{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2606.03577", "kind": "paper", "url": "https://arxiv.org/abs/2606.03577"}, {"id": "gh:aim-uofa/reasonmatch", "kind": "repo", "url": "https://github.com/aim-uofa/ReasonMatch"}, {"kind": "website", "url": "https://aim-uofa.github.io/reasonmatch/"}], "authors": ["Hao Zhong", "Muzhi Zhu", "Shenyan Zeng", "Anzhou Li", "Cong Chen", "Hua Geng", "Duochao Shi", "Wentao Ye", "Tao Lin", "Hao Chen", "Chunhua Shen"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2606.03577", "source_url": "https://arxiv.org/abs/2606.03577"}, "categories": ["General AI", "Reasoning", "Vision & 3D"], "description": {"en": "ReasonMatch-Bench evaluates wide-baseline matching and spatial reasoning in MLLMs, with benchmarks stratified by viewpoint displacement and matching granularity across indoor, outdoor, and object-centric scenarios."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2606.03577", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2606.03577", "title": "ReasonMatch-Bench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "key": "claire-radar:2606.03577", "modality": null, "name": "ReasonMatch-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_reasonmatch-bench_7b565eab", "origin_source": "arxiv", "recognition_confidence": "1.0", "record_sha256": "9cd4a709e2fbdd2c9523db3eedab5617876bccabb511942dcfff2bcae42e8bd6", "relation": "introduces", "review_state": "unreviewed", "source_url": "https://arxiv.org/abs/2606.03577"}, "publisher": null, "released": "2026-06-02", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2606.03577", "source_url": "https://arxiv.org/abs/2606.03577"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2606-03577", "source": "claire_radar", "source_benchmark_id": "2606.03577", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Unknown", "applicationDomains": ["General AI"], "area": "Vision & 3D", "attention": {"asOf": "2026-10-05", "attemptedAt": "2026-10-04T15:31:48.030916Z", "githubRepo": "https://github.com/aim-uofa/ReasonMatch", "githubScope": "benchmark_repo", "githubStars": 19, "hfDailySubmittedAt": null, "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": 16, "hfPaperUrl": "https://huggingface.co/papers/2606.03577", "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": "2026-09-25", "state": "stale"}}, "status": "fresh"}, "capabilities": ["Reasoning"], "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": "2361e0e31b29ed4c7cfa0919493ebca44904464faa7f0f7a2e88f51ddf52b4e2", "model": "deepseek-v4-flash"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "ReasonMatch-Bench evaluates wide-baseline matching and spatial reasoning in MLLMs, with benchmarks stratified by viewpoint displacement and matching granularity across indoor, outdoor, and object-centric scenarios.", "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": "We introduce ReasonMatch-Bench, a benchmark stratified by viewpoint displacement and matching granularity across indoor, outdoor, and object-centric scenarios, and show that current MLLMs still struggle with fine-grained wide-baseline correspondence: on a difficult 90-sample subset, human annotators achieve 84.0 F1, while the best existing baseline reaches 37.2."}, "familyId": "bmf_726b0c5520b5", "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/aim-uofa/ReasonMatch", "url": "https://github.com/aim-uofa/ReasonMatch"}, {"basis": "existing-metric-association", "linkedUrl": "https://github.com/aim-uofa/ReasonMatch", "url": "https://huggingface.co/papers/2606.03577"}, {"basis": "linked-from-project", "linkedUrl": "https://github.com/aim-uofa/ReasonMatch", "url": "https://aim-uofa.github.io/reasonmatch/"}, {"basis": "linked-from-report", "linkedUrl": "https://github.com/aim-uofa/ReasonMatch", "url": "https://arxiv.org/abs/2606.03577"}], "scope": "benchmark_repo", "url": "https://github.com/aim-uofa/ReasonMatch", "verification": "source-linked"}], "status": "linked"}, "heat": null, "id": "bm_reasonmatch-bench_7b565eab", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": "https://github.com/aim-uofa/ReasonMatch", "data": null, "hfPaper": "https://huggingface.co/papers/2606.03577", "paper": "https://arxiv.org/abs/2606.03577", "pdf": "https://arxiv.org/pdf/2606.03577", "project": "https://aim-uofa.github.io/reasonmatch/", "report": "https://arxiv.org/abs/2606.03577"}, "metrics": [], "motivation": "Wide-baseline matching (WBM) requires integrating geometric understanding, viewpoint changes, fine-grained perception, and occlusion reasoning, making it a challenging testbed for spatial reasoning in multimodal large language models (MLLMs) deployed in physical environments.", "name": "ReasonMatch-Bench", "oneLine": "ReasonMatch-Bench evaluates wide-baseline matching and spatial reasoning in MLLMs, with benchmarks stratified by viewpoint displacement and matching granularity across indoor, outdoor, and object-centric scenarios.", "paperTitle": "Eliciting Complex Spatial Reasoning in MLLMs through Wide-Baseline Matching", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2606.03577", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-10-14T01:46:58.058199Z"}, "publications": [], "publishers": [{"name": "AIM-UoFA", "organizationType": "academic-lab", "role": "benchmark-publisher", "sourceUrl": "https://github.com/aim-uofa/ReasonMatch"}], "ranking": {}, "readiness": "Runnable", "recognitionConfidence": 1.0, "relation": "introduces", "releasedAt": "2026-06-02", "source": {"authors": ["Hao Zhong", "Muzhi Zhu", "Shenyan Zeng", "Anzhou Li", "Cong Chen", "Hua Geng", "Duochao Shi", "Wentao Ye", "Tao Lin", "Hao Chen", "Chunhua Shen"], "categories": ["cs.CV"], "id": "2606.03577", "title": "Eliciting Complex Spatial Reasoning in MLLMs through Wide-Baseline Matching", "type": "arxiv", "url": "https://arxiv.org/abs/2606.03577"}, "sourceUpdatedAt": "2026-06-02T12:46:34Z", "topics": ["Reasoning"], "venueAttempts": [], "whyItMatters": "Addresses the lack of systematic evaluation for spatial reasoning in MLLMs, offering a public benchmark and reproducible training recipe to advance visual correspondence understanding."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
