IneqAgent uses LLMs and symbolic math to prove inequalities in Lean4
Researchers have introduced IneqAgent, a new agentic framework designed for formal inequality proving in the Lean4 theorem prover. The system combines large language model (LLM) reasoning with symbolic computation to handle mathematical tasks that require exact computation and intermediate lemmas. A key component is Lemma-level Monte Carlo Tree Search, which coordinates the construction of…
Key points
- IneqAgent combines LLM reasoning with symbolic computation for inequality proving in Lean4.
- Using DeepSeek-V4-Flash, the agent solved 175 of 186 benchmark problems (94.09%).
- The system achieved Lean-verified strengthenings of Dittert's conjecture in dimensions four and five.
The framework utilizes complementary backends, including sum-of-squares, Bernstein, and successive difference-substitution, to generate exact certificates for local Lean proofs. These are integrated with LLM-generated steps, and the agent uses Lean feedback to construct complete proofs that are checked against original theorem statements. This approach aims to bridge the gap between heuristic LLM reasoning and rigorous formal verification.
In testing on four benchmarks, IneqAgent paired with the DeepSeek-V4-Flash model solved 175 of 186 problems, achieving a 94.09% success rate. This represents the highest overall success rate among the evaluated methods. The study also notes that IneqAgent achieves higher success rates at low LLM inference cost thresholds compared to a Direct Agent using the same backbone. Additionally, the system produced Lean-verified strengthenings of Dittert's conjecture in dimensions four and five and certified the Holens--Djokovic (4,4) inequality.
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Research
All →- OpenAI offers U.S. university labs free ChatGPT Business seats with Rosalind access · 2 src
- Anthropic reports small AI models may self-replicate on IoT devices within months · 29 src
- Researchers propose LLM-referenced C-test for evaluating supra-intelligence · 8 src
- Researchers study permutation diffusion LLMs for recommendation reranking · 1 src
- Researchers compare Non-Generative verifier to Astra LLM for RAG claim checks · 1 src
Comments
via GitHub Discussions