定理证明论文趋势 2023–2026

主要结论

• 定理证明论文从 2023 年的 3 篇增长到 2026 年的 38 篇，三年间扩大到约 13 倍。
• ICLR 是发定理证明论文最多的会议，2023–2026 年共 32 篇，占四个会议总量的 40%。
• 交叉方向里涨得最快的是数学推理：同时属于两个主题的论文从 2023 年的 1 篇增至 2026 年的 16 篇。
• 在「程序综合」的 5 个主题里，它的 2026 年论文数排第 3；同组增长最快的形式验证 ML是 3.7 倍，定理证明是 13 倍。

代表论文

按年份和会议均匀挑选，方便快速看到每一届的样貌

• NeurIPS 2026A Theoretical Framework for Self-Play Theorem Proving Algorithms
• ICML 2026Automated Formal Proofs of Combinatorial Identities via Wilf–Zeilberger Guidance and LLMs
• ICLR 2026LoC-Decomp: LLM Autoformalization via Logical Concept Decomposition and Iterative Feedback Correction
• NeurIPS 20253D-Prover: Diversity Driven Theorem Proving With Determinantal Point Processes
• ICML 2025Position: Trustworthy AI Agents Require the Integration of Large Language Models and Formal Methods
• ICLR 2025Alchemy: Amplifying Theorem-Proving Capability Through Symbolic Mutation
• CVPR 2025Automated Proof of Polynomial Inequalities via Reinforcement Learning
• NeurIPS 2024FVEL: Interactive Formal Verification Environment with Large Language Models via Theorem Proving
• ICML 2024Position: The No Free Lunch Theorem, Kolmogorov Complexity, and the Role of Inductive Biases in Machine Learning
• ICLR 2024A Characterization Theorem for Equivariant Networks with Point-wise Activations
• ICML 2023Extending Conformal Prediction to Hidden Markov Models with Exact Validity via de Finetti's Theorem for Markov Chains
• ICLR 2023Formal Mathematics Statement Curriculum Learning
• NeurIPS 2026Compositional Generalization Certificates via the Van Kampen Theorem
• ICML 2026Editable Proof Sketch for Automated Theorem Proving
• ICLR 2026Premise Selection for a Lean Hammer
• NeurIPS 2025Ineq-Comp: Benchmarking Human-Intuitive Compositional Reasoning in Automated Theorem Proving of Inequalities
• ICML 2025ProofAug: Efficient Neural Theorem Proving via Fine-grained Proof Structure Analysis
• ICLR 2025ImProver: Agent-Based Automated Proof Optimization
• NeurIPS 2024Proving Olympiad Algebraic Inequalities without Human Demonstrations
• ICML 2024Subgoal-based Demonstration Learning for Formal Theorem Proving

全部定理证明论文

共 80 篇，第 1 / 2 页

1. ICLR 2026A Representer Theorem for Hawkes Processes via Penalized Least Squares Minimization
2. ICLR 2026EvolProver: Advancing Automated theorem proving by Evolving Formalized Problems via Symmetry and Difficulty
3. ICLR 2026FormalML: A Benchmark for Evaluating Formal Subgoal Completion in Machine Learning Theory
4. ICLR 2026GAR: Generative Adversarial Reinforcement Learning for Formal Theorem Proving
5. ICLR 2026Goedel-Prover-V2: Scaling Formal Theorem Proving with Scaffolded Data Synthesis and Self-Correction
6. ICLR 2026Hilbert: Recursively Building Formal Proofs with Informal Reasoning
7. ICLR 2026Let's Explore Step by Step: Generating Provable Formal Statements with Deductive Exploration
8. ICLR 2026LoC-Decomp: LLM Autoformalization via Logical Concept Decomposition and Iterative Feedback Correction
9. ICLR 2026Mathesis: Towards Formal Theorem Proving from Natural Languages
10. ICLR 2026Neural Theorem Proving for Verification Conditions: A Real-World Benchmark
11. ICLR 2026Premise Selection for a Lean Hammer
12. ICLR 2026Process-Verified Reinforcement Learning for Theorem Proving via Lean
13. ICLR 2026ProofBridge: Auto-Formalization of Natural Language Proofs in Lean via Joint Embeddings
14. ICLR 2026ReForm: Reflective Autoformalization with Prospective Bounded Sequence Optimization
15. ICLR 2026VERINA: Benchmarking Verifiable Code Generation
16. ICML 2026A Minimal Agent for Automated Theorem Proving
17. ICML 2026APE-Bench: Evaluating Automated Proof Engineering for Formal Math Libraries
18. ICML 2026Automated Formal Proofs of Combinatorial Identities via Wilf–Zeilberger Guidance and LLMs
19. ICML 2026Beyond Theorem Proving: Formulation, Framework and Benchmark for Formal Problem-Solving
20. ICML 2026BrokenMath: A Benchmark for Sycophancy in Theorem Proving with LLMs
21. ICML 2026Can LLMs Reason Like Automated Theorem Provers for Rust Verification? VCoT-Bench: Evaluating via Verification Chain of Thought
22. ICML 2026Compile to Compress: Boosting Formal Theorem Provers by Compiler Outputs
23. ICML 2026Editable Proof Sketch for Automated Theorem Proving
24. ICML 2026Enhancing Neural Theorem Proving via High-Quality Proof Selection and Verifier Feedback
25. ICML 2026Euclean: Automated Geometry Problem Formalization with Unified Verification in Lean
26. ICML 2026Faults in Our Formal Benchmarking: Dataset Defects and Evaluation Failures in Lean Theorem Proving
27. ICML 2026From LLM-Generated Conjectures to Lean Formalizations: Automated Polynomial Inequality Proving via Sum-of-Squares Certificates
28. ICML 2026Hallucination is a Consequence of Space-Optimality: A Rate-Distortion Theorem for Membership Testing
29. ICML 2026miniF2F-Dafny: LLM-Guided Mathematical Theorem Proving via Auto-Active Verification
30. ICML 2026OptProver: Bridging Olympiad and Optimization through Continual Training in Formal Theorem Proving
31. ICML 2026SorryDB: Can AI Provers Complete Real-World Lean Theorems?
32. ICML 2026Towards Execution-Grounded Automated AI Research
33. NeurIPS 2026A Theoretical Framework for Self-Play Theorem Proving Algorithms
34. NeurIPS 2026Compositional Generalization Certificates via the Van Kampen Theorem
35. NeurIPS 2026DAPS: Dependency-Aware Premise Selection for LLM Theorem Proving
36. NeurIPS 2026Distilling LLM Feedback for Lean Theorem Proving
37. NeurIPS 2026How does RL Post-training Induce Skill Composition? A Case Study on Countdown
38. NeurIPS 2026LeanSearch v2: Global Premise Retrieval for Lean 4 Theorem Proving
39. CVPR 2025Automated Proof of Polynomial Inequalities via Reinforcement Learning
40. ICLR 2025Alchemy: Amplifying Theorem-Proving Capability Through Symbolic Mutation
41. ICLR 2025CARTS: Advancing Neural Theorem Proving with Diversified Tactic Calibration and Bias-Resistant Tree Search
42. ICLR 2025Deep Learning Alternatives Of The Kolmogorov Superposition Theorem
43. ICLR 2025ImProver: Agent-Based Automated Proof Optimization
44. ICLR 2025LeanAgent: Lifelong Learning for Formal Theorem Proving
45. ICLR 2025miniCTX: Neural Theorem Proving with (Long-)Contexts
46. ICLR 2025Proving Olympiad Inequalities by Synergizing LLMs and Symbolic Reasoning
47. ICML 2025MA-LoT: Model-Collaboration Lean-based Long Chain-of-Thought Reasoning enhances Formal Theorem Proving
48. ICML 2025Position: Trustworthy AI Agents Require the Integration of Large Language Models and Formal Methods
49. ICML 2025ProofAug: Efficient Neural Theorem Proving via Fine-grained Proof Structure Analysis
50. NeurIPS 20253D-Prover: Diversity Driven Theorem Proving With Determinantal Point Processes

常见问题

2023–2026 年四大会议一共有多少篇定理证明论文？

共 80 篇，按年份分别是 2023 年 3 篇、2024 年 19 篇、2025 年 20 篇、2026 年 38 篇。

NeurIPS 2026 有多少篇定理证明论文？

NeurIPS 2026 有 6 篇定理证明论文，上一届是 9 篇。

ICLR 2026 有多少篇定理证明论文？

ICLR 2026 有 15 篇定理证明论文，上一届是 7 篇。

ICML 2026 有多少篇定理证明论文？

ICML 2026 有 17 篇定理证明论文，上一届是 3 篇。

CVPR 2025 有多少篇定理证明论文？

CVPR 2025 有 1 篇定理证明论文。

哪个会议的定理证明论文最多？

ICLR，2023–2026 年共 32 篇。