[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-816":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":10,"language":11,"languages":10,"totalLinesOfCode":10,"stars":12,"forks":13,"watchers":14,"openIssues":15,"contributorsCount":16,"subscribersCount":16,"size":16,"stars1d":17,"stars7d":18,"stars30d":19,"stars90d":16,"forks30d":16,"starsTrendScore":20,"compositeScore":21,"rankGlobal":10,"rankLanguage":10,"license":22,"archived":23,"fork":23,"defaultBranch":24,"hasWiki":25,"hasPages":23,"topics":26,"createdAt":10,"pushedAt":10,"updatedAt":29,"readmeContent":30,"aiSummary":31,"trendingCount":16,"starSnapshotCount":16,"syncStatus":32,"lastSyncTime":33,"discoverSource":34},816,"meta-harness","stanford-iris-lab\u002Fmeta-harness","stanford-iris-lab","Reference code for the Meta-Harness paper.","https:\u002F\u002Fyoonholee.com\u002Fmeta-harness\u002F",null,"Python",1055,103,9,4,0,26,54,205,78,19.05,"MIT License",false,"main",true,[27,28],"harness-engineering","llm-agents","2026-06-12 02:00:19","# Meta-Harness\n\n![Meta-Harness](assets\u002Frepo.png)\n\nMeta-Harness is a framework for automated search over task-specific model harnesses: the code around a fixed base model that decides what to store, retrieve, and show while the model works. This repo contains the framework and two reference experiments from the paper.\nThe paper is [Meta-Harness: End-to-End Optimization of Model Harnesses](https:\u002F\u002Farxiv.org\u002Fabs\u002F2603.28052).\n\n**If you end up building something cool with Meta-Harness, please let us know!** We would be happy to showcase it here in the main README and link to your repository, artifact, blog post, paper, or whatever else is most useful.\n\n## Contents\n\n- The reusable framework and onboarding flow for applying Meta-Harness to a new domain.\n- Two paper reference experiments under `reference_examples\u002F`:\n  - [`reference_examples\u002Ftext_classification\u002F`](reference_examples\u002Ftext_classification\u002FREADME.md): memory-system search for text classification.\n  - [`reference_examples\u002Fterminal_bench_2\u002F`](reference_examples\u002Fterminal_bench_2\u002FREADME.md): scaffold evolution for Terminal-Bench 2.0.\n- The optimized Terminal-Bench 2 harness from the paper lives in the separate artifact repo: [stanford-iris-lab\u002Fmeta-harness-tbench2-artifact](https:\u002F\u002Fgithub.com\u002Fstanford-iris-lab\u002Fmeta-harness-tbench2-artifact).\n\n## Quick Start\n\nText classification:\n\n```bash\ncd reference_examples\u002Ftext_classification\nuv sync\nuv run python meta_harness.py --iterations 1\n```\n\nTerminal-Bench 2 smoke task:\n\n```bash\ncd reference_examples\u002Fterminal_bench_2\nuv sync\nuv run bash scripts\u002Frun_eval.sh agents.baseline_kira:AgentHarness full 1 1 -i extract-elf\n```\n\nUse the subdir READMEs for setup details, expected runtime, and additional commands.\n\n## Applying Meta-Harness To A New Domain\n\nStart by pointing your coding assistant to [`ONBOARDING.md`](ONBOARDING.md) and having a conversation with it.\nThis should produce a `domain_spec.md` file with concrete details on how to proceed with implementing Meta-Harness for your domain.\n\nThe shipped examples currently assume Claude Code as the proposer agent. To use a different proposer agent, adapt the example `claude_wrapper.py` scripts in [`reference_examples\u002Ftext_classification\u002Fclaude_wrapper.py`](reference_examples\u002Ftext_classification\u002Fclaude_wrapper.py) or [`reference_examples\u002Fterminal_bench_2\u002Fclaude_wrapper.py`](reference_examples\u002Fterminal_bench_2\u002Fclaude_wrapper.py). The main requirement is a wrapper that cleanly logs proposer interactions.\n\n## Release Note\n\nThis is a cleaned up version of the code we used for the paper. It has not been tested beyond verifying that it runs. Please let us know if anything goes wrong.\n\n## Citation\n\nIf this repository is useful for your research, please cite the paper:\n\n```bibtex\n@misc{lee2026metaharnessendtoendoptimizationmodel,\n      title={Meta-Harness: End-to-End Optimization of Model Harnesses},\n      author={Yoonho Lee and Roshen Nair and Qizheng Zhang and Kangwook Lee and Omar Khattab and Chelsea Finn},\n      year={2026},\n      eprint={2603.28052},\n      archivePrefix={arXiv},\n      primaryClass={cs.AI},\n      url={https:\u002F\u002Farxiv.org\u002Fabs\u002F2603.28052},\n}\n```\n","Meta-Harness 是一个用于自动化搜索特定任务模型框架的工具，围绕固定的基础模型决定存储、检索和展示的内容。项目使用 Python 编写，提供了一个可复用的框架及两个参考实验，支持文本分类和终端基准测试等场景的应用。其核心技术特点是能够对模型周围的代码进行端到端优化，以提高特定任务的表现。适用于需要针对特定应用场景优化基础模型性能的研究人员或开发者。项目采用 MIT 许可证开放源码，并鼓励社区贡献新的应用案例。",2,"2026-06-11 02:39:33","CREATED_QUERY"]