{"record": {"aliases": ["MC-CXR: A Multi-Context Chest X-ray Benchmark for Context-Induced Disruption in Vision-Language Models"], "artifacts": [{"id": "arxiv:2608.24118", "kind": "paper", "url": "https://arxiv.org/abs/2608.24118"}], "authors": ["Junhyeok Lee", "Songsoo Kim", "Kyu Sung Choi"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2608.24118", "source_url": "https://arxiv.org/abs/2608.24118"}, "categories": ["Biology & Drug Discovery", "Multimodal"], "description": {"en": "MC-CXR is a benchmark of 240 chest X-ray cases (2,522 instances) for evaluating context-induced disruption in vision-language models. It pairs reliable and misleading context across text and prior imaging, with visual overlays. Defines three task families and two paired metrics: switch-to-wrong rate and context-aligned error rate."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2608.24118", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2608.24118", "title": "MC-CXR"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "key": "claire-radar:2608.24118", "modality": null, "name": "MC-CXR", "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": "true", "origin_record_id": "bm_mc-cxr-a-multi-context-chest-x-ray-benchma_1a149873", "origin_source": "arxiv", "recognition_confidence": "0.6", "record_sha256": "afa6fcbb271dcb6b18ac2db0f966dabe8f2f50eac11b1ab801936dfacdab5d7e", "relation": "introduces", "review_model": "deepseek-v4-pro", "review_state": "ai-reviewed", "reviewed_at": "2026-08-27T04:18:56.318436Z", "source_url": "https://arxiv.org/abs/2608.24118"}, "publisher": null, "released": "2026-08-25", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2608.24118", "source_url": "https://arxiv.org/abs/2608.24118"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2608-24118", "source": "claire_radar", "source_benchmark_id": "2608.24118", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": ["MC-CXR: A Multi-Context Chest X-ray Benchmark for Context-Induced Disruption in Vision-Language Models"], "annotation": "Unknown", "applicationDomains": ["Biology & Drug Discovery"], "area": "Multimodal", "attention": {"asOf": "2026-08-26", "attemptedAt": "2026-08-26T11:32:18.902760Z", "githubRepo": null, "githubScope": null, "githubStars": null, "hfDailySubmittedAt": null, "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": null, "hfPaperUrl": "https://huggingface.co/papers/2608.24118", "observedAt": "2026-08-24T11:31:30.653116Z", "signalStatus": {"githubStars": {"state": "not_applicable"}, "hfDatasetDownloads": {"state": "not_applicable"}, "hfDatasetLikes": {"state": "not_applicable"}, "hfPaperUpvotes": {"checkedAt": "2026-08-26T11:32:18.902760Z", "state": "unavailable"}}, "status": "unavailable"}, "attentionForecast": {"confidence": "Medium", "horizon": "7d", "reason": "Clinical AI relevance and dataset availability on PhysioNet likely attract medical imaging and VLM researchers.", "score": 70}, "capabilities": [], "confidence": "Medium", "construction": "Unknown", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-08-27T04:22:25.189955Z", "inputHash": "913f7206eab8533f99cbdacbc971d32924b7f9e0cc3c2729f60371d20bbdd9fb", "model": "deepseek-v4-flash"}, "curation": {"canonicalNameEvidence": "We introduce Multi-Context Chest X-ray (MC-CXR), a benchmark", "canonicalNameSource": "abstract", "decisionReason": "The dataset is available on PhysioNet, providing a public reuse path, and the benchmark defines clear paired metrics.", "model": "deepseek-v4-pro", "reviewedAt": "2026-08-27T04:18:56.318436Z", "state": "ai-reviewed"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "MC-CXR is a benchmark of 240 chest X-ray cases (2,522 instances) for evaluating context-induced disruption in vision-language models. It pairs reliable and misleading context across text and prior imaging, with visual overlays. Defines three task families and two paired metrics: switch-to-wrong rate and context-aligned error rate.", "displayEligible": true, "domainCuration": {"method": "rules-v1", "state": "auto"}, "evaluationMode": "public_reusable", "evidence": {"reasonCodes": ["benchmark term in abstract", "benchmark release stated in one sentence", "evaluation protocol evidence"], "snippet": "We introduce Multi-Context Chest X-ray (MC-CXR), a benchmark of 240 cases expanded into 2,522 instances that isolates context-induced disruption through paired perturbation."}, "familyId": "bmf_bfa02518247e", "firstSeenAt": "2026-08-26", "heat": null, "id": "bm_mc-cxr-a-multi-context-chest-x-ray-benchma_1a149873", "indexedAt": "2026-08-26T06:08:22.244942Z", "industrySectors": ["Pharma & Biotech"], "links": {"code": null, "data": null, "paper": "https://arxiv.org/abs/2608.24118", "pdf": "https://arxiv.org/pdf/2608.24118", "project": null, "report": "https://arxiv.org/abs/2608.24118"}, "metrics": [], "motivation": "Vision-language models (VLMs) are increasingly used in clinical pipelines where a chest X-ray is interpreted alongside retrieved reports, preliminary notes, or prior imaging.", "name": "MC-CXR", "oneLine": "MC-CXR is a benchmark of 240 chest X-ray cases (2,522 instances) for evaluating context-induced disruption in vision-language models. It pairs reliable and misleading context across text and prior imaging, with visual overlays. Defines three task families and two paired metrics: switch-to-wrong rate and context-aligned error rate.", "paperTitle": "MC-CXR: A Multi-Context Chest X-ray Benchmark for Context-Induced Disruption in Vision-Language Models", "primaryDomain": "Biology & Drug Discovery", "publication": {"evidence": "15 pages, 3 figures, 4 tables. Accepted to Findings of EMNLP 2026", "evidenceLevel": "author-claim", "evidenceUrl": "https://arxiv.org/abs/2608.24118", "source": "arxiv-comments", "status": "acceptance_claimed", "venue": "Findings of EMNLP 2026", "verifiedAt": "2026-08-26T06:08:22.244942Z"}, "publications": [], "ranking": {"30d": {"confidence": "Low", "coverage": 0.0, "growth": {"components": {"githubStars": {"percentile": null, "value": null}, "hfDatasetDownloads": {"percentile": null, "value": null}, "hfPaperUpvotes": {"percentile": null, "value": null}}, "confidence": "Low", "coverage": 0.0, "elapsedDays": null, "historyDate": null, "rank": null, "score": null, "windowDays": 30}, "level": {"components": {"githubStars": {"percentile": null, "value": null}, "hfDatasetDownloads": {"percentile": null, "value": null}, "hfPaperUpvotes": {"percentile": null, "value": null}}, "confidence": "Low", "coverage": 0.0, "rank": null, "score": null}, "method": "attention-ranking-v12", "rank": null, "score": null}, "90d": {"confidence": "Low", "coverage": 0.0, "growth": {"components": {"githubStars": {"percentile": null, "value": null}, "hfDatasetDownloads": {"percentile": null, "value": null}, "hfPaperUpvotes": {"percentile": null, "value": null}}, "confidence": "Low", "coverage": 0.0, "elapsedDays": null, "historyDate": null, "rank": null, "score": null, "windowDays": 90}, "level": {"components": {"githubStars": {"percentile": null, "value": null}, "hfDatasetDownloads": {"percentile": null, "value": null}, "hfPaperUpvotes": {"percentile": null, "value": null}}, "confidence": "Low", "coverage": 0.0, "rank": null, "score": null}, "method": "attention-ranking-v12", "rank": null, "score": null}, "today": {"confidence": "Low", "coverage": 0.0, "growth": {"components": {"githubStars": {"percentile": null, "value": null}, "hfDatasetDownloads": {"percentile": null, "value": null}, "hfPaperUpvotes": {"percentile": null, "value": null}}, "confidence": "Low", "coverage": 0.0, "elapsedDays": 2, "historyDate": "2026-08-24", "rank": null, "score": null, "windowDays": 1}, "level": {"components": {"githubStars": {"percentile": null, "value": null}, "hfDatasetDownloads": {"percentile": null, "value": null}, "hfPaperUpvotes": {"percentile": null, "value": null}}, "confidence": "Low", "coverage": 0.0, "rank": null, "score": null}, "method": "attention-ranking-v12", "rank": null, "score": null}}, "readiness": "Paper only", "recognitionConfidence": 0.6, "relation": "introduces", "releasedAt": "2026-08-25", "source": {"authors": ["Junhyeok Lee", "Songsoo Kim", "Kyu Sung Choi"], "categories": ["cs.CL"], "id": "2608.24118", "title": "MC-CXR: A Multi-Context Chest X-ray Benchmark for Context-Induced Disruption in Vision-Language Models", "type": "arxiv", "url": "https://arxiv.org/abs/2608.24118"}, "sourceUpdatedAt": "2026-08-25T06:28:17Z", "topics": ["Multimodal"], "venueAttempts": [{"decisionRaw": "15 pages, 3 figures, 4 tables. Accepted to Findings of EMNLP 2026", "evidence": [{"level": "author-claim", "observedAt": "2026-08-26T06:08:22.244942Z", "rawValue": "15 pages, 3 figures, 4 tables. Accepted to Findings of EMNLP 2026", "sourceType": "arxiv-comments", "sourceUrl": "https://arxiv.org/abs/2608.24118"}], "reviewStatus": "accepted", "venueName": "Findings of EMNLP 2026"}], "whyItMatters": "VLMs in clinical pipelines may be disrupted by plausible but misleading context, reducing diagnostic accuracy. MC-CXR isolates this disruption with paired perturbation, revealing high switch rates and text-visual asymmetry, supporting safer deployment."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
