Google Research Cogentic：多智能体自动化数学证明发现
Banger paper from Google Research on multi-agent proof discovery.
多智能体协同解决高难度数学证明的硬核实践，架构设计对 Agent 编排有直接参考价值。
Banger paper from Google Research on multi-agent proof discovery.
Google Research 关于多智能体证明发现的研究论文非常精彩。
(bookmark it)
（收藏它）
It's really interesting to see this emerging multi-agent pattern: not enforcing too much execution structure and pairing it with dedicated agents for advising and verification.
看到这种新兴的多智能体模式真的很有趣：不强制施加过多的执行结构，而是将其与专门用于建议和验证的智能体配对。
I think it is generally applicable as well. Great read.
我认为这种方法也普遍适用。值得一读。
Here is how it works:
其工作原理如下：
Cogentic runs on Gemini and works on open problems in theoretical computer science, starting from the problem statement with no expert hints.
Cogentic 基于 Gemini 运行，致力于解决理论计算机科学中的开放性问题，从问题陈述开始，不提供专家提示。
The system works in rounds, and the orchestrator decides how many provers to run in each round. Every prover gets one direction to work on, such as a specific bound or a counterexample search, plus a short briefing that a summarizer agent writes from earlier attempts and verifier feedback. Each summarizer writes its briefing independently, so provers in the same round read different summaries of the same history.
该系统以轮次方式工作，编排者决定每轮运行多少个证明者。每个证明者获得一个方向进行工作，例如特定界限或反例搜索，以及由摘要智能体根据早期尝试和验证反馈编写的简短简报。每个摘要智能体独立编写其简报，因此同一轮次的证明者阅读的是相同历史的不同摘要。
Each draft goes through two adversarial verifiers. One checks the draft on its own, and the other reads all of the round's drafts side by side to catch shared mistakes. A draft is accepted only if both pass it.
每个草稿需经过两个对抗性验证器。一个独立检查草稿，另一个并排阅读该轮次的所有草稿以捕捉共同错误。只有当两者都通过时，草稿才会被接受。
The agents share state through two disk documents. A record logs every attempt with the objection it failed on, and a ledger stores verified lemmas and ruled-out directions. An auditor extracts correct lemmas from rejected proofs, verifies them again independently, and adds them to the ledger.
智能体通过两个磁盘文档共享状态。记录日志记录了每次尝试及其失败的异议，分类账存储已验证的引理和被排除的方向。审计员从被拒绝的证明中提取正确的引理，独立再次验证后将其添加到分类账中。
A separate process advisor reads the verification logs across rounds and updates the instructions given to provers and verifiers. The orchestrator and the advisor can't give mathematical opinions, so the provers provide all the math.
一个独立的流程顾问跨轮次读取验证日志，并更新提供给证明者和验证者的指令。编排者和顾问不能提供数学意见，因此所有数学工作由证明者完成。
It produced new results on five open problems in online learning, auction theory, and mechanism design, each checked by domain experts. Most problems took around 100 Gemini calls, and the hardest took around 1,000.
它在在线学习、拍卖理论和机制设计领域的五个开放问题上产生了新结果，每个结果均由领域专家检查。大多数问题大约需要 100 次 Gemini 调用，最难的问题大约需要 1,000 次。
Paper: https://arxiv.org/abs/2609.40324
论文：https://arxiv.org/abs/2609.40324
Chat with Paper: https://academy.dair.ai/papers/cogentic-multi-agent-orchestration-for-automated-proof-discovery-2609.40324
与论文对话：https://academy.dair.ai/papers/cogentic-multi-agent-orchestration-for-automated-proof-discovery-2609.40324
更进一步：量化金融体系
看懂新闻只是起点——沿量化金融路径，把它变成能交付的工程能力