AutoDesign: Meta-Harness Optimisation for Long-Horizon Agentic Design

LinkYaxin Luo, Haobin Jiang, Jialv Zou, Xu Huang, Wenhao Yan, Haodong Li, Zhengrong Yue, Jing Li, Xiaofu Chen, Xiaohan Zhao, Jiacheng Liu, Jiacheng Cui, Zhiqiang Shen, Xiaotong LiResearch

Researchers introduce AutoDesign, a framework that leverages autonomous code agents to transform multimodal content into structured media outputs like conference posters. A meta-harness optimiser recursively improves the agent harness using rollout feedback, outperforming existing commercial systems.

Key points

  1. AutoDesign employs a meta-harness optimiser that guides code agents to iteratively refine their harness based on rollout feedback and human design priors.
  2. The approach was evaluated using PosterBench, a benchmark comprising 100 papers across five disciplines, alongside a 10-paper subset called PosterBench-mini.
  3. On the PosterBench Main Track, AutoDesign reached a score of 78.32, outperforming the commercial tool Claude Design by 7.45 points.
  4. Integrating the learned DesignHarness across seven code-agent-model configurations improved the average PosterBench score from 54.99 to 67.39 (+12.4%).
  5. During autonomous runs, the system executed 253 tool calls and 11 editing turns within 40 minutes at a cost of under $3 per poster.

Why it matters

Long-horizon agentic workflows often struggle in complex design tasks due to static harnesses. By using meta-optimisation to accumulate reusable design experience, AutoDesign offers a path towards robust and cost-effective autonomous visual generation.

Analysis

The study highlights that agent performance in open-ended creative tasks benefits heavily from self-improving operational harnesses, not just raw model capability. Achieving conference-grade poster quality in blind evaluations for under three dollars indicates that agentic design workflows are moving closer to practical deployment.

Mentioned links

Found in

Summary generated by AI (Gemini 3.8 Flash, 27 September 2026). It may contain errors – the original source is authoritative.

Related entries

Similar in content, found by the AI search.

  • Article:Yaxin Luo, Haobin Jiang, Jialv Zou, Xu Huang, Wenhao Yan, Haodong Li, Zhengrong Yue, Jing Li, Xiaofu Chen, Xiaohan Zhao, Jiacheng Liu, Jiacheng Cui, Zhiqiang Shen, Xiaotong Li

    AutoDesign: Meta-Harness Optimisation for Long-Horizon Agentic Design

    Researchers introduce AutoDesign, a framework that converts multimodal sources such as academic papers into structured visual formats like conference posters. A meta-harness optimiser guides a code agent to recursively improve its harness based on rollout feedback.

    AI & LLMs· Research

  • Article:Jiaqi Liu et al.

    AutoResearchClaw: Self-Reinforcing Autonomous Research with Human-AI Collaboration

    A research team led by Jiaqi Liu introduces AutoResearchClaw, an autonomous multi-agent research pipeline. Moving away from rigid linear workflows, the system implements iterative error recovery, verifiable reporting, and adaptable human intervention modes. On the ARC-Bench benchmark, it outperforms AI Scientist v2 by 54.7%.

    AI & LLMs· Research

  • Article:Shuai Shao, Kangning Zhang, Qingyao Li, Shijian Wang, Hao Wang, Wenxiang Jiao, Yuan Lu, Yi Guo, Weiwen Liu, Weinan Zhang

    Harness-R1: Learning to Edit Executable Runtime Harnesses from Agent Failure Trajectories

    Researchers introduce Harness-R1, a framework that trains a dedicated agent to adapt and patch the runtime harness of LLM agents based on failure logs. Using online reinforcement learning, a separate 9B model learns to generate executable patches for context handling, tool mediation, and error recovery to directly improve task completion.

    AI & LLMs· Research

  • Article:Mehdi Allahyari

    The Harness Beats the Model: Demystifying Agent Harness Engineering

    Following prompt and context engineering, harness engineering marks the third wave in developing autonomous AI systems. Mehdi Allahyari explains why the software infrastructure wrapping an LLM matters more than raw model scale for reliability and self-correction.

    AI & LLMs· Resource

  • Link:Powerformer, Inc.

    OpenDesign: Local Open-Source Alternative to Claude Design

    OpenDesign is an open-source, local-first AI design workspace released under the Apache-2.0 licence, positioned as an alternative to proprietary cloud platforms like Claude Design. It connects existing local coding agents into a structured design engine for generating prototypes, dashboards, slides, and HTML video.

    AI & LLMs· Tool

Back to the knowledge base

Let's talk about your project

Get in touch

Locations

  • Mattersburg
    Johann Nepomuk Bergerstraße 7/2/14
    7210 Mattersburg, Austria
  • Vienna
    Ungargasse 64-66/3/404
    1030 Wien, Austria

Parts of this content were created with the assistance of AI.