AI Paper Atlas

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

The Open Proof Corpus: A Large-Scale Study of LLM-Generated Mathematical Proofs

Jasper Dekoninck, Ivo Petrov, Kristian Minchev, Miroslav Marinov, Maria Drencheva, Lyuba Konova, Milen Milenov Shumanov, Kaloyan Tsvetkov, Nikolay Drenchev, Lazar D. Todorov, Kalina Nikolova, Nikolay Georgiev and 4 more

ICLR 2026 · Large language models › Agents, code and math · approach: Mathematical and structural reasoning benchmarks · 28 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
GRPO fine-tune R1-Qwen3-8B on OPC human labelsR1-Qwen3-8B · OPC proof judgingaccuracy: 70.7 -> 83.8 (pass@1); 71.3 -> 88.1 (maj@5)Improved
fine-tune on OPC (high-school only), test on undergraduate problemsR1-Qwen3-8B · Putnam undergraduate judging (OOD)accuracy: 72.1 -> 75.0Improved
judge own proofs instead of others' proofsLLM judges · OPC judging breakdownaccuracy: e.g. Gemini 79.4 own vs 86.9-85.9 by othersHurt
provide ground-truth solution to the judge (ablation)LLM judges · OPC proof judgingaccuracy: GPT-5 89.3 -> 89.0; o3 83.1 -> 85.9; Gemini-2.5-Pro 85.4 -> 82.7No clear effect
require valid proof instead of correct final answero3 · MathArena subsetaccuracy: 59.5% of answers contain correct proofHurt
use best-of-n selection with o4-mini judgeo4-mini · OPC best-of-n subsetproof accuracy: 26% -> 36% (standard), 47% (ranking)Improved
use pairwise ranking selection instead of discrete or continuous scoringbest-of-n · OPC best-of-n small subsetproof accuracy: about +10%Improved
use Rank (Swiss) instead of Rank (Bracket)best-of-n · OPC best-of-n small subsetproof accuracy: +3% marginImproved
increase n beyond 5 for discrete and continuous selectionbest-of-n · OPC best-of-n small subsetproof accuracyNo clear effect
natural-language instead of formal proof generationGemini-2.5-Pro vs Goedel-Prover-v2 · PutnamBench subsetproof accuracy: <19% -> almost 83%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