{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2606.19338", "kind": "paper", "url": "https://arxiv.org/abs/2606.19338"}, {"id": "gh:internlm/rngbench", "kind": "repo", "url": "https://github.com/InternLM/RNGBench"}], "categories": ["General AI", "Multimodal"], "description": {"en": "RNG-Bench evaluates multimodal LLMs in controllable non-Markov games, requiring reconstruction of past observations and acting on them, with two games: Matching Pairs and 3D Maze."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2606.19338", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2606.19338", "title": "RNG-Bench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "key": "claire-radar:2606.19338", "modality": null, "name": "RNG-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_rng-bench_1fb2eb04", "origin_source": "arxiv", "recognition_confidence": "1.0", "record_sha256": "458f2224b208b9e77af78b26cea21052035161d2a1602994b2d7afffd222f781", "relation": "aggregates", "review_state": "unreviewed", "source_url": "https://arxiv.org/abs/2606.19338"}, "publisher": null, "released": "2026-06-17", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2606.19338", "source_url": "https://arxiv.org/abs/2606.19338"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2606-19338", "source": "claire_radar", "source_benchmark_id": "2606.19338", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Unknown", "applicationDomains": ["General AI"], "area": "Multimodal", "attention": {"asOf": "2026-10-05", "attemptedAt": "2026-10-04T15:31:48.030916Z", "githubRepo": "https://github.com/InternLM/RNGBench", "githubScope": "benchmark_repo", "githubStars": 42, "hfDailySubmittedAt": null, "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": 51, "hfPaperUrl": "https://huggingface.co/papers/2606.19338", "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": "2026-10-02", "state": "stale"}}, "status": "fresh"}, "capabilities": [], "confidence": "High", "construction": "Transform Existing", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-08-20T15:22:10.044731Z", "inputHash": "47e264b08d3d68b5084f6a5e801e641cd2e92b8b8cc6c9ac7434569b9803aade", "model": "deepseek-v4-flash"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "RNG-Bench evaluates multimodal LLMs in controllable non-Markov games, requiring reconstruction of past observations and acting on them, with two games: Matching Pairs and 3D Maze.", "domainCuration": {"method": "rules-v1", "state": "auto"}, "evidence": {"reasonCodes": ["exact named benchmark artifact released in abstract", "benchmark release stated in one sentence", "evaluation protocol evidence"], "snippet": "We introduce RNG-Bench (Reconstructive Non-Markov Games), a benchmark suite designed to isolate a base model's ability to reconstruct past observations and act on them during multi-step interaction."}, "familyId": "bmf_4e8d6f247af1", "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/InternLM/RNGBench", "url": "https://github.com/InternLM/RNGBench"}, {"basis": "existing-metric-association", "linkedUrl": "https://github.com/InternLM/RNGBench", "url": "https://huggingface.co/papers/2606.19338"}], "scope": "benchmark_repo", "url": "https://github.com/InternLM/RNGBench", "verification": "source-linked"}], "status": "linked"}, "heat": null, "id": "bm_rng-bench_1fb2eb04", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": "https://github.com/InternLM/RNGBench", "data": null, "hfPaper": "https://huggingface.co/papers/2606.19338", "paper": "https://arxiv.org/abs/2606.19338", "pdf": "https://arxiv.org/pdf/2606.19338", "project": null, "report": "https://arxiv.org/abs/2606.19338"}, "metrics": [], "motivation": "Deploying multimodal foundation models as closed-loop policies increasingly requires conditioning actions on observations that are no longer visible.", "name": "RNG-Bench", "oneLine": "RNG-Bench evaluates multimodal LLMs in controllable non-Markov games, requiring reconstruction of past observations and acting on them, with two games: Matching Pairs and 3D Maze.", "paperTitle": "Beyond the Current Observation: Evaluating Multimodal Large Language Models in Controllable Non-Markov Games", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2606.19338", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-10-14T01:46:58.058199Z"}, "publications": [], "publishers": [{"name": "InternLM", "organizationType": "benchmark-organization", "role": "benchmark-publisher", "sourceUrl": "https://github.com/InternLM/RNGBench"}], "ranking": {}, "readiness": "Runnable", "recognitionConfidence": 1.0, "relation": "aggregates", "releasedAt": "2026-06-17", "source": {"authors": ["Shengyuan Ding", "Xilin Wei", "Xinyu Fang", "Haodong Duan", "Dahua Lin", "Jiaqi Wang", "Yuhang Zang"], "categories": ["cs.CV"], "id": "2606.19338", "title": "Beyond the Current Observation: Evaluating Multimodal Large Language Models in Controllable Non-Markov Games", "type": "arxiv", "url": "https://arxiv.org/abs/2606.19338"}, "sourceUpdatedAt": "2026-06-17T17:59:34Z", "topics": ["Multimodal"], "venueAttempts": [], "whyItMatters": "Addresses the gap in evaluating models' abilities to remember and act on hidden state, which is critical for real-world deployments where observations are partial."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
