EvoX 论文提出把 Agent Harness 视作运行时策略
EvoX: Rethinking the Agent Harness as a Runtime Policy
一篇关于 agent 工程的硬核论文：EvoX 用 Rust 写的运行时层去管住模型落地时的工具调用、中断恢复和结果校验，不靠记忆和插件就拿到 691 组测试里 56.2% 的成功率。做 agent 框架或工程化落地的开发者建议看看它对 GLM-5.3 和 Kimi K3 提升最大的原因。
arXiv 论文 2611.15006 提出 EvoX，一个基于 Rust 的多模型 Agent 运行时，核心思想是把 harness 看作补充模型语义策略的运行时操作策略（The Second Policy）。
在 5 个基准、6 个基础模型、6 个 agent harness 共 691 个任务-模型对的评测中，EvoX 取得 56.2% 的最高成功率，比其余 5 个 harness 的平均成绩高约 3.
0 个百分点。其机制由三部分组成：Action Realization（解决工具调用不匹配）、Execution Continuity（维持被中断的工作流）、Outcome Grounding（防止无依据的完成声明），全程不依赖记忆、进化或扩展插件。
论文同时发布 Contrast 评测框架，通过显式公平约束、请求来源验证和基于证据的裁定来避免压制各 harness 的原生行为。在 GLM-5.
3 上 EvoX 高出 10.7 个百分点、Kimi K3 上高出 7.
9 个百分点，说明 harness 层面的执行控制对不同模型的效果差异显著。
EvoX: Rethinking the Agent Harness as a Runtime Policy
Large language model (LLM) agents increasingly interact with tools and environments, yet plausible model decisions do not always translate into successful execution due to tool-call mismatches, interrupted workflows, and unsupported completion claims. We present EvoX, a Rust-based, multi-model agent runtime guided by The Second Policy, which views the harness as an Operational Policy complementing the foundation model's Semantic Policy. EvoX combines model-facing guidance with deterministic runtime control through three complementary mechanisms: Action Realization, Execution Continuity, and Outcome Grounding. To evaluate harnesses without suppressing their native behaviors, we introduce Contrast, a controlled and auditable framework integrating explicit fairness constraints, provider-bound request verification, and evidence-based adjudication. Across five benchmarks, six foundation models, and six agent harnesses, EvoX achieves the highest numerical success rate of 56.2% over 691 common task--model pairs, exceeding the average of the other five harnesses by approximately 3.0 percentage points. Its larger advantages on GLM-5.3 (+10.7 points) and Kimi K3 (+7.9 points) highlight the model-dependent effectiveness of harness design and the potential for cross-provider adaptability. Achieved without memory, evolution, or extensions, these results establish EvoX as a competitive multi-model agent runtime and underscore the importance of harness-level operational control in realizing foundation-model capabilities.
This report is categorized under AITOP's "Papers" desk. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows. Key entities and technical topics under continuous tracking include EvoX, agent, 论文, GLM-5.3.
Source attribution: curated from arXiv cs.AI (@Haoyang Zhang, Qumeng Sun, Xiao Zhang, Xiang Liu, Zhidong Fan, Junjie Wang). Cross-source corroboration: 7 independent sources have covered this development (including rohanpaul_ai, berryxia, pandaily), confirming industry-wide signal strength.