{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2609.07823", "kind": "paper", "url": "https://arxiv.org/abs/2609.07823"}, {"id": "gh:risys-lab/safire", "kind": "repo", "url": "https://github.com/RISys-Lab/SAFIRE"}, {"kind": "website", "url": "https://risys-lab.github.io/SAFIRE/"}], "authors": ["Pengfei Li", "Naufal Suryanto", "Sicheng Zhang", "Mohammad Alsharid", "Muzammal Naseer"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2609.07823", "source_url": "https://arxiv.org/abs/2609.07823"}, "categories": ["General AI", "Multimodal", "Reasoning", "Safety", "Safety & Trustworthiness"], "description": {"en": "Evaluates MLLMs on 193K multiple-choice VQA over 83K fire and smoke images across 20 scenarios and 10 reasoning dimensions."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2609.07823", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2609.07823", "title": "SAFIRE"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "key": "claire-radar:2609.07823", "modality": null, "name": "SAFIRE", "openness": {"code_license": null, "data_license": null, "evidence": [], "status": "unknown"}, "provenance": {"admission_policy_version": "2026-09-09.1", "confidence": "High", "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_safire_2546967e", "origin_source": "arxiv", "recognition_confidence": "1.0", "record_sha256": "4cd35ccfb06c4a26989207a971b382251bb2385ad042b90c4041ddb7145a22e3", "relation": "introduces", "review_model": "deepseek-v4-pro", "review_state": "ai-reviewed", "reviewed_at": "2026-09-09T06:27:57.312318Z", "source_url": "https://arxiv.org/abs/2609.07823"}, "publisher": null, "released": "2026-09-07", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2609.07823", "source_url": "https://arxiv.org/abs/2609.07823"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2609-07823", "source": "claire_radar", "source_benchmark_id": "2609.07823", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Unknown", "applicationDomains": ["General AI"], "area": "Safety & Trustworthiness", "attention": {"asOf": "2026-10-05", "attemptedAt": "2026-10-04T15:31:48.030916Z", "githubRepo": "https://github.com/RISys-Lab/SAFIRE", "githubScope": "benchmark_repo", "githubStars": 5, "hfDailySubmittedAt": null, "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": 2, "hfPaperUrl": "https://huggingface.co/papers/2609.07823", "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": {"observedAt": "2026-10-04T15:31:48.030916Z", "state": "fresh"}}, "status": "fresh"}, "attentionForecast": {"confidence": "Medium", "horizon": "7d", "reason": "Safety-critical focus, large-scale data, and EMNLP 2026 acceptance suggest moderate community interest; public code and project page facilitate uptake.", "score": 65}, "capabilities": ["Reasoning"], "confidence": "High", "construction": "Transform Existing", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-09-09T06:27:57.312318Z", "inputHash": "6ba06e7707ce346a3c0bc12f678f20111c921d20ad846b12ee2bdf9aa0455728", "model": "deepseek-v4-pro"}, "curation": {"admissionPolicyVersion": "2026-09-09.1", "canonicalNameEvidence": "We introduce SAFIRE, a large-scale benchmark for fire-smoke understanding in MLLMs", "canonicalNameSource": "abstract", "decisionReason": "The benchmark has a clearly defined task suite, scoring via accuracy, and publicly available datasets, model, and code for independent evaluation.", "evaluationEvidence": {"reuse": "All datasets, models, and code are available at https://risys-lab.github.io/SAFIRE/.", "scoring": "Evaluating ten open-source MLLMs (8B-38B) yields an average accuracy of 61.9%", "task": "The benchmark comprises 83K captioned images from 20 scenarios and 193K MCVQA generated from a 9.7K-image subset, spanning 10 evaluation dimensions from basic perception to higher-order reasoning."}, "model": "deepseek-v4-pro", "reviewedAt": "2026-09-09T06:27:57.312318Z", "state": "ai-reviewed"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "Evaluates MLLMs on 193K multiple-choice VQA over 83K fire and smoke images across 20 scenarios and 10 reasoning dimensions.", "displayEligible": true, "domainCuration": {"method": "rules-v1", "state": "auto"}, "evaluationMode": "score_submission", "evidence": {"reasonCodes": ["exact coined title identity tied to benchmark evidence", "benchmark release stated in one sentence", "evaluation protocol evidence", "public artifact URL"], "snippet": "We introduce SAFIRE, a large-scale benchmark for fire-smoke understanding in MLLMs, comprising 83K captioned images from 20 scenarios and 193K multiple-choice VQA (MCVQA) generated from a 9.7K-image subset, spanning 10 evaluation dimensions from basic perception to higher-order reasoning."}, "familyId": "bmf_bdd12e76e245", "firstSeenAt": "2026-09-09", "heat": null, "id": "bm_safire_2546967e", "indexedAt": "2026-09-09T06:18:29.794093Z", "industrySectors": [], "links": {"code": "https://github.com/RISys-Lab/SAFIRE", "data": null, "paper": "https://arxiv.org/abs/2609.07823", "pdf": "https://arxiv.org/pdf/2609.07823", "project": "https://risys-lab.github.io/SAFIRE/", "report": "https://arxiv.org/abs/2609.07823"}, "metrics": [], "motivation": "Multimodal Large Language Models (MLLMs) show strong progress on vision-language tasks, yet their reliability in safety-critical settings remains underexplored.", "name": "SAFIRE", "oneLine": "Evaluates MLLMs on 193K multiple-choice VQA over 83K fire and smoke images across 20 scenarios and 10 reasoning dimensions.", "paperTitle": "SAFIRE: Safety-Critical Benchmark for Fine-grained Fire and Smoke Understanding in Multimodal LLMs", "primaryDomain": "General AI", "publication": {"evidence": "Accepted at EMNLP 2026 Findings. Project page: https://risys-lab.github.io/SAFIRE/ | Github: https://github.com/RISys-Lab/SAFIRE", "evidenceLevel": "author-claim", "evidenceUrl": "https://arxiv.org/abs/2609.07823", "source": "arxiv-comments", "status": "acceptance_claimed", "venue": "EMNLP 2026 Findings", "verifiedAt": "2026-09-09T06:18:29.794093Z"}, "publications": [], "publishers": [{"name": "RISys-Lab", "organizationType": "academic-lab", "role": "benchmark-publisher", "sourceUrl": "https://risys-lab.github.io/SAFIRE/"}], "ranking": {"30d": {"confidence": "High", "coverage": 0.85, "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": 30, "historyDate": "2026-09-05", "rank": null, "score": null, "windowDays": 30}, "level": {"components": {"githubStars": {"normalization": "log-max", "normalized": 0.26418589365088024, "value": 5}, "hfDatasetDownloads": {"normalization": "log-max", "normalized": null, "value": null}, "hfPaperUpvotes": {"normalization": "log-max", "normalized": 0.2045589896236431, "value": 2}}, "confidence": "High", "coverage": 0.85, "rank": 85, "score": 28}, "method": "attention-ranking-v13", "rank": 85, "score": 28}, "90d": {"confidence": "Medium", "coverage": 0.7, "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": 0.2391676802567412, "value": 5}, "hfDatasetDownloads": {"normalization": "log-max", "normalized": null, "value": null}, "hfPaperUpvotes": {"normalization": "log-max", "normalized": 0.19411886932144953, "value": 2}}, "confidence": "Medium", "coverage": 0.7, "rank": 236, "score": 31}, "method": "attention-ranking-v13", "rank": 236, "score": 31}}, "readiness": "Runnable", "recognitionConfidence": 1.0, "relation": "introduces", "releasedAt": "2026-09-07", "source": {"authors": ["Pengfei Li", "Naufal Suryanto", "Sicheng Zhang", "Mohammad Alsharid", "Muzammal Naseer"], "categories": ["cs.CV", "cs.AI", "cs.CL"], "id": "2609.07823", "title": "SAFIRE: Safety-Critical Benchmark for Fine-grained Fire and Smoke Understanding in Multimodal LLMs", "type": "arxiv", "url": "https://arxiv.org/abs/2609.07823"}, "sourceUpdatedAt": "2026-09-07T17:59:50Z", "topics": ["Multimodal", "Safety", "Reasoning"], "venueAttempts": [{"decisionRaw": "Accepted at EMNLP 2026 Findings. Project page: https://risys-lab.github.io/SAFIRE/ | Github: https://github.com/RISys-Lab/SAFIRE", "evidence": [{"level": "author-claim", "observedAt": "2026-09-09T06:18:29.794093Z", "rawValue": "Accepted at EMNLP 2026 Findings. Project page: https://risys-lab.github.io/SAFIRE/ | Github: https://github.com/RISys-Lab/SAFIRE", "sourceType": "arxiv-comments", "sourceUrl": "https://arxiv.org/abs/2609.07823"}], "reviewStatus": "accepted", "venueName": "EMNLP 2026 Findings"}], "whyItMatters": "Provides safety-critical, context-aware evaluation for fire-smoke understanding, addressing a gap in existing benchmarks that lack diverse real-world scenarios."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
