Research map / Large language models / Agents, code and math
Hilbert: Recursively Building Formal Proofs with Informal Reasoning
Sumanth Varambally, Thomas Voice, Yanchao Sun, Zhifeng Chen, Rose Yu, Ke Ye
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add Hilbert recursive decomposition with Gemini 2.5 Pro | Goedel-Prover-V2-32B · MiniF2F-test | pass rate (%): 74.6 -> 99.2 | Improved |
| add Hilbert with Gemini 2.5 Pro | DeepSeek-Prover-V2-7B non-CoT · MiniF2F-test | pass rate (%): 61.3 -> 98.4 | Improved |
| add Hilbert on PutnamBench | Goedel-Prover-V2-32B · PutnamBench | solved problems (%): 13.4 -> 70.0 | Improved |
| use gpt-oss-120b instead of Gemini 2.5 Pro as reasoner (ablation) | Hilbert + Goedel-Prover-V2-32B · PutnamBench | solved problems (%): 70.0 -> 13.3 | Hurt |
| use Gemini 2.5 Pro instead of Gemini 2.5 Flash as reasoner (ablation) | Hilbert + Goedel-Prover-V2-32B · MiniF2F-test | pass rate (%): 94.7 -> 99.2 | Improved |
| add theorem retrieval (ablation) | Hilbert + Goedel-Prover-V2-32B · MiniF2F-test | pass rate (%): 97.9 -> 99.2 | Improved |
| add theorem retrieval (ablation) | Hilbert + DeepSeek-Prover-V2-7B · MiniF2F-test | pass rate (%): 97.1 -> 98.4 | Improved |
| add theorem retrieval (ablation) | Hilbert + Goedel-Prover-V2-32B · MiniF2F-test | reasoner tokens per sample: 4.0M -> 2.3M | Improved |
| add recursive subgoal decomposition up to depth 3 (ablation) | Hilbert + Goedel-Prover-V2-32B · MiniF2F-test | pass rate (%): 75 -> 98.7 | Improved |
| add shallow-solve by reasoner for subgoals (ablation) | Hilbert · MiniF2F-test | convergence with depth | Improved |
Closest papers by meaning
- HERMES: Towards Efficient and Verifiable Mathematical Reasoning in LLMs (ICML 2026 · similarity 0.92)
- APOLLO: Automated LLM and Lean Collaboration for Advanced Formal Reasoning (NeurIPS 2025 · similarity 0.92)
- MPS-Prover: Advancing Stepwise Theorem Proving by Multi-Perspective Search and Data Curation (NeurIPS 2025 · similarity 0.91)
- FaithSieve: Fine-Grained Evaluation of Math Proofs with Faithful Formal Evidence (NeurIPS 2026 · similarity 0.91)
- ImProver: Agent-Based Automated Proof Optimization (ICLR 2025 · similarity 0.90)
- Numina-Lean-Agent: An Open and General Agentic Reasoning System for Formal Mathematics (ICML 2026 · similarity 0.90)
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.
