{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2606.27499", "kind": "paper", "url": "https://arxiv.org/abs/2606.27499"}, {"id": "gh:yyyujintang/dmv-bench", "kind": "repo", "url": "https://github.com/yyyujintang/DMV-Bench"}], "authors": ["Yujin Tang", "Chenming Shang", "Ruize Xu", "Nikhil Singh"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2606.27499", "source_url": "https://arxiv.org/abs/2606.27499"}, "categories": ["Agents", "Agents & Tool Use", "General AI", "Multimodal"], "description": {"en": "DMV-Bench evaluates visual memory of multimodal agents in an interactive shopping environment with 1,000 product variants. Agents run autonomous sessions and must recall cued product images via exact URL match, with text leakage controlled."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2606.27499", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2606.27499", "title": "DMV-Bench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "key": "claire-radar:2606.27499", "modality": null, "name": "DMV-Bench", "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_dmv-bench_d16bc24a", "origin_source": "arxiv", "recognition_confidence": "1.0", "record_sha256": "68336019f85403f35cc58b637844d887fb6851c7c62ca2e9fe143925f7532db8", "relation": "introduces", "review_state": "unreviewed", "source_url": "https://arxiv.org/abs/2606.27499"}, "publisher": null, "released": "2026-06-25", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2606.27499", "source_url": "https://arxiv.org/abs/2606.27499"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2606-27499", "source": "claire_radar", "source_benchmark_id": "2606.27499", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Unknown", "applicationDomains": ["General AI"], "area": "Agents & Tool Use", "attention": {"asOf": "2026-10-05", "attemptedAt": "2026-10-04T15:31:48.030916Z", "githubRepo": "https://github.com/yyyujintang/DMV-Bench", "githubScope": "benchmark_repo", "githubStars": 3, "hfDailySubmittedAt": null, "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": 1, "hfPaperUrl": "https://huggingface.co/papers/2606.27499", "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", "lastSuccessfulDate": "2026-10-04", "state": "stale"}}, "status": "fresh"}, "capabilities": [], "confidence": "High", "construction": "Interactive Environment", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-08-20T15:22:10.044731Z", "inputHash": "4228d80f2293441b65e01027ef3ace973887d6a3a795b65dd5354248bcd0fee9", "model": "deepseek-v4-flash"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "DMV-Bench evaluates visual memory of multimodal agents in an interactive shopping environment with 1,000 product variants. Agents run autonomous sessions and must recall cued product images via exact URL match, with text leakage controlled.", "domainCuration": {"method": "rules-v1", "state": "auto"}, "evidence": {"reasonCodes": ["coined title prefix ending in Bench or Benchmark", "benchmark release stated in one sentence", "evaluation protocol evidence", "public artifact URL"], "snippet": "We introduce DMV-Bench (Code: https://github.com/yyyujintang/DMV-Bench), the first interactive benchmark for multimodal-agent visual memory."}, "familyId": "bmf_bac3db190dd7", "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/yyyujintang/DMV-Bench", "url": "https://github.com/yyyujintang/DMV-Bench"}, {"basis": "existing-metric-association", "linkedUrl": "https://github.com/yyyujintang/DMV-Bench", "url": "https://huggingface.co/papers/2606.27499"}, {"basis": "linked-from-report", "linkedUrl": "https://github.com/yyyujintang/DMV-Bench", "url": "https://arxiv.org/abs/2606.27499"}], "scope": "benchmark_repo", "url": "https://github.com/yyyujintang/DMV-Bench", "verification": "source-linked"}], "status": "linked"}, "heat": null, "id": "bm_dmv-bench_d16bc24a", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": "https://github.com/yyyujintang/DMV-Bench", "data": null, "hfPaper": "https://huggingface.co/papers/2606.27499", "paper": "https://arxiv.org/abs/2606.27499", "pdf": "https://arxiv.org/pdf/2606.27499", "project": null, "report": "https://arxiv.org/abs/2606.27499"}, "metrics": [], "motivation": "Research on agent memory has matured rapidly, but almost entirely on the text side: few existing benchmarks ask, in an interactive environment, when an agent genuinely needs to remember what it saw rather than what it could write down.", "name": "DMV-Bench", "oneLine": "DMV-Bench evaluates visual memory of multimodal agents in an interactive shopping environment with 1,000 product variants. Agents run autonomous sessions and must recall cued product images via exact URL match, with text leakage controlled.", "paperTitle": "DMV-Bench: Diagnosing Long-Horizon Multimodal Agents' Visual Memory with Incidental Cue Injection", "primaryDomain": "General AI", "publication": {"evidence": "Accepted to EMNLP 2026 Main Conference. 17 pages, 8 figures, 18 tables", "evidenceLevel": "author-claim", "evidenceUrl": "https://arxiv.org/abs/2606.27499", "source": "arxiv-comments", "status": "acceptance_claimed", "venue": "EMNLP 2026 Main Conference", "verifiedAt": "2026-10-14T01:46:58.058199Z"}, "publications": [], "publishers": [{"name": "DMV-Bench Team", "organizationType": "academic-lab", "role": "benchmark-publisher", "sourceUrl": "https://github.com/yyyujintang/DMV-Bench"}], "ranking": {}, "readiness": "Runnable", "recognitionConfidence": 1.0, "relation": "introduces", "releasedAt": "2026-06-25", "source": {"authors": ["Yujin Tang", "Chenming Shang", "Ruize Xu", "Nikhil Singh"], "categories": ["cs.CV", "cs.AI", "cs.CL"], "id": "2606.27499", "title": "DMV-Bench: Diagnosing Long-Horizon Multimodal Agents' Visual Memory with Incidental Cue Injection", "type": "arxiv", "url": "https://arxiv.org/abs/2606.27499"}, "sourceUpdatedAt": "2026-06-25T19:34:25Z", "topics": ["Agents", "Multimodal"], "venueAttempts": [{"decisionRaw": "Accepted to EMNLP 2026 Main Conference. 17 pages, 8 figures, 18 tables", "evidence": [{"level": "author-claim", "observedAt": "2026-10-14T01:46:58.058199Z", "rawValue": "Accepted to EMNLP 2026 Main Conference. 17 pages, 8 figures, 18 tables", "sourceType": "arxiv-comments", "sourceUrl": "https://arxiv.org/abs/2606.27499"}], "reviewStatus": "accepted", "venueName": "EMNLP 2026 Main Conference"}], "whyItMatters": "Existing agent memory benchmarks focus on text; DMV-Bench isolates visual memory needs in interactive settings, offering a controlled protocol for comparing agent architectures on pixel-based recall across varying session lengths."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
