Agent harness
15 papers we read, extracted findings from, and mapped onto how 5ac works.
Build or Adopt? Custom-Built vs. General-Purpose LLM Agent Harnesses for Automotive GUI Testing Keep the in-house harness or adopt a general one: Volvo ran both Arsham Gholamzadeh Khoee, Ruyu Fu, Jiaqi Wei +4 · 7 authors · Chalmers University of Technology and Volvo Group · 2026 · arXiv:2610.33636 Trial by Trace: Benchmarking AI Agent Harnesses for Reliability and Failure Attribution Benchmarks grade the model, but who grades the broken harness? Zhiyong Cao, Hengjie Zhu, Yongxin Wang +20 · 23 authors · Chinese Academy of Sciences, Tsinghua University, and Beijing Institute of Technology · 2026 · arXiv:2610.32285 How Do Coding-Agent Harnesses Self-Evolve? An Empirical Study of Retrospective Adaptation Skills get patched after every session — what proves the patch helps the next one? Christoph Treude, Sebastian Baltes · 2 authors · Singapore Management University and Heidelberg University · 2026 · arXiv:2610.34938 Rethinking Verification in Agent Harnesses Where in the agent lifecycle is checking worth the spend? Zhengxin Zhang, Erkang Zhu, Yin Lin +3 · 6 authors · Cornell University and Alibaba Group · 2026 · arXiv:2610.34577 HarnessMM: Recursive Self-Improvement for Harness Token Efficiency After every tool response, how much reasoning actually needs to run again? Yuhua Jiang, Chang Xu, Feifei Gao +1 · 4 authors · Tsinghua University and Microsoft Research Asia · 2026 · arXiv:2610.32813 Context Is a Pipeline, Not a Window: Information Attrition in LLM Agent Harnesses Measure agents by the context window and you miss the output being cut Yuvraj Sehgal, Mahsa Panahandeh, Naser Ezzati-Jivan · 3 authors · Brock University and University of Ottawa · 2026 · arXiv:2610.36114 Your Policy Is Secretly a Harness: The Decision Geometry of Agent Interfaces A trained policy never leaves its harness behind Jian Mu, Ruibo Shen, Yao Shu · 3 authors · The Hong Kong University of Science and Technology (Guangzhou) · 2026 · arXiv:2610.37255 A portable specification for the agent harness: which model ran inside which harness Benchmark results change with the harness, yet nobody records which harness Nicolas Gertler, Claudio Novelli, Luciano Floridi · 3 authors · Yale University (Digital Ethics Center) and University of Bologna · 2026 · arXiv:2610.40917 τ-bench: measuring whether tool-using agents follow domain rules, and stay consistent across runs An agent calls tools by the rules once: what happens on the second run? Shunyu Yao, Noah Shinn, Pedram Razavi +1 · 4 authors · Sierra · 2024 · arXiv:2406.12045 OSWorld: measuring agents on a real computer instead of a fabricated environment Demos click smoothly: what happens on a real computer? Tianbao Xie, Danyang Zhang, Jixuan Chen +14 · 17 authors · The University of Hong Kong · 2024 · arXiv:2404.07972 SWE-bench: how many real GitHub issues can a language model actually fix? Models are good at toy functions: what happens with a real issue in a real repo? Carlos E. Jimenez, John Yang, Alexander Wettig +4 · 7 authors · Princeton University · 2023 · arXiv:2310.06770 WebArena: a realistic web environment to measure whether an agent can actually do the work Agents look good on fake websites: what happens on real ones? Shuyan Zhou, Frank F. Xu, Hao Zhu +9 · 12 authors · Carnegie Mellon University · 2023 · arXiv:2307.13854 Voyager: capability accumulates in the skill library, not in the model Voyager keeps capability in the skill library, not in the model Guanzhi Wang, Yuqi Xie, Yunfan Jiang +5 · 8 authors · NVIDIA, Caltech, and UT Austin · 2023 · arXiv:2305.16291 Reflexion: an agent that learns from verbal feedback instead of weight updates The agent got it wrong: keep the lesson in words, not in retraining Noah Shinn, Federico Cassano, Edward Berman +3 · 6 authors · Northeastern University · 2023 · arXiv:2303.11366 ReAct: an agent that reasons and acts inside the same loop Why does an agent have to reason and act at the same time? Shunyu Yao, Jeffrey Zhao, Dian Yu +4 · 7 authors · Princeton University and Google Research (Brain Team) · 2022 · arXiv:2210.03629
Không có paper nào khớp bộ lọc này.