{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2607.01360", "kind": "paper", "url": "https://arxiv.org/abs/2607.01360"}], "authors": ["Cuong Chi Le", "Aashish Yadavally", "Minh Le-Anh", "Tien N. Nguyen"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2607.01360", "source_url": "https://arxiv.org/abs/2607.01360"}, "categories": ["Code & Software", "Software & AI Compute", "cs.SE"], "description": {"en": "PAIR-Bench evaluates code improvement by transforming incorrect programs into more correct ones through feedback-guided refinement. It uses progressive hinting with failure-region and hint-depth controls to measure repair of targeted failures, generalization beyond the hint, preservation of correct behavior, and required assistance."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2607.01360", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2607.01360", "title": "PAIR-Bench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "institutions": ["University of Central Florida", "The University of Texas at Dallas"], "institutions_reference": {"basis": "openalex_linked_institutions", "doi": "10.48550/arxiv.2607.01360", "retrieved_at": "2026-11-04", "source_key": "claire-radar:2607.01360", "source_url": "https://openalex.org/W7167258645"}, "key": "claire-radar:2607.01360", "modality": null, "name": "PAIR-Bench", "openness": {"code_license": null, "data_license": null, "evidence": [], "status": "unknown"}, "provenance": {"confidence": "Medium", "crawl_bundle": "claire_radar_2026-10-14", "crawled_at": "2026-10-14T09:44:51Z", "data_status": "primary-source-indexed", "display_eligible": "unknown", "origin_record_id": "bm_pair-bench_c3cd8d09", "origin_source": "arxiv", "recognition_confidence": "0.85", "record_sha256": "7dbf934643df9f26e78359250e43a129e9cd4e2503a2bbb23ac3faff1db08f23", "relation": "introduces", "review_state": "unreviewed", "source_url": "https://arxiv.org/abs/2607.01360"}, "publisher": null, "released": "2026-07-01", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2607.01360", "source_url": "https://arxiv.org/abs/2607.01360"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2607-01360", "source": "claire_radar", "source_benchmark_id": "2607.01360", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Unknown", "applicationDomains": ["Software & AI Compute"], "area": "Code & Software", "attention": {"asOf": "2026-10-05", "attemptedAt": "2026-10-04T15:31:48.030916Z", "githubRepo": null, "githubScope": null, "githubStars": null, "hfDailySubmittedAt": null, "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": null, "hfPaperUrl": "https://huggingface.co/papers/2607.01360", "observedAt": "2026-10-04T03:02:42.754940Z", "signalStatus": {"githubStars": {"state": "not_applicable"}, "hfDatasetDownloads": {"state": "not_applicable"}, "hfDatasetLikes": {"state": "not_applicable"}, "hfPaperUpvotes": {"checkedAt": "2026-10-04T15:31:48.030916Z", "state": "unavailable"}}, "status": "unavailable"}, "capabilities": [], "confidence": "Medium", "construction": "Unknown", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-08-20T15:22:10.044731Z", "inputHash": "7ff056914072d0fa9354c5d0518a824348b6b0efb0313c5b220eb8259131b4de", "model": "deepseek-v4-flash"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "PAIR-Bench evaluates code improvement by transforming incorrect programs into more correct ones through feedback-guided refinement. It uses progressive hinting with failure-region and hint-depth controls to measure repair of targeted failures, generalization beyond the hint, preservation of correct behavior, and required assistance.", "domainCuration": {"method": "rules-v1", "state": "auto"}, "evidence": {"reasonCodes": ["exact named benchmark artifact released in abstract", "benchmark release stated in one sentence"], "snippet": "We introduce PAIR-Bench, a progressive and adaptive benchmark for evaluating code improvement: transforming an incorrect or incomplete program into a more correct one through feedback-guided refinement."}, "familyId": "bmf_7121e1a7ef1c", "firstSeenAt": "2026-08-19", "githubIndex": {"candidateCount": 0, "indexedAt": "2026-09-17T04:40:34.111665+00:00", "repositories": [], "status": "not-found"}, "heat": null, "id": "bm_pair-bench_c3cd8d09", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": ["Software & Cloud"], "links": {"code": null, "data": null, "hfPaper": "https://huggingface.co/papers/2607.01360", "paper": "https://arxiv.org/abs/2607.01360", "pdf": "https://arxiv.org/pdf/2607.01360", "project": null, "report": "https://arxiv.org/abs/2607.01360"}, "metrics": [], "motivation": "Large language models (LLMs) are typically evaluated on code generation and program repair using binary functional correctness: a generated program or patch either passes or fails a test suite.", "name": "PAIR-Bench", "oneLine": "PAIR-Bench evaluates code improvement by transforming incorrect programs into more correct ones through feedback-guided refinement. It uses progressive hinting with failure-region and hint-depth controls to measure repair of targeted failures, generalization beyond the hint, preservation of correct behavior, and required assistance.", "paperTitle": "Benchmarking Code Improvement with Progressive, Adaptive, and Interactive Feedback", "primaryDomain": "Software & AI Compute", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2607.01360", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-10-14T01:46:58.058199Z"}, "publications": [], "ranking": {}, "readiness": "Paper only", "recognitionConfidence": 0.85, "relation": "introduces", "releasedAt": "2026-07-01", "source": {"authors": ["Cuong Chi Le", "Aashish Yadavally", "Minh Le-Anh", "Tien N. Nguyen"], "categories": ["cs.SE"], "id": "2607.01360", "title": "Benchmarking Code Improvement with Progressive, Adaptive, and Interactive Feedback", "type": "arxiv", "url": "https://arxiv.org/abs/2607.01360"}, "sourceUpdatedAt": "2026-07-01T18:20:27Z", "topics": ["cs.SE"], "venueAttempts": [], "whyItMatters": "Traditional binary pass/fail metrics miss partial progress and refinement trajectories. PAIR-Bench provides finer-grained, progressive metrics to assess how LLMs improve code through feedback, offering practical value for developing and selecting models for code improvement tasks."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
