Research map / Large language models / Agents, code and math
SITA: A Framework for Structure-to-Instance Theorem Autoformalization
Chenyi Li, Wanli Ma, Zichen Wang, Zaiwen Wen
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 |
|---|---|---|---|
| apply SITA pipeline instead of direct generation with DeepSeek-V3 | DeepSeek-V3 · Optimization structure-to-instance formalization | file success rate: 0.0% -> 27.2% | Improved |
| apply SITA pipeline instead of direct generation with DeepSeek-R1 | DeepSeek-R1 · Optimization structure-to-instance formalization | file success rate: 0.0% -> 57.14% | Improved |
| apply SITA pipeline to DeepSeek-V3 for definitions | DeepSeek-V3 · Optimization structure-to-instance formalization | definition success rate: 27.9% -> 91.0% | Improved |
| apply SITA pipeline to DeepSeek-R1 for majority-vote score | DeepSeek-R1 · Optimization structure-to-instance formalization | majority-vote score: 46.0 -> 76.9 | Improved |
| use DeepSeek-R1 instead of DeepSeek-V3 as SITA backbone | SITA · Optimization structure-to-instance formalization | file success rate: 27.2% -> 57.14% | Improved |
| add same-category exemplar in prompt (ablation) | SITA · Optimization structure-to-instance formalization (ablation) | file success rate: 9.5% -> 57.14% | Improved |
| add same-category exemplar in prompt (ablation) | SITA · Optimization structure-to-instance formalization (ablation) | syntactic correctness rate: 74.1% -> 90.7% | Improved |
| add same-category exemplar in prompt (ablation) | SITA · Optimization structure-to-instance formalization (ablation) | proof success rate: 42.5% -> 51.23% | Improved |
| add error knowledge base correction-tip retrieval (ablation) | SITA · Optimization structure-to-instance formalization (ablation) | file success rate: 45.8% -> 57.14% | Improved |
| add error knowledge base correction-tip retrieval (ablation) | SITA · Optimization structure-to-instance formalization (ablation) | syntactic correctness rate: 83.4% -> 90.7% | Improved |
| add error knowledge base correction-tip retrieval (ablation) | SITA · Optimization structure-to-instance formalization (ablation) | proof success rate: 49.8% -> 51.23% | Improved |
| add secondary proof refinement stage (ablation) | SITA · Optimization structure-to-instance formalization (ablation) | proof success rate: 23.45% -> 51.23% | Improved |
Closest papers by meaning
- APOLLO: Automated LLM and Lean Collaboration for Advanced Formal Reasoning (NeurIPS 2025 · similarity 0.88)
- ImProver: Agent-Based Automated Proof Optimization (ICLR 2025 · similarity 0.87)
- miniF2F-Lean Revisited: Reviewing Limitations and Charting a Path Forward (NeurIPS 2025 · similarity 0.87)
- LeanAgent: Lifelong Learning for Formal Theorem Proving (ICLR 2025 · similarity 0.87)
- ATLAS: Autoformalizing Theorems through Lifting, Augmentation, and Synthesis of Data (NeurIPS 2025 · similarity 0.87)
- Autoformalize Mathematical Statements by Symbolic Equivalence and Semantic Consistency (NeurIPS 2024 · similarity 0.87)
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.
