AI 精选动态
智能评分 78
Sakana Fugu Technical Report
AI 推荐理由
差异点：首次提出通过动态路由多模型实现跨基准 SoTA 成果核心解读
Sakana AI 在技术报告中介绍了 Fugu 概念，即通过路由器和深度多智能体导师动态组合 GPT-5.5、Gemini-3.1-Pro、Claude Opus 4.8 等模型来处理不同查询。该系统使用 SFT、进化策略和 GRPO 训练，并在 SWE‑Bench Pro、Terminal Bench、LiveCodeBench、GPQA‑Diamond、CharXiv 等基准上实现了 SoTA 成果。
全文
Sakana AI (@SakanaAILabs) 转发了 alphaXiv (@askalphaxiv) 的帖子：
Sakana Fugu Technical Report
Instead of training one larger model, Sakana AI trains an orchestrator that reads each query and dynamically routes or composes GPT-5.5, Gemini-3.1-Pro, Claude Opus 4.8 and other agents into query-specific workflows.
With Fugu being the fast router, and Fugu-Ultra being the deep multi-agent conductor, trained with SFT, evolutionary strategies and GRPO to build adaptive scaffolds.
The idea is to have the model pick GPT for math, Gemini for science and recall, Opus for debugging, then synthesize them when no single agent is best.
This router is able to get SoTA results across SWE-Bench Pro, Terminal Bench, LiveCodeBench, GPQA-Diamond, CharXiv and more, demonstrating the potential of orchestration being a practical alternative beyond training.
