智能体
arxiv.org ·
扩展工具框架而非上下文：从无策略脚手架到可复用专家智能体
论文提出 Growing Harness 失败引导训练范式，从仅暴露固定模型与工具接口、不含任务求解控制器的无策略脚手架出发，利用函数级执行轨迹把失败定位到有限代码面，由优化器联合修复一组失败，并用成功优先的留出集门控回滚损害既有能力的修复序列，使控制结构从任务反馈中逐步生长为共享工具框架。在 BrowseComp-Plus 与 WebArena-Verified 上使用 4B 至 120B 三种部署模型，六种基准-模型组合中五种平均成功率最高，第六种落后最优 0.7 个百分点；相比 Tool-Calling 智能体减少 76.0–91.8% 的 LLM 调用、降低 74.4–98.6% 部署推理成本。WebArena-Verified 上其成功率跨模型规模保持 44.7–45.3%，而 Tool-Calling 在 4B 模型降至 6.7%。消融表明轨迹局部编辑、联合修复与门控回滚均提升最终成功率。
AI推荐理由 arXiv 新论文（v1，2026-09-22 提交），提出 Growing Harness 失败引导训练范式：从无策略脚手架出发，用函数级执行轨迹定位失败、联合修复并加成功优先的留出回滚门，把重复控制逻辑沉淀为可复用代码。在 BrowseComp-Plus 与 WebArena-Verified 上，用 4B–120B 三种部署模型测试，六种设置中五种平均成功率最高，第六种落后 0.7 个百分点；相比 Tool-Calling 智能体减少 76.0–91.8% 的 LLM 调用、降低 74.4–98.6% 推理成本，WebArena-Verified 上跨模型规模成功率保持 44.7–45.3%，而 Tool-Calling 在 4B 模型下降到 6.7%。消融显示轨迹局部编辑、联合修复与门控回滚各自有效。局限是仅摘要级证据、基准与实现细节未展开，且为单篇预印本，尚无独立复现。
82
正文
Computer Science > Artificial Intelligence
arXiv:2609.26760 (cs)
[Submitted on 22 Sep 2026]
Title:Grow the Harness, Not the Context: From Strategy-Free Scaffolds to Reusable Specialist Agents
Authors:Laizhen Li, Jiarui Li, Juanjuan Zhao, Kejiang Ye, Ye Li, Cheng-zhong Xu, Xitong Gao
View a PDF of the paper titled Grow the Harness, Not the Context: From Strategy-Free Scaffolds to Reusable Specialist Agents, by Laizhen Li and 6 other authors
View PDF
HTML (experimental)
Abstract:Large language model (LLM) agents often handle streams of related tasks, yet standard harnesses repeatedly ask the model to reconstruct the same control decisions inside each task's context. We study whether task feedback can instead turn recurring control into reusable executable code, while reserving LLM calls for task-specific semantic reasoning. We introduce Growing Harness, a failure-guided training paradigm that learns the agent harness itself from a strategy-free scaffold that exposes fixed model and tool interfaces but encodes no task-solving controller. Function-level execution traces localize each failure to a bounded code surface, an optimizer repairs a window of failures jointly, and a success-first held-out gate rolls back repair sequences that harm prior capability. Accepted edits accumulate in one shared harness, allowing its control structure to emerge from task feedback. Across BrowseComp-Plus and WebArena-Verified with three deployment models from 4B to 120B parameters, Growing Harness achieves the highest mean success in five of six benchmark-model settings and trails the best mean by 0.7 pp. in the sixth. Relative to a Tool-Calling agent, it reduces LLM calls by 76.0-91.8% and deployed-agent inference cost by 74.4-98.6%. On WebArena-Verified, its success remains 44.7-45.3% across model scales, whereas Tool-Calling falls to 6.7% with the 4B model. Ablations show that trace-local edits, joint repair, and gate-based rollback each improve final success. These results show that persistent program growth can move recurring control out of model context and into low-cost code, yielding reusable specialist agents that remain effective with smaller deployment models.
Comments:
Subjects:
Artificial Intelligence (cs.AI); Software Engineering (cs.SE)
Cite as:
arXiv:2609.26760 [cs.AI]
(or
arXiv:2609.26760v1 [cs.AI] for this version)
https://doi.org/10.48550/arXiv.2609.26760
Focus to learn more
arXiv-issued DOI via DataCite (pending registration)
Submission history
From: Laizhen Li [view email]
[v1]
Tue, 22 Sep 2026 17:40:45 UTC (353 KB)
Full-text links:
Access Paper:
View a PDF of the paper titled Grow the Harness, Not the Context: From Strategy-Free Scaffolds to Reusable Specialist Agents, by Laizhen Li and 6 other authors
View PDF
HTML (experimental)
TeX Source
view license
Current browse context:
cs.AI
new
|
recent
| 2026-09
Change to browse by:
cs
cs.SE
References & Citations
NASA ADS
Google Scholar
Semantic Scholar
export BibTeX citation
Loading...
BibTeX formatted citation
×
loading...
Data provided by:
Bookmark
Bibliographic Tools
Bibliographic and Citation Tools
Bibliographic Explorer Toggle
Bibliographic Explorer (What is the Explorer?)
Connected Papers Toggle
Connected Papers (What is Connected Papers?)
Litmaps Toggle
Litmaps (What is Litmaps?)
scite.ai Toggle
scite Smart Citations (What are Smart Citations?)
Code, Data, Media
Code, Data and Media Associated with this Article
alphaXiv Toggle
alphaXiv (What is alphaXiv?)
Links to Code Toggle
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub Toggle
DagsHub (What is DagsHub?)
GotitPub Toggle
Gotit.pub (What is GotitPub?)
Huggingface Toggle
Hugging Face (What is Huggingface?)
ScienceCast Toggle
ScienceCast (What is ScienceCast?)
Demos
Demos
Replicate Toggle
Replicate (What is Replicate?)
Spaces Toggle
Hugging Face Spaces (What is Spaces?)
Spaces Toggle
TXYZ.AI (What is TXYZ.AI?)
Related Papers
Recommenders and Search Tools
Link to Influence Flower
Influence Flower (What are Influence Flowers?)
Core recommender toggle
CORE Recommender (What is CORE?)
Author
Venue
Institution
Topic
About arXivLabs
arXivLabs: experimental projects with community collaborators
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Which authors of this paper are endorsers? |
Disable MathJax (What is MathJax?)