{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2606.30182", "kind": "paper", "url": "https://arxiv.org/abs/2606.30182"}, {"id": "gh:epoch-research/mirrorcode", "kind": "repo", "url": "https://github.com/epoch-research/MirrorCode"}], "authors": ["Tom Adamczewski", "David Owen", "David Rein", "Florian Brand", "Giles Edkins", "Allen Hart", "Daniel O'Connell"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2606.30182", "source_url": "https://arxiv.org/abs/2606.30182"}, "categories": ["Language & Knowledge", "Software & AI Compute", "cs.AI"], "description": {"en": "MirrorCode evaluates AI agents on reimplementing entire software projects from behavior only, matching outputs on end-to-end tests across 25 programs spanning Unix utilities, data serialization, bioinformatics, interpreters, static analysis, cryptography, and compression."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2606.30182", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2606.30182", "title": "MirrorCode"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "institutions": ["University of Warwick"], "institutions_reference": {"basis": "openalex_linked_institutions", "doi": "10.48550/arxiv.2606.30182", "retrieved_at": "2026-11-04", "source_key": "claire-radar:2606.30182", "source_url": "https://openalex.org/W7166643087"}, "key": "claire-radar:2606.30182", "modality": null, "name": "MirrorCode", "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_mirrorcode_114f6eb8", "origin_source": "arxiv", "recognition_confidence": "1.0", "record_sha256": "e7b71b8a84aa964a5db022062f43221c8adb28527a67993b7acd98a31bcfbf09", "relation": "introduces", "review_state": "unreviewed", "source_url": "https://arxiv.org/abs/2606.30182"}, "publisher": null, "released": "2026-06-29", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2606.30182", "source_url": "https://arxiv.org/abs/2606.30182"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2606-30182", "source": "claire_radar", "source_benchmark_id": "2606.30182", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Unknown", "applicationDomains": ["Software & AI Compute"], "area": "Language & Knowledge", "attention": {"asOf": "2026-10-05", "attemptedAt": "2026-10-04T15:31:48.030916Z", "githubRepo": "https://github.com/epoch-research/MirrorCode", "githubScope": "benchmark_repo", "githubStars": 75, "hfDailySubmittedAt": null, "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": null, "hfPaperUrl": "https://huggingface.co/papers/2606.30182", "observedAt": "2026-10-04T15:31:48.030916Z", "signalStatus": {"githubStars": {"observedAt": "2026-10-04T15:31:48.030916Z", "state": "fresh"}, "hfDatasetDownloads": {"state": "not_applicable"}, "hfDatasetLikes": {"state": "not_applicable"}, "hfPaperUpvotes": {"checkedAt": "2026-10-04T15:31:48.030916Z", "state": "unavailable"}}, "status": "fresh"}, "capabilities": [], "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": "67c61e915e20b607b728481872859b99bae0e3b00f5681543f13e73b2ec5d59e", "model": "deepseek-v4-flash"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "MirrorCode evaluates AI agents on reimplementing entire software projects from behavior only, matching outputs on end-to-end tests across 25 programs spanning Unix utilities, data serialization, bioinformatics, interpreters, static analysis, cryptography, and compression.", "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", "public artifact URL"], "snippet": "To address these challenges, we introduce MirrorCode, a long-horizon coding benchmark based on reimplementing entire software projects."}, "familyId": "bmf_5f4db04ac40d", "firstSeenAt": "2026-08-19", "githubIndex": {"candidateCount": 0, "indexedAt": "2026-09-17T04:40:34.111665+00:00", "repositories": [{"evidence": [{"basis": "existing-code-link", "linkedUrl": "https://github.com/epoch-research/MirrorCode", "url": "https://github.com/epoch-research/MirrorCode"}, {"basis": "existing-metric-association", "linkedUrl": "https://github.com/epoch-research/MirrorCode", "url": "https://huggingface.co/papers/2606.30182"}, {"basis": "linked-from-report", "linkedUrl": "https://github.com/epoch-research/MirrorCode", "url": "https://arxiv.org/abs/2606.30182"}], "scope": "benchmark_repo", "url": "https://github.com/epoch-research/MirrorCode", "verification": "source-linked"}], "status": "linked"}, "heat": null, "id": "bm_mirrorcode_114f6eb8", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": ["Software & Cloud"], "links": {"code": "https://github.com/epoch-research/MirrorCode", "data": null, "hfPaper": "https://huggingface.co/papers/2606.30182", "paper": "https://arxiv.org/abs/2606.30182", "pdf": "https://arxiv.org/pdf/2606.30182", "project": null, "report": "https://arxiv.org/abs/2606.30182"}, "metrics": [], "motivation": "AI models are rapidly improving at autonomous coding, as shown by benchmark progress and one-off demonstrations such as AI implementing a C compiler.", "name": "MirrorCode", "oneLine": "MirrorCode evaluates AI agents on reimplementing entire software projects from behavior only, matching outputs on end-to-end tests across 25 programs spanning Unix utilities, data serialization, bioinformatics, interpreters, static analysis, cryptography, and compression.", "paperTitle": "MirrorCode: AI can rebuild entire programs from behavior alone", "primaryDomain": "Software & AI Compute", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2606.30182", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-10-14T01:46:58.058199Z"}, "publications": [], "publishers": [{"name": "Epoch Research", "organizationType": "academic-lab", "role": "benchmark-publisher", "sourceUrl": "https://github.com/epoch-research/MirrorCode"}], "ranking": {}, "readiness": "Runnable", "recognitionConfidence": 1.0, "relation": "introduces", "releasedAt": "2026-06-29", "source": {"authors": ["Tom Adamczewski", "David Owen", "David Rein", "Florian Brand", "Giles Edkins", "Allen Hart", "Daniel O'Connell"], "categories": ["cs.AI"], "id": "2606.30182", "title": "MirrorCode: AI can rebuild entire programs from behavior alone", "type": "arxiv", "url": "https://arxiv.org/abs/2606.30182"}, "sourceUpdatedAt": "2026-07-17T10:56:56Z", "topics": ["cs.AI"], "venueAttempts": [], "whyItMatters": "Existing coding benchmarks focus on shorter tasks, while long-horizon reimplementation remains hard to compare. MirrorCode provides a standardized, repeatable measure of autonomous software engineering capability."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": [{"key": "model-reports:mirrorcode", "name": "MirrorCode", "relation": "equivalent", "slug": "mirrorcode", "source": "model_reports"}]}
