arXiv cs.AI· Tianshi Xu, Huifeng Wen, Meng Li·· 2 天前AI 评分41
Life-Harness：不改模型权重，用运行时框架适配确定性 LLM 智能体
Adapting the Interface, Not the Model: Runtime Harness Adaptation for Deterministic LLM Agents
AI 导读
研究者提出 Life-Harness，一种生命周期感知的运行时框架，在不改变模型权重和评测环境的前提下提升冻结 LLM 智能体的表现。它从训练轨迹中把反复出现的交互失败转化为可复用干预，覆盖环境契约、流程技能、动作实现和轨迹调控。
正文 · 原文
Abstract:LLM agents are shaped not only by their language models, but also by the runtime harness that mediates observation, tool use, action execution, feedback interpretation, and trajectory control. While existing agent adaptation methods mainly update model parameters, many failures in deterministic, rule-governed domains stem from mismatches at the model--environment interface. We propose Life-Harness, a lifecycle-aware runtime harness that improves frozen LLM agents without changing model weights or evaluation environments. Life-Harness evolves from training trajectories by converting recurring interaction failures into reusable interventions across environment contracts, procedural skills, action realization, and trajectory regulation, and remains fixed for evaluation on unseen tasks. We evaluate Life-Harness across eight environments from AgentBench, $\tau$-bench, $\tau^2$-bench, and AppWorld using 18 open-weight backbones, and further test three frontier models on seven environments. With Qwen3-4B-Instruct, Life-Harness exceeds Meta-Harness and AHE by 28.7 and 7.5 percentage points in mean accuracy across the seven shared environments, respectively, while using 8.5\% and 68.4\% fewer optimization tokens. The evolved harnesses also benefit GPT 5.5, Opus 4.8, and Gemini 3.1 Pro, yielding an average accuracy gain of 3.5 percentage points. Code is available at this https URL.
| Comments: | Work in progress |
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2605.22166 [cs.AI] |
| (or arXiv:2605.22166v3 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2605.22166 arXiv-issued DOI via DataCite |
Submission history
From: Tianshi Xu [view email]
[v1]
Thu, 21 May 2026 08:36:49 UTC (1,723 KB)
[v2]
Wed, 27 May 2026 04:34:36 UTC (1,724 KB)
[v3]
Thu, 15 Oct 2026 03:20:30 UTC (1,704 KB)
来源：arXiv cs.AI · arxiv.org