摘要：新一代前沿模型不断拓展着能力的边界。在 Harness（包含运行循环、Shell 终端、文件系统以及编写并运行自身代码的能力环境）的辅助下，模型能够承担日益繁重的工作。

越来越多的研究将 Coding-Agent Harness 视为知识工作的基础设施，而非仅仅是一个编程工具，新一波基准测试也正从这一维度展开测评。大多数现有基准偏向生成类任务：根据需求简报构建财务模型、备忘录或电子表格，并由评分细则或另一个模型给出评判。很少有评测涉及知识工作中成本最高昂的环节——调查性（investigative）工作。在审计、税务、尽职调查、监管合规报告、保险理赔或任何针对数字的争议中，答案早已存在——潜藏在跨多个系统、不完整且不一致的证据之中——必须被完整还原并精确到分毫不差。

我们推出了 kw-bench，这是一个基于专业人员向我们描述的真实业务场景构建的基准测试。其首个任务族为跨越三代 SAP 系统的五年期增值税（VAT）重构——涵盖老旧的 ECC、部分迁移的 S/4HANA 以及由咨询顾问重新设计的系统，且全部以原始系统提取物形式提供：包含页面表头、本地数字格式、残缺的关联表，没有任何说明系统如何连接的文档。评分机制为确定性评估，不设 LLM 评委。

我们在同一模型 DeepSeek v4.1 Flash 上运行了四种 Harness（Claude Code、Codex、OpenCode 和 micro-cc），测试时长均为两小时。两个独立且与模型解耦的 Harness 交付了成果（OpenCode 0.542，micro-cc 0.528）；而两个实验室自研 Harness 在脱离自家原生模型后毫无建树（得分均为 0.00）。击垮它们的是惰性（inertia）：即在漫长会话中让大量微弱信号保持存活的能力。但在前沿模型上，该任务被轻而易举攻破。在 Claude Opus 5.5 上运行 micro-cc 和 Claude Code，分别在 22 分钟和 25 分钟内取得了 0.9986 和 1.000 的得分。原本耗时数周的复杂业务现在只需几分钟和几美元；而在前沿模型之下，Harness 决定了模型能走多远。

Abstract:Each new generation of frontier model pushes the edge of what is possible. Inside a harness -- a loop, a shell, a filesystem, the ability to write and run its own code -- a model can take on increasingly difficult work.
A growing body of work treats the coding-agent harness as infrastructure for knowledge work rather than a coding tool, and a new wave of benchmarks tests it there. Most of them are generative: build a financial model, a memo, a spreadsheet from a brief, judged by a rubric or by another model. Few test the expensive part of knowledge work, which is investigative. In audit, tax, due diligence, regulatory reporting, claims or any dispute over numbers, the answer already exists -- hidden in incomplete and inconsistent evidence spread across systems -- and has to be recovered and defended to the cent.
We introduce kw-bench, a benchmark built from real engagements specialists described to us. Its first task family is a five-year VAT reconstruction across three generations of SAP -- legacy ECC, a partial S/4HANA migration and a consultant's redesign -- delivered as raw system extracts: page headers, local number formats, partial link tables, no documentation of how anything connects. Grading is deterministic, with no LLM judge.
We ran four harnesses -- Claude Code, Codex, OpenCode and micro-cc -- on the same model, DeepSeek v4.1 Flash, for two hours. The two independent, model-agnostic harnesses delivered (OpenCode 0.542, micro-cc 0.528); the two lab harnesses, away from their home model, delivered nothing (0.00 each). What defeated them was inertia: keeping many weak signals alive across a long session. On a frontier model the task collapses. micro-cc and Claude Code on Claude Opus 5.5 scored 0.9986 and 1.000, in 22 and 25 minutes. A weeks-long engagement now costs minutes and a few dollars; below the frontier, the harness decides how far a model gets.