{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2609.01437", "kind": "paper", "url": "https://arxiv.org/abs/2609.01437"}, {"kind": "website", "url": "https://self-developing-agents.github.io/"}], "authors": ["Yuhao Wu", "Jingyuan Zhang", "Jiajun Shi", "Xinping Lei", "Qingshui Gu", "Yuxuan Zhang", "Zexuan Wang", "Chen He", "Chen Huang", "Maojia Song", "Zhiyuan Zeng", "Shaowen Wang", "Jinkai Liu", "Yunfeng Shi", "Jiaheng Liu", "Shen Yan", "Wenhao Huang", "Ge Zhang", "Wenxuan Zhang"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2609.01437", "source_url": "https://arxiv.org/abs/2609.01437"}, "categories": ["Agents", "Code & Software", "General AI"], "description": {"en": "HarnessDev evaluates LLMs on creating and evolving runnable agent harnesses, shifting evaluation from task outputs to the infrastructure capabilities themselves."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2609.01437", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2609.01437", "title": "HarnessDev"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "key": "claire-radar:2609.01437", "modality": null, "name": "HarnessDev", "openness": {"code_license": null, "data_license": null, "evidence": [], "status": "unknown"}, "provenance": {"confidence": "High", "crawl_bundle": "claire_radar_2026-10-14", "crawled_at": "2026-10-14T09:44:51Z", "data_status": "primary-source-indexed", "display_eligible": "true", "origin_record_id": "bm_harnessdev_06493943", "origin_source": "arxiv", "recognition_confidence": "1.0", "record_sha256": "f03bfbab5583f7c4b0e87feadcc01f0c2a98260153803764edd97979a394c65f", "relation": "introduces", "review_model": "deepseek-v4-pro", "review_state": "ai-reviewed", "reviewed_at": "2026-09-03T00:59:34.317835Z", "source_url": "https://arxiv.org/abs/2609.01437"}, "publisher": null, "released": "2026-09-01", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2609.01437", "source_url": "https://arxiv.org/abs/2609.01437"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2609-01437", "source": "claire_radar", "source_benchmark_id": "2609.01437", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Unknown", "applicationDomains": ["General AI"], "area": "Code & Software", "attention": {"asOf": "2026-09-02", "attemptedAt": "2026-09-02T06:30:32.214812Z", "githubRepo": null, "githubScope": null, "githubStars": null, "hfDailySubmittedAt": null, "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": null, "hfPaperUrl": "https://huggingface.co/papers/2609.01437", "observedAt": "2026-08-31T04:56:28.279469Z", "signalStatus": {"githubStars": {"state": "not_applicable"}, "hfDatasetDownloads": {"state": "not_applicable"}, "hfDatasetLikes": {"state": "not_applicable"}, "hfPaperUpvotes": {"checkedAt": "2026-09-02T06:30:32.214812Z", "state": "unavailable"}}, "status": "unavailable"}, "attentionForecast": {"confidence": "Medium", "horizon": "7d", "reason": "The niche focus on agent harness evolution may attract moderate attention from the agentic AI and LLM evaluation communities.", "score": 60}, "capabilities": [], "confidence": "High", "construction": "Unknown", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-09-03T00:59:34.317835Z", "inputHash": "e6a0ce8ca3685617e26696ad9f019ef0d001b776103af9fcf1a2528c46066a85", "model": "deepseek-v4-pro"}, "curation": {"canonicalNameEvidence": "We introduce HarnessDev, a benchmark that shifts the unit of evaluation from task outputs to runnable infrastructure.", "canonicalNameSource": "abstract", "decisionReason": "The benchmark has a formal name, a defined evaluation object, and a public path via the project page, satisfying the criteria for publication.", "model": "deepseek-v4-pro", "reviewedAt": "2026-09-03T00:59:34.317835Z", "state": "ai-reviewed"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "HarnessDev evaluates LLMs on creating and evolving runnable agent harnesses, shifting evaluation from task outputs to the infrastructure capabilities themselves.", "displayEligible": true, "domainCuration": {"method": "rules-v1", "state": "auto"}, "evaluationMode": "score_submission", "evidence": {"reasonCodes": ["exact coined title identity tied to benchmark evidence", "benchmark release stated in one sentence", "evaluation protocol evidence", "public artifact URL"], "snippet": "We introduce HarnessDev, a benchmark that shifts the unit of evaluation from task outputs to runnable infrastructure."}, "familyId": "bmf_55aa01ef3a3a", "firstSeenAt": "2026-09-02", "heat": null, "id": "bm_harnessdev_06493943", "indexedAt": "2026-09-02T05:21:06.631990Z", "industrySectors": [], "links": {"code": null, "data": null, "paper": "https://arxiv.org/abs/2609.01437", "pdf": "https://arxiv.org/pdf/2609.01437", "project": "https://self-developing-agents.github.io/", "report": "https://arxiv.org/abs/2609.01437"}, "metrics": [], "motivation": "As agents move from research prototypes to deployed tools, their capability increasingly depends on model-external execution infrastructure, commonly termed the agent harness.", "name": "HarnessDev", "oneLine": "HarnessDev evaluates LLMs on creating and evolving runnable agent harnesses, shifting evaluation from task outputs to the infrastructure capabilities themselves.", "paperTitle": "HarnessDev: Can LLMs Create and Evolve Their Own Agent Harness?", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2609.01437", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-09-02T05:21:06.631990Z"}, "publications": [], "ranking": {"30d": {"confidence": "Low", "coverage": 0.0, "growth": {"components": {"githubStars": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "hfDatasetDownloads": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "hfPaperUpvotes": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}}, "confidence": "Low", "coverage": 0.0, "elapsedDays": null, "historyDate": null, "rank": null, "score": null, "windowDays": 30}, "level": {"components": {"githubStars": {"normalization": "log-max", "normalized": null, "value": null}, "hfDatasetDownloads": {"normalization": "log-max", "normalized": null, "value": null}, "hfPaperUpvotes": {"normalization": "log-max", "normalized": null, "value": null}}, "confidence": "Low", "coverage": 0.0, "rank": null, "score": null}, "method": "attention-ranking-v13", "rank": null, "score": null}, "90d": {"confidence": "Low", "coverage": 0.0, "growth": {"components": {"githubStars": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "hfDatasetDownloads": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "hfPaperUpvotes": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}}, "confidence": "Low", "coverage": 0.0, "elapsedDays": null, "historyDate": null, "rank": null, "score": null, "windowDays": 90}, "level": {"components": {"githubStars": {"normalization": "log-max", "normalized": null, "value": null}, "hfDatasetDownloads": {"normalization": "log-max", "normalized": null, "value": null}, "hfPaperUpvotes": {"normalization": "log-max", "normalized": null, "value": null}}, "confidence": "Low", "coverage": 0.0, "rank": null, "score": null}, "method": "attention-ranking-v13", "rank": null, "score": null}, "today": {"confidence": "Low", "coverage": 0.0, "growth": {"components": {"githubStars": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "hfDatasetDownloads": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "hfPaperUpvotes": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}}, "confidence": "Low", "coverage": 0.0, "elapsedDays": 2, "historyDate": "2026-08-31", "rank": null, "score": null, "windowDays": 1}, "level": {"components": {"githubStars": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "hfDatasetDownloads": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "hfPaperUpvotes": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "llmAttentionForecast": {"maxContribution": 0.25, "percentile": 0.68, "role": "bonus", "value": 68}}, "confidence": "Low", "coverage": 0.0, "rank": null, "score": 17}, "method": "attention-ranking-v13", "rank": null, "score": 17}}, "readiness": "Paper only", "recognitionConfidence": 1.0, "relation": "introduces", "releasedAt": "2026-09-01", "source": {"authors": ["Yuhao Wu", "Jingyuan Zhang", "Jiajun Shi", "Xinping Lei", "Qingshui Gu", "Yuxuan Zhang", "Zexuan Wang", "Chen He", "Chen Huang", "Maojia Song", "Zhiyuan Zeng", "Shaowen Wang", "Jinkai Liu", "Yunfeng Shi", "Jiaheng Liu", "Shen Yan", "Wenhao Huang", "Ge Zhang", "Wenxuan Zhang"], "categories": ["cs.SE", "cs.CL"], "id": "2609.01437", "title": "HarnessDev: Can LLMs Create and Evolve Their Own Agent Harness?", "type": "arxiv", "url": "https://arxiv.org/abs/2609.01437"}, "sourceUpdatedAt": "2026-09-01T15:45:33Z", "topics": ["Agents"], "venueAttempts": [], "whyItMatters": "It addresses the gap in evaluating whether models can autonomously build and improve the scaffolding needed for agent operation, providing insight into infrastructure-level self-development."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
