---
title: "AutoDesign：面向长周期智能体设计的元优化框架"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2608.13560v1"
permalink: "https://feed.astralor.com/items/itm_b8c1eb736a42ef3959de6e87"
published: "2026-08-13T17:59:57.000Z"
score: 76
channel: "ai"
tags: ["AutoDesign","Claude Design","DesignHarness","PosterBench","元学习","智能体"]
---

# AutoDesign：面向长周期智能体设计的元优化框架

## AI 点评

> 这篇论文值得关注的地方在于它把'让智能体自我优化'从口号变成了可量化验证的框架，尤其用不到3美元的成本完成高质量海报生成，对资源有限的团队很有吸引力。对做Agent产品、自动化设计工具或内容管线的人，这套方法论很可能可以直接借鉴。

## 速览

AutoDesign提出了一种元优化框架，通过代码智能体基于反馈递归改进模型-工具系统，实现长周期智能体设计的自举优化。论文聚焦学术论文转海报任务，构建含100篇论文的PosterBench基准。AutoDesign在PosterBench上得78.32分，超越闭源商业系统Claude Design 7.45分；在七种代码智能体配置下，集成学习到的DesignHarness平均提升12.4%。全自主循环中，40分钟内执行253次工具调用和11次编辑轮次，成本不到3美元，达到会议海报平均质量水平。系统盲测显示AutoDesign获得最高人类偏好。

## 关键点

- 提出AutoDesign元优化框架，利用代码智能体根据交互反馈递归改进模型-工具系统，实现长期自我提升。
- 构建PosterBench基准，含100篇论文主轨和10篇论文子集，覆盖五个学科。
- AutoDesign在PosterBench主轨获得78.32分，超过Claude Design 7.45分；学到的DesignHarness在七种配置中平均提升12.4%。
- 全自主循环40分钟、成本低于3美元完成253次工具调用和11次编辑轮次，人类评估达会议海报平均质量。

## 全文

Transforming multimodal sources into condensed and structured media outputs can be fundamentally conceptualized as a long-horizon agentic process centered on a model-harness system. While an ideal harness system should align with human design priors and accumulate reusable experience through empirical exploration to drive recursive self-improvement, existing paradigms remain static and fall short of this capability. In this paper, we present AutoDesign, a framework that aligns with human design priors, where a meta-harness optimizer guides a code agent to recursively improve harness based on rollout feedback. To instantiate and evaluate this framework, we focus on the academic paper-to-poster generation task and introduce PosterBench, comprising a 100-paper Main Track spanning five disciplines and PosterBench-mini, a shared 10-paper subset for controlled evaluation. On the PosterBench Main Track, AutoDesign achieves the highest score of 78.32, surpassing the closed-source commercial system Claude Design by 7.45 points. Across seven controlled code-agent-model configurations, integrating the learned DesignHarness consistently improves performance, increasing the average PosterBench Score from 54.99 to 67.39 (+12.4%). In a fully autonomous long-horizon loop, it executes 253 tool calls and 11 editing turns within 40 minutes for under $3, reaching average conference-poster quality in human evaluation. A system-blind human study further demonstrates that AutoDesign achieves the highest human preference among evaluated systems.

---

引用前请核对原文;本文由星际流动整理:https://feed.astralor.com/items/itm_b8c1eb736a42ef3959de6e87