{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2606.01600", "kind": "paper", "url": "https://arxiv.org/abs/2606.01600"}, {"kind": "website", "url": "https://huiqiongli.github.io/RoboTrustBench/"}], "authors": ["Huiqiong Li", "Jiayu Wang", "Zhiting Mei", "Anirudha Majumdar", "Jingjing Chen", "Bin Zhu"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2606.01600", "source_url": "https://arxiv.org/abs/2606.01600"}, "categories": ["Robot manipulation", "Robotics", "Robotics & Embodied AI"], "description": {"en": "RoboTrustBench evaluates the trustworthiness of video world models for robotic manipulation across four scenarios: Normal, Constraint-Sensitive, Counterfactual, and Adversarial. It contains 1,207 expert-validated instruction-image pairs from DROID episodes and a six-dimensional evaluation protocol with 13 fine-grained criteria."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2606.01600", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2606.01600", "title": "RoboTrustBench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "key": "claire-radar:2606.01600", "modality": null, "name": "RoboTrustBench", "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_robotrustbench_454946c5", "origin_source": "arxiv", "recognition_confidence": "1.0", "record_sha256": "fd482039a529cbe9c8ac656241a6f5aa5fee3807f336569567e93da79a3432a0", "relation": "introduces", "review_state": "unreviewed", "source_url": "https://arxiv.org/abs/2606.01600"}, "publisher": null, "released": "2026-06-01", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2606.01600", "source_url": "https://arxiv.org/abs/2606.01600"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2606-01600", "source": "claire_radar", "source_benchmark_id": "2606.01600", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Unknown", "applicationDomains": ["Robotics & Embodied AI"], "area": "Robotics & Embodied AI", "attention": {"asOf": "2026-10-05", "attemptedAt": "2026-10-04T15:31:48.030916Z", "githubRepo": null, "githubScope": null, "githubStars": null, "hfDailySubmittedAt": null, "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": 0, "hfPaperUrl": "https://huggingface.co/papers/2606.01600", "observedAt": "2026-10-04T03:02:42.754940Z", "signalStatus": {"githubStars": {"state": "not_applicable"}, "hfDatasetDownloads": {"state": "not_applicable"}, "hfDatasetLikes": {"state": "not_applicable"}, "hfPaperUpvotes": {"checkedAt": "2026-10-04T15:31:48.030916Z", "lastSuccessfulDate": "2026-09-30", "state": "stale"}}, "status": "stale"}, "capabilities": ["Robot manipulation"], "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": "15548aa1d2295b999b236d057b81f446c1a22b8293446670ef77579a87a37f08", "model": "deepseek-v4-flash"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "RoboTrustBench evaluates the trustworthiness of video world models for robotic manipulation across four scenarios: Normal, Constraint-Sensitive, Counterfactual, and Adversarial. It contains 1,207 expert-validated instruction-image pairs from DROID episodes and a six-dimensional evaluation protocol with 13 fine-grained criteria.", "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 RoboTrustBench, a benchmark for evaluating the trustworthiness of video world models under four scenarios: Normal, Constraint-Sensitive, Counterfactual, and Adversarial."}, "familyId": "bmf_2309dd08a194", "firstSeenAt": "2026-08-19", "githubIndex": {"candidateCount": 0, "indexedAt": "2026-09-17T04:40:34.111665+00:00", "repositories": [], "status": "not-found"}, "heat": null, "id": "bm_robotrustbench_454946c5", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": ["Robotics"], "links": {"code": null, "data": null, "hfPaper": "https://huggingface.co/papers/2606.01600", "paper": "https://arxiv.org/abs/2606.01600", "pdf": "https://arxiv.org/pdf/2606.01600", "project": "https://huiqiongli.github.io/RoboTrustBench/", "report": "https://arxiv.org/abs/2606.01600"}, "metrics": [], "motivation": "Video world models are increasingly used in robotic manipulation, yet existing benchmarks mostly evaluate them under valid, feasible, and safe instructions.", "name": "RoboTrustBench", "oneLine": "RoboTrustBench evaluates the trustworthiness of video world models for robotic manipulation across four scenarios: Normal, Constraint-Sensitive, Counterfactual, and Adversarial. It contains 1,207 expert-validated instruction-image pairs from DROID episodes and a six-dimensional evaluation protocol with 13 fine-grained criteria.", "paperTitle": "RoboTrustBench: Benchmarking the Trustworthiness of Video World Models for Robotic Manipulation", "primaryDomain": "Robotics & Embodied AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2606.01600", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-10-14T01:46:58.058199Z"}, "publications": [], "publishers": [{"name": "RoboTrustBench Team", "organizationType": "academic-lab", "role": "benchmark-publisher", "sourceUrl": "https://huiqiongli.github.io/RoboTrustBench/"}], "ranking": {}, "readiness": "Inspectable", "recognitionConfidence": 1.0, "relation": "introduces", "releasedAt": "2026-06-01", "source": {"authors": ["Huiqiong Li", "Jiayu Wang", "Zhiting Mei", "Anirudha Majumdar", "Jingjing Chen", "Bin Zhu"], "categories": ["cs.CV", "cs.CL", "cs.RO"], "id": "2606.01600", "title": "RoboTrustBench: Benchmarking the Trustworthiness of Video World Models for Robotic Manipulation", "type": "arxiv", "url": "https://arxiv.org/abs/2606.01600"}, "sourceUpdatedAt": "2026-06-01T02:56:09Z", "topics": ["Robotics"], "venueAttempts": [], "whyItMatters": "Existing benchmarks for video world models largely overlook trustworthiness aspects such as constraint reasoning, counterfactual grounding, and safety. RoboTrustBench provides a structured evaluation to assess these capabilities, offering practical guidance for selecting and improving models for safe robotic manipulation."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
