{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2607.15216", "kind": "paper", "url": "https://arxiv.org/abs/2607.15216"}, {"id": "gh:stanford-aimi/symbal", "kind": "repo", "url": "https://github.com/Stanford-AIMI/Symbal"}], "authors": ["Maya Varma", "Jean-Benoit Delbrouck", "Sophie Ostmeier", "Akshay Chaudhari", "Curtis Langlotz"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2607.15216", "source_url": "https://arxiv.org/abs/2607.15216"}, "categories": ["Biology & Drug Discovery", "Vision & 3D", "cs.CV"], "description": {"en": "SymbalBench evaluates automated detection of systematic misalignments in MLLM-generated image captions. It comprises 420 vision-language datasets (1.7 million image-text pairs) from natural and medical domains, each annotated with known systematic misalignments."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2607.15216", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2607.15216", "title": "SymbalBench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "key": "claire-radar:2607.15216", "modality": null, "name": "SymbalBench", "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_symbalbench_84a0aab6", "origin_source": "arxiv", "recognition_confidence": "1.0", "record_sha256": "60441aae96a48495dbe70365655265f8f6f817ce4e50d2193b1303cc3fbc36ed", "relation": "introduces", "review_state": "unreviewed", "source_url": "https://arxiv.org/abs/2607.15216"}, "publisher": null, "released": "2026-07-16", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2607.15216", "source_url": "https://arxiv.org/abs/2607.15216"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2607-15216", "source": "claire_radar", "source_benchmark_id": "2607.15216", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Unknown", "applicationDomains": ["Biology & Drug Discovery"], "area": "Vision & 3D", "attention": {"asOf": "2026-10-05", "attemptedAt": "2026-10-04T15:31:48.030916Z", "githubRepo": "https://github.com/Stanford-AIMI/Symbal", "githubScope": "benchmark_repo", "githubStars": null, "hfDailySubmittedAt": null, "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": null, "hfPaperUrl": "https://huggingface.co/papers/2607.15216", "observedAt": "2026-10-04T03:02:42.754940Z", "signalStatus": {"githubStars": {"checkedAt": "2026-10-04T15:31:48.030916Z", "state": "unavailable"}, "hfDatasetDownloads": {"state": "not_applicable"}, "hfDatasetLikes": {"state": "not_applicable"}, "hfPaperUpvotes": {"checkedAt": "2026-10-04T15:31:48.030916Z", "state": "unavailable"}}, "status": "unavailable"}, "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": "0718cfa0851993fa77465aa34998a9780399ae8aa865c14932469950b3c6d10a", "model": "deepseek-v4-flash"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "SymbalBench evaluates automated detection of systematic misalignments in MLLM-generated image captions. It comprises 420 vision-language datasets (1.7 million image-text pairs) from natural and medical domains, each annotated with known systematic misalignments.", "domainCuration": {"method": "rules-v1", "state": "auto"}, "evidence": {"reasonCodes": ["exact named benchmark artifact released in abstract", "benchmark release stated in one sentence", "evaluation protocol evidence", "public artifact URL"], "snippet": "As our second key contribution, we introduce SymbalBench, a benchmark designed to evaluate automated methods on our proposed task."}, "familyId": "bmf_26aed8f715b6", "firstSeenAt": "2026-08-19", "githubIndex": {"candidateCount": 1, "indexedAt": "2026-09-17T04:40:34.111665+00:00", "repositories": [], "status": "candidates"}, "heat": null, "id": "bm_symbalbench_84a0aab6", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": ["Pharma & Biotech"], "links": {"code": "https://github.com/Stanford-AIMI/Symbal", "data": null, "hfPaper": "https://huggingface.co/papers/2607.15216", "paper": "https://arxiv.org/abs/2607.15216", "pdf": "https://arxiv.org/pdf/2607.15216", "project": null, "report": "https://arxiv.org/abs/2607.15216"}, "metrics": [], "motivation": "Multimodal large language models (MLLMs) often introduce errors when generating image captions, resulting in misaligned image-text pairs.", "name": "SymbalBench", "oneLine": "SymbalBench evaluates automated detection of systematic misalignments in MLLM-generated image captions. It comprises 420 vision-language datasets (1.7 million image-text pairs) from natural and medical domains, each annotated with known systematic misalignments.", "paperTitle": "Symbal: Detecting Systematic Misalignments in Model-Generated Captions", "primaryDomain": "Biology & Drug Discovery", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2607.15216", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-10-14T01:46:58.058199Z"}, "publications": [], "publishers": [{"name": "Stanford AIMI", "organizationType": "academic-lab", "role": "benchmark-publisher", "sourceUrl": "https://github.com/Stanford-AIMI/Symbal"}], "ranking": {"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": "Runnable", "recognitionConfidence": 1.0, "relation": "introduces", "releasedAt": "2026-07-16", "source": {"authors": ["Maya Varma", "Jean-Benoit Delbrouck", "Sophie Ostmeier", "Akshay Chaudhari", "Curtis Langlotz"], "categories": ["cs.CV", "cs.AI"], "id": "2607.15216", "title": "Symbal: Detecting Systematic Misalignments in Model-Generated Captions", "type": "arxiv", "url": "https://arxiv.org/abs/2607.15216"}, "sourceUpdatedAt": "2026-07-16T17:18:37Z", "topics": ["cs.CV"], "venueAttempts": [], "whyItMatters": "MLLM-generated captions often contain recurring errors tied to visual features, which can degrade downstream tasks. SymbalBench provides a standardized testbed to assess methods for surfacing such systematic captioning failures, aiding dataset auditing and model improvement."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
