Sakana AI 提出 Conductor：用自然语言编排多智能体
Learning to Orchestrate Agents in Natural Language with the Conductor
Sakana AI 发布论文《Learning to Orchestrate Agents in Natural Language with the Conductor》，该工作已被 ICLR2026 接收。研究用强化学习训练一个 7B 的 Conductor 模型，让它以自然语言输出协作流程，指定调用哪个智能体、分配什么子任务以及它们能看到哪些历史消息，从而编排 GPT-5、Gemini、Claude 及开源模型组成的模型池。Conductor 会随任务难度自适应：简单事实问题只调用一个模型，困难编码问题则自动搭建 planner-executor-verifier 流水线。论文称 7B Conductor 在 LiveCodeBench 达到 83.9%、GPQA-Diamond 达到 87.5%，超过池中每个单独模型，并以更低成本优于 Mixture-of-Agents 等多智能体基线；它还支持递归测试时扩展，可把自己选为 worker 并在失败时生成纠正流程。该研究与几天前公布的 TRINITY 一起支撑其新的多智能体系统 Sakana Fugu。
推荐理由
AI 评分 59
论文展示了 7B Conductor 用自然语言编排多个前沿模型，并给出 LiveCodeBench 与 GPQA-Diamond 上的具体成绩。
主题
正文
当前显示原文，中文全文译文尚未就绪。
TL;DR
For the past few years, humans have been doing “prompt engineering” to coax the best performance out of different LLMs. In this work, we explored what happens if we train an AI to do that job instead.
In this work, by training a Conductor model with RL, we found that it naturally learns to write highly effective, custom instructions for a whole pool of other models. It essentially learns to ‘manage’ them in natural language.
What surprised us most was how it dynamically adapts. For simple factual questions, it just queries one model. But for hard coding problems, it autonomously spins up a whole pipeline of planners, coders, and verifiers.
Summary
Introducing our new work: “Learning to Orchestrate Agents in Natural Language with the Conductor” accepted at ICLR2026
What if we trained an AI not to solve problems directly, but to act as a manager that delegates tasks to a diverse team of other AIs?
To solve complex tasks, humans rarely work alone; we form teams, delegate, and communicate. Yet, multi-agent AI systems currently rely heavily on rigid, human-designed workflows or simple routers that just pick a single model. We wanted an AI that could dynamically build its own team.
We trained a 7B Conductor model using Reinforcement Learning to orchestrate a pool of frontier models (including GPT-5, Gemini, Claude, and open-source models available during the period leading up to ICLR2026).
Instead of executing code, the Conductor outputs a collaborative workflow in natural language. For any given question, the Conductor specifies:
Which agent to call; What specific subtask to give them (acting as an expert prompt engineer); What previous messages they can see in their context window
Through pure end-to-end reward maximization, amazing behaviors emerged. The Conductor learned to adapt to task difficulty: it 1-shots simple factual questions, but autonomously spins up complex planner-executor-verifier pipelines for hard coding problems.
The results are very promising: The 7B Conductor surpasses the performance of every individual worker model in its pool, setting new records on LiveCodeBench (83.9%) and GPQA-Diamond (87.5%) at the time of publication. It also significantly outperforms expensive multi-agent baselines like Mixture-of-Agents at a fraction of the cost.
One of our favorite features: Recursive Test-Time Scaling! By allowing the Conductor to select itself as a worker, it reads its own team’s prior output, realizes if it failed, and spins up a corrective workflow on the fly. This opens a new axis for scaling compute during inference.
This research proves that language models can become elite meta-prompt engineers, dynamically harnessing collective intelligence.
Alongside our TRINITY research which we announced a few days earlier, this foundational research powers our new multi-agent system: Sakana Fugu! 🐡