Cogentic: Multi-Agent Orchestration for Automated Proof Discovery
Cogentic, a multi-agent proof harness using Gemini, produced expert-verified novel results on five open math and CS problems.
Cogentic is a multi-agent harness for automated proof discovery on open research problems. An orchestrator spreads independent provers across proof directions, runs adversarial verification, and stores confirmed intermediate results in a persistent ledger for later rounds. Using Gemini as the base model, the system produced novel results on five open problems in online learning, auction theory, and mechanism design. Domain experts independently verified each result, with details published in companion papers.
- Orchestrator assigns independent provers across competing proof directions.
- Adversarial verifiers check outputs before a persistent verified ledger.
- Gemini-based runs yielded novel results on five open problems.
- Results span online learning, auction theory, and mechanism design.
- Domain experts independently verified each result.
From the original article
1 min read at arxiv.orgWe present Cogentic, a multi-agent harness for automated proof discovery on open research problems. While frontier language models can generate strong mathematical ideas in a single shot, single-shot generation is often insufficient for open problems that require exploring multiple competing conjectures, overcoming subtle technical obstructions, and retaining intermediate progress over a long horizon.
By Yang Cai · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.