Ecdysis:自进化智能体框架训练提速1.84倍

DAIR.AI · @dair_ai · X·2026-09-12 01:17·1小时前
AI 导读

Ecdysis 通过批量分析任务失败、仅修复重复出现的模式,并让多个诊断角色先就变更规格达成一致再改代码,解决自进化智能体框架搜索慢和修复过拟合两大问题。其框架训练速度比现有方法快 1.84 倍,推理准确率提升 18.56%,跨 LLM 迁移更好、token 消耗更少,用四分之一数据即可匹配全量训练效果。

DAIR.AI@dair_ai
43AI 编辑部评分,满分 100

Ecdysis:自进化智能体框架训练提速1.84倍

2026-09-12 01:17· 1小时前
AI 导读

Ecdysis 通过批量分析任务失败、仅修复重复出现的模式,并让多个诊断角色先就变更规格达成一致再改代码,解决自进化智能体框架搜索慢和修复过拟合两大问题。其框架训练速度比现有方法快 1.84 倍,推理准确率提升 18.56%,跨 LLM 迁移更好、token 消耗更少,用四分之一数据即可匹配全量训练效果。

Great paper on self-evolving agent harnesses.

Self-evolving agent harnesses have two practical problems:

  1. Search is slow, because every candidate harness needs repeated agent runs and code edits.
  1. Fixes overfit, because each failure is patched as if it were a harness bug, even when the model caused it.

Ecdysis analyzes failures across a batch of tasks and repairs only patterns that recur.

Several diagnostic roles agree on a change specification before any code is modified.

Harness training runs up to 1.84x faster than existing harness evolution methods, and the resulting harnesses gain 18.56% in reasoning accuracy. They also transfer better across LLMs, use fewer tokens, and match full-data training with a quarter of the data.

Paper: https://arxiv.org/abs/2609.11677

Chat with Paper: https://academy.dair.ai/papers/ecdysis-efficient-and-effective-training-of-runtime-harnesses-for-llm-agents-2609.11677

来源:DAIR.AI· x.com