---
format: "aidr-story-markdown/v1"
id: "f06e4da304d537b8c12541ec4b2e31276b997548298df7ed72fbd41522063505"
canonical_url: "https://aidr.today/f06e4da3?lang=en"
title: "Doc2Harness: Benchmarking LLMs on Long-Horizon Harness Engineering of Coding Agents"
lang: "en"
requested_lang: "en"
available_langs: ["en","vi"]
translation_fallback: null
fallback_fields: []
published_at: "2026-10-15T00:00:00.000Z"
category: "Tools"
topics: ["agent","devtools"]
source_urls: ["https://huggingface.co/papers/2610.20344"]
summary: "Doc2Harness is a two-stage benchmark designed to evaluate long-horizon harness engineering for large language model-driven coding agents, where a build model constructs a candidate harness from a contract skeleton and development document, and the frozen harness is then tested end to end using a scripted mock model."
---

# Doc2Harness: Benchmarking LLMs on Long\-Horizon Harness Engineering of Coding Agents

> [Open the canonical story](<https://aidr.today/f06e4da3?lang=en>)

**Published:** 2026-10-15T00:00:00.000Z
**Category:** Tools
**Topics:** agent, devtools

## Summary

Doc2Harness is a two\-stage benchmark designed to evaluate long\-horizon harness engineering for large language model\-driven coding agents, where a build model constructs a candidate harness from a contract skeleton and development document, and the frozen harness is then tested end to end using a scripted mock model\.

## Sources

- [Story source](<https://huggingface.co/papers/2610.20344>)