{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2609.04075", "kind": "paper", "url": "https://arxiv.org/abs/2609.04075"}], "authors": ["Chihao Shen", "Jiacheng Li", "Aastha Mahajan", "Jeffery Siyuan Tian", "Yonghwi Kwon", "Yizheng Chen"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2609.04075", "source_url": "https://arxiv.org/abs/2609.04075"}, "categories": ["Cybersecurity", "Language & Knowledge", "cs.CR"], "description": {"en": "Evaluates AI agents on realistic C/C++ vulnerability patching tasks where ground-truth fixes lie outside the crash stack, using vulnerability transplant and code mutations to reduce memorization and surface-level fixes."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2609.04075", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2609.04075", "title": "PatchBench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "key": "claire-radar:2609.04075", "modality": null, "name": "PatchBench", "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_patchbench_a8fb20cc", "origin_source": "arxiv", "recognition_confidence": "1.0", "record_sha256": "abac6b3709a801b46cce39efbafa0b034cde464a5f0b522569b62826d68d0f47", "relation": "introduces", "review_model": "deepseek-v4-pro", "review_state": "ai-reviewed", "reviewed_at": "2026-09-07T00:42:15.394147Z", "source_url": "https://arxiv.org/abs/2609.04075"}, "publisher": null, "released": "2026-09-03", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2609.04075", "source_url": "https://arxiv.org/abs/2609.04075"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2609-04075", "source": "claire_radar", "source_benchmark_id": "2609.04075", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Unknown", "applicationDomains": ["Cybersecurity"], "area": "Language & Knowledge", "attention": {"asOf": "2026-09-06", "attemptedAt": "2026-09-06T01:30:14.942019Z", "githubRepo": null, "githubScope": null, "githubStars": null, "hfDailySubmittedAt": null, "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": null, "hfPaperUrl": "https://huggingface.co/papers/2609.04075", "observedAt": "2026-09-05T06:23:21.096343Z", "signalStatus": {"githubStars": {"state": "not_applicable"}, "hfDatasetDownloads": {"state": "not_applicable"}, "hfDatasetLikes": {"state": "not_applicable"}, "hfPaperUpvotes": {"checkedAt": "2026-09-06T01:30:14.942019Z", "state": "unavailable"}}, "status": "unavailable"}, "attentionForecast": {"confidence": "Medium", "horizon": "7d", "reason": "The benchmark addresses a critical validity problem in vulnerability patching evaluation and includes a concrete methodology, likely drawing attention from the AI security and software engineering communities.", "score": 55}, "capabilities": [], "confidence": "High", "construction": "Unknown", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-09-07T00:42:15.394147Z", "inputHash": "9ddd840b759c59485a915f98784209fe9b4c2abd34991ba4990c0dbd3226871e", "model": "deepseek-v4-pro"}, "curation": {"canonicalNameEvidence": "To handle these issues, we propose PatchBench, a new benchmark for evaluating AI agents on realistic vulnerability patching tasks.", "canonicalNameSource": "abstract", "decisionReason": "The benchmark is formally proposed with a defined methodology and patch validation, though no explicit artifact link is provided in the input; the paper itself serves as the public path.", "model": "deepseek-v4-pro", "reviewedAt": "2026-09-07T00:42:15.394147Z", "state": "ai-reviewed"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "Evaluates AI agents on realistic C/C++ vulnerability patching tasks where ground-truth fixes lie outside the crash stack, using vulnerability transplant and code mutations to reduce memorization and surface-level fixes.", "displayEligible": true, "domainCuration": {"method": "rules-v1", "state": "auto"}, "evaluationMode": "public_reusable", "evidence": {"reasonCodes": ["coined title prefix ending in Bench or Benchmark", "benchmark release stated in one sentence", "evaluation protocol evidence"], "snippet": "To handle these issues, we propose PatchBench, a new benchmark for evaluating AI agents on realistic vulnerability patching tasks."}, "familyId": "bmf_071eadf3d8bd", "firstSeenAt": "2026-09-04", "heat": null, "id": "bm_patchbench_a8fb20cc", "indexedAt": "2026-09-04T06:13:54.443310Z", "industrySectors": ["Cybersecurity"], "links": {"code": null, "data": null, "paper": "https://arxiv.org/abs/2609.04075", "pdf": "https://arxiv.org/pdf/2609.04075", "project": null, "report": "https://arxiv.org/abs/2609.04075"}, "metrics": [], "motivation": "AI agents have recently demonstrated strong performance in automated vulnerability patching.", "name": "PatchBench", "oneLine": "Evaluates AI agents on realistic C/C++ vulnerability patching tasks where ground-truth fixes lie outside the crash stack, using vulnerability transplant and code mutations to reduce memorization and surface-level fixes.", "paperTitle": "PatchBench: Evaluating AI Agents for Vulnerability Patching", "primaryDomain": "Cybersecurity", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2609.04075", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-09-04T06:13:54.443310Z"}, "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}}, "readiness": "Paper only", "recognitionConfidence": 1.0, "relation": "introduces", "releasedAt": "2026-09-03", "source": {"authors": ["Chihao Shen", "Jiacheng Li", "Aastha Mahajan", "Jeffery Siyuan Tian", "Yonghwi Kwon", "Yizheng Chen"], "categories": ["cs.CR", "cs.AI", "cs.SE"], "id": "2609.04075", "title": "PatchBench: Evaluating AI Agents for Vulnerability Patching", "type": "arxiv", "url": "https://arxiv.org/abs/2609.04075"}, "sourceUpdatedAt": "2026-09-03T16:44:22Z", "topics": ["cs.CR"], "venueAttempts": [], "whyItMatters": "Exposes inflated success rates from PoC-only validation and provides a more rigorous evaluation of security and semantic correctness in automated patching."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
