@16: 2.35% -> 6.52%Improved
add PhysLib to contextDeepSeek-R1-8B · LeanPhysBench Overallpass@16: 3.33% -> 5.83%Improved
add PhysLib to context (average over models)8 LLMs · LeanPhysBench Overallpass@16: 11.90%Improved
add PhysLib to contextGemini-2.5-pro · LeanPhysBench Comp-Hardpass@16: 0.00% -> 0.00%No clear effect
use expert Lean prover instead of general LLM (Goedel-Prover-V2-8B vs Gemini-2.5-pro) with PhysLibexpert provers · LeanPhysBench Overallpass@16: 40.00% -> 14.67%Hurt
use expert Lean prover instead of Claude on Comp-Hardexpert provers · LeanPhysBench Comp-Hardpass@16: 0.00% -> 5.88%Improved

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