{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2609.03331", "kind": "paper", "url": "https://arxiv.org/abs/2609.03331"}, {"id": "gh:lab-klc/fpco-dialog", "kind": "repo", "url": "https://github.com/lab-klc/FPCO-Dialog"}], "authors": ["Jiayuan Ma", "Yuqi Lu", "Weiyang Guo", "Chenrui Wang", "Junyi Shu", "Xuebo Liu", "Min Zhang", "Jing Li"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2609.03331", "source_url": "https://arxiv.org/abs/2609.03331"}, "categories": ["General AI", "Multimodal"], "description": {"en": "A benchmark with 1,080 images and 10,800 question turns across 10-turn dialogues, evaluating correction and cooperation when false visual premises recur, scored by CorrTP@K and two detectors."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2609.03331", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2609.03331", "title": "FPCO-Dialog"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "key": "claire-radar:2609.03331", "modality": null, "name": "FPCO-Dialog", "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": "true", "origin_record_id": "bm_fpco-dialog_09eb84bc", "origin_source": "arxiv", "recognition_confidence": "1.0", "record_sha256": "84b0630033e3c06d1076f350db73009257f8879b25dd5d6861801c67773b6b9f", "relation": "introduces", "review_model": "deepseek-v4-pro", "review_state": "ai-reviewed", "reviewed_at": "2026-09-07T00:42:15.394147Z", "source_url": "https://arxiv.org/abs/2609.03331"}, "publisher": null, "released": "2026-09-03", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2609.03331", "source_url": "https://arxiv.org/abs/2609.03331"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2609-03331", "source": "claire_radar", "source_benchmark_id": "2609.03331", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Mixed", "applicationDomains": ["General AI"], "area": "Multimodal", "attention": {"asOf": "2026-09-04", "attemptedAt": "2026-09-04T06:36:46.783164Z", "githubRepo": "https://github.com/lab-klc/FPCO-Dialog", "githubScope": "benchmark_repo", "githubStars": 1, "hfDailySubmittedAt": null, "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": null, "hfPaperUrl": "https://huggingface.co/papers/2609.03331", "observedAt": "2026-09-04T06:36:46.783164Z", "signalStatus": {"githubStars": {"observedAt": "2026-09-04T06:36:46.783164Z", "state": "fresh"}, "hfDatasetDownloads": {"state": "not_applicable"}, "hfDatasetLikes": {"state": "not_applicable"}, "hfPaperUpvotes": {"checkedAt": "2026-09-04T06:36:46.783164Z", "state": "unavailable"}}, "status": "fresh"}, "attentionForecast": {"confidence": "Medium", "horizon": "7d", "reason": "The benchmark addresses an emerging concern in multimodal dialogue reliability with a sizable public dataset, likely to draw attention from VLM robustness researchers.", "score": 35}, "capabilities": [], "confidence": "High", "construction": "Aggregate Existing", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-09-07T00:42:15.394147Z", "inputHash": "fbb2d046d9a9ea18859e4566de4e4a05fa6fb83b119c963e1ccce2ff38ce21b3", "model": "deepseek-v4-pro"}, "curation": {"canonicalNameEvidence": "FPCO-Dialog: A Multi-Turn False-Premise Benchmark for Correction and Cooperation in Vision-Language Models", "canonicalNameSource": "paper_title", "decisionReason": "The paper includes a GitHub repository with dataset and code, a clear 10-turn protocol, and defined scoring metrics, supporting a reusable public benchmark.", "model": "deepseek-v4-pro", "reviewedAt": "2026-09-07T00:42:15.394147Z", "state": "ai-reviewed"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "A benchmark with 1,080 images and 10,800 question turns across 10-turn dialogues, evaluating correction and cooperation when false visual premises recur, scored by CorrTP@K and two detectors.", "displayEligible": true, "domainCuration": {"method": "rules-v1", "state": "auto"}, "evaluationMode": "public_reusable", "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 FPCO-Dialog, a benchmark for evaluating correction and cooperation behavior in VLMs under repeated false premises."}, "familyId": "bmf_b4dae55fad8b", "firstSeenAt": "2026-09-04", "heat": null, "id": "bm_fpco-dialog_09eb84bc", "indexedAt": "2026-09-04T00:41:56.430646Z", "industrySectors": [], "links": {"code": "https://github.com/lab-klc/FPCO-Dialog", "data": null, "paper": "https://arxiv.org/abs/2609.03331", "pdf": "https://arxiv.org/pdf/2609.03331", "project": null, "report": "https://arxiv.org/abs/2609.03331"}, "metrics": [], "motivation": "Vision-language models (VLMs) are increasingly deployed in multi-turn settings where users may describe visual content with incorrect assumptions.", "name": "FPCO-Dialog", "oneLine": "A benchmark with 1,080 images and 10,800 question turns across 10-turn dialogues, evaluating correction and cooperation when false visual premises recur, scored by CorrTP@K and two detectors.", "paperTitle": "FPCO-Dialog: A Multi-Turn False-Premise Benchmark for Correction and Cooperation in Vision-Language Models", "primaryDomain": "General AI", "publication": {"evidence": "Accepted at EMNLP2026 Main Conference. For code and data, see https://github.com/lab-klc/FPCO-Dialog", "evidenceLevel": "author-claim", "evidenceUrl": "https://arxiv.org/abs/2609.03331", "source": "arxiv-comments", "status": "acceptance_claimed", "venue": "EMNLP2026 Main Conference", "verifiedAt": "2026-09-04T00:41:56.430646Z"}, "publications": [], "publishers": [{"name": "lab-klc", "organizationType": "academic-lab", "role": "benchmark-publisher", "sourceUrl": "https://github.com/lab-klc/FPCO-Dialog"}], "ranking": {"30d": {"confidence": "Low", "coverage": 0.55, "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": 30}, "level": {"components": {"githubStars": {"normalization": "log-max", "normalized": 0.10323645750017778, "value": 1}, "hfDatasetDownloads": {"normalization": "log-max", "normalized": null, "value": null}, "hfPaperUpvotes": {"normalization": "log-max", "normalized": null, "value": null}}, "confidence": "Low", "coverage": 0.55, "rank": 94, "score": 28}, "method": "attention-ranking-v13", "rank": 94, "score": 28}, "90d": {"confidence": "Low", "coverage": 0.55, "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": 0.09399878027980717, "value": 1}, "hfDatasetDownloads": {"normalization": "log-max", "normalized": null, "value": null}, "hfPaperUpvotes": {"normalization": "log-max", "normalized": null, "value": null}}, "confidence": "Low", "coverage": 0.55, "rank": 256, "score": 28}, "method": "attention-ranking-v13", "rank": 256, "score": 28}, "today": {"confidence": "Low", "coverage": 0.25, "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": 1, "historyDate": "2026-09-03", "rank": null, "score": null, "windowDays": 1}, "level": {"components": {"githubStars": {"normalization": "percentile", "normalized": 0.7142857142857143, "percentile": 0.7142857142857143, "value": 1}, "hfDatasetDownloads": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "hfPaperUpvotes": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "llmAttentionForecast": {"maxContribution": 0.25, "percentile": 0.45, "role": "bonus", "value": 45}}, "confidence": "Low", "coverage": 0.25, "rank": 5, "score": 54}, "method": "attention-ranking-v13", "rank": 5, "score": 54}}, "readiness": "Runnable", "recognitionConfidence": 1.0, "relation": "introduces", "releasedAt": "2026-09-03", "source": {"authors": ["Jiayuan Ma", "Yuqi Lu", "Weiyang Guo", "Chenrui Wang", "Junyi Shu", "Xuebo Liu", "Min Zhang and Jing Li"], "categories": ["cs.CL"], "id": "2609.03331", "title": "FPCO-Dialog: A Multi-Turn False-Premise Benchmark for Correction and Cooperation in Vision-Language Models", "type": "arxiv", "url": "https://arxiv.org/abs/2609.03331"}, "sourceUpdatedAt": "2026-09-03T03:35:29Z", "topics": ["Multimodal"], "venueAttempts": [{"decisionRaw": "Accepted at EMNLP2026 Main Conference. For code and data, see https://github.com/lab-klc/FPCO-Dialog", "evidence": [{"level": "author-claim", "observedAt": "2026-09-04T00:41:56.430646Z", "rawValue": "Accepted at EMNLP2026 Main Conference. For code and data, see https://github.com/lab-klc/FPCO-Dialog", "sourceType": "arxiv-comments", "sourceUrl": "https://arxiv.org/abs/2609.03331"}], "reviewStatus": "accepted", "venueName": "EMNLP2026 Main Conference"}], "whyItMatters": "It isolates persistent refusal to correct false visual premises in multi-turn VLM use, a key reliability signal for deployed visual assistants."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
