arXiv AI By Peng Kuang, Haibo Jin, Dehao Wu, Feiyang Deng, Xiaopeng Yuan, Jerry Wang, Haohan Wang

Composing Task-specific Agent Harnesses at Test Time with Reusable Primitives

Read the original on arXiv AI →

The paper introduces Harness Primitives—reusable agent harness mechanisms mined from failed task trajectories—and a framework called STITCH that selects and composes these primitives into task‑specific harnesses at test time. This approach avoids generating or debugging harness code for each task, achieving up to 12‑point gains in task success over fixed harness baselines and outperforming human‑designed harnesses like Codex CLI. STITCH also demonstrates minimal test‑time overhead (2.7%) and scales efficiently with the size of the primitive library.

Machine-generated by The Flow from the publisher's headline and feed description — not written or checked by a human. The full article lives at arXiv AI.

arXiv Computation and Language
Sep 15

ModularRSI: Modular and Generalizable Recursive Harness Self-Improvement

arXiv:2609.14857v1 Announce Type: new Abstract: Recent work extends recursive self-improvement (RSI) to agent harnesses for long-horizon coding and terminal tasks, enabling agents to improve executio...

By Siwei Wu, Jincheng Ren, Yizhi Li, Haau-Sing Li, Chengran Yang, Yuxuan Zhang, Weicheng Gu, Jian Yang, Riza Batista-Navarro, Chuanyi Zhang, Xianglong Liu, Ming Zhou, Bryan Dai, Chenghua Lin
arXiv AI
Nov 6

openJiuwen: Beyond Static Harnesses for Long-Horizon Coding Agents

openJiuwen is an open-source harness that addresses two key challenges for long-horizon coding agents: Structural Composability and Runtime Adaptivity. It offers a shared execution substrate and Rail-based capability composition for single agents, delegated sub-agents, and Swarm Flow, enabling developers to build sophisticated agent systems without repeatedly rebuilding orchestration. The harness also adapts runtime decisions based on evolving evidence such as diagnostics and execution outcomes, and it achieves strong performance on SWE-bench Verified and Terminal-Bench 2.1, surpassing the best official leaderboard points by 3.4 and 3.39 percentage points.

By Tao Yu, Xinyu Zhang, Qianqian Chen, Xiaoneng Xiang, Chia Kwangyang, Xingchen Huang, Ran Chen, Yangkai Ding, Zheng Wang, Yeo Boon Hong, Bingzheng Gan, Enrui Hu, Shuo Cheng, Deyang Li, Ruifeng Shi, Hongbo Wang, Qi Ye, Xuefeng Jin, Zhangchun Zhao
arXiv AI
Sep 18

An Empirical Study of Harness Design for Coding Agents

The study investigates how individual components of a coding harness—planning, action space, and context management—affect autonomous coding agents’ performance. By fixing the execution loop and varying these components across 176 settings on SWE‑Bench Verified and Terminal‑Bench 2.1, the authors find that context management is most valuable when context windows are tight, staging rule‑based elision before LLM summarization yields the best efficiency, planning serves as an accuracy scaffold for weaker models and a cost saver for stronger ones, and predefined tools help models with limited bash skills while bash‑capable models benefit from a bash‑only interface. Trajectory‑level analysis shows that context management lengthens execution paths, planning alters where trajectories terminate, and the action space determines code granularity, offering a modular framework for future harness design.

By Run-Ze Fan, Zihao Zhang, Simin Ma, Yebowen Hu, Shouju Wang, Kaiqiang Song, Fei Liu, Hamed Zamani, Xiaoyang Wang