{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2605.26177", "kind": "paper", "url": "https://arxiv.org/abs/2605.26177"}], "authors": ["Hanyu Li", "Yichi Zhang", "Speed Zhu", "Hang Su", "Jun Zhu", "Yinpeng Dong"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2605.26177", "source_url": "https://arxiv.org/abs/2605.26177"}, "categories": ["Code & Software", "Reasoning", "Software & AI Compute"], "description": {"en": "RepoMirage is a two-stage evaluation suite built on SWE-Bench Verified that applies semantics-preserving repository-level perturbations and extended tasks to probe repository context reasoning in code agents."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2605.26177", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2605.26177", "title": "RepoMirage"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "institutions": ["Beijing University of Posts and Telecommunications", "Tsinghua University", "Tencent (China)"], "institutions_reference": {"basis": "openalex_linked_institutions", "doi": "10.48550/arxiv.2605.26177", "retrieved_at": "2026-11-04", "source_key": "claire-radar:2605.26177", "source_url": "https://openalex.org/W7162566353"}, "key": "claire-radar:2605.26177", "modality": null, "name": "RepoMirage", "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": "unknown", "origin_record_id": "bm_repomirage_322a2daa", "origin_source": "arxiv", "recognition_confidence": "1.0", "record_sha256": "57bc6df9804ca912422555969909e7de5d020358862569ea176785adf91ce9f8", "relation": "introduces", "review_state": "unreviewed", "source_url": "https://arxiv.org/abs/2605.26177"}, "publisher": null, "released": "2026-05-25", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2605.26177", "source_url": "https://arxiv.org/abs/2605.26177"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2605-26177", "source": "claire_radar", "source_benchmark_id": "2605.26177", "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/2605.26177", "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": ["Reasoning"], "confidence": "High", "construction": "Unknown", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-08-20T15:22:10.044731Z", "inputHash": "cad8c803504208bfc383b6e69168d3ddfababa4d8939051acf7b1c39d5ced261", "model": "deepseek-v4-flash"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "RepoMirage is a two-stage evaluation suite built on SWE-Bench Verified that applies semantics-preserving repository-level perturbations and extended tasks to probe repository context reasoning in code agents.", "domainCuration": {"method": "rules-v1", "state": "auto"}, "evidence": {"reasonCodes": ["exact coined title identity tied to benchmark evidence", "benchmark release stated in one sentence", "evaluation protocol evidence"], "snippet": "To investigate this question, we introduce RepoMirage, a two-stage evaluation suite built on SWE-Bench Verified that adopts perturbation as a diagnostic tool to increase the demand for context reasoning by transforming how the repository is exposed."}, "familyId": "bmf_defa1b39bb7c", "firstSeenAt": "2026-08-19", "githubIndex": {"candidateCount": 0, "indexedAt": "2026-09-17T04:40:34.111665+00:00", "repositories": [], "status": "not-found"}, "heat": null, "id": "bm_repomirage_322a2daa", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": ["Software & Cloud"], "links": {"code": null, "data": null, "hfPaper": "https://huggingface.co/papers/2605.26177", "paper": "https://arxiv.org/abs/2605.26177", "pdf": "https://arxiv.org/pdf/2605.26177", "project": null, "report": "https://arxiv.org/abs/2605.26177"}, "metrics": [], "motivation": "Code agents are currently having skillful performance on repository-level software engineering benchmarks, but it remains unclear whether success on end-to-end tasks such as issue resolution truly reflects repository context reasoning, the ability to identify the task-relevant information across multiple files and reason over the relations among them.", "name": "RepoMirage", "oneLine": "RepoMirage is a two-stage evaluation suite built on SWE-Bench Verified that applies semantics-preserving repository-level perturbations and extended tasks to probe repository context reasoning in code agents.", "paperTitle": "RepoMirage: Probing Repository Context Reasoning in Code Agents with Perturbations", "primaryDomain": "Software & AI Compute", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2605.26177", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-10-14T01:46:58.058199Z"}, "publications": [], "ranking": {}, "readiness": "Paper only", "recognitionConfidence": 1.0, "relation": "introduces", "releasedAt": "2026-05-25", "source": {"authors": ["Hanyu Li", "Yichi Zhang", "Speed Zhu", "Hang Su", "Jun Zhu", "Yinpeng Dong"], "categories": ["cs.SE", "cs.AI"], "id": "2605.26177", "title": "RepoMirage: Probing Repository Context Reasoning in Code Agents with Perturbations", "type": "arxiv", "url": "https://arxiv.org/abs/2605.26177"}, "sourceUpdatedAt": "2026-05-25T06:26:43Z", "topics": ["Reasoning"], "venueAttempts": [], "whyItMatters": "It aims to isolate repository context reasoning from end-to-end issue resolution performance, revealing a gap that could inform structure-aware agent design."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
