{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2609.07047", "kind": "paper", "url": "https://arxiv.org/abs/2609.07047"}, {"id": "gh:collab-gen/memobench", "kind": "repo", "url": "https://github.com/Collab-Gen/MEMOBench"}], "authors": ["Haiyang Sun", "Haoxiao Wang", "Junming Chen", "Weicheng Fang", "Zihao Su", "Jingkun Yi", "Wenyou Yi", "Hao Chen", "Zhou Zhao"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2609.07047", "source_url": "https://arxiv.org/abs/2609.07047"}, "categories": ["Robot manipulation", "Robotics", "Robotics & Embodied AI"], "description": {"en": "Evaluates robotic manipulation policies on memory fidelity via 30 history-dependent tasks with annotated checkpoints measuring storage, update, and compression operations."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2609.07047", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2609.07047", "title": "MEMOBench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "key": "claire-radar:2609.07047", "modality": null, "name": "MEMOBench", "openness": {"code_license": null, "data_license": null, "evidence": [], "status": "unknown"}, "provenance": {"admission_policy_version": "2026-09-09.1", "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_memobench_dd60fcde", "origin_source": "arxiv", "recognition_confidence": "1.0", "record_sha256": "d779f08862baf98d62155ad9a6e1b9d6ee2f3df53c09bc37114af39574610208", "relation": "introduces", "review_model": "deepseek-v4-pro", "review_state": "ai-reviewed", "reviewed_at": "2026-09-09T06:27:57.312318Z", "source_url": "https://arxiv.org/abs/2609.07047"}, "publisher": null, "released": "2026-09-07", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2609.07047", "source_url": "https://arxiv.org/abs/2609.07047"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2609-07047", "source": "claire_radar", "source_benchmark_id": "2609.07047", "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": "https://github.com/Collab-Gen/MEMOBench", "githubScope": "benchmark_repo", "githubStars": 3, "hfDailySubmittedAt": null, "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": 1, "hfPaperUrl": "https://huggingface.co/papers/2609.07047", "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": {"observedAt": "2026-10-04T15:31:48.030916Z", "state": "fresh"}}, "status": "fresh"}, "attentionForecast": {"confidence": "Medium", "horizon": "7d", "reason": "Robotic memory is an active research area and the benchmark provides unique process-level metrics with accessible code and data.", "score": 55}, "capabilities": ["Robot manipulation"], "confidence": "High", "construction": "Interactive Environment", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-09-09T06:27:57.312318Z", "inputHash": "dcc8a6da035d961e470e125ba06b4bbbf4e84f3e1518db86a82257d9df403be5", "model": "deepseek-v4-pro"}, "curation": {"admissionPolicyVersion": "2026-09-09.1", "canonicalNameEvidence": "We present \\textbf{MEMOBench}, a benchmark for process level memory evaluation in robotic manipulation.", "canonicalNameSource": "abstract", "decisionReason": "Clear memory operation definitions with quantitative rates, public code and data, and executable checkpoints support independent evaluation.", "evaluationEvidence": {"reuse": "The project page is available at https://github.com/Collab-Gen/MEMOBench and datasets on HuggingFace.", "scoring": "Each checkpoint pairs coarse to fine language with a simulator predicate and labels one memory operation: Storage, Update, or Compression, defining Memory Storage Rate, Memory Update Rate, and Memory Compression Rate.", "task": "MEMOBench includes 30 history dependent tasks, 1,500 expert demonstrations, and 4,200 executable checkpoint instances from 84 templates."}, "model": "deepseek-v4-pro", "reviewedAt": "2026-09-09T06:27:57.312318Z", "state": "ai-reviewed"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "Evaluates robotic manipulation policies on memory fidelity via 30 history-dependent tasks with annotated checkpoints measuring storage, update, and compression operations.", "displayEligible": true, "domainCuration": {"method": "rules-v1", "state": "auto"}, "evaluationMode": "score_submission", "evidence": {"reasonCodes": ["coined title prefix ending in Bench or Benchmark", "benchmark release stated in one sentence", "evaluation protocol evidence", "public artifact URL"], "snippet": "We present \\textbf{MEMOBench}, a benchmark for process level memory evaluation in robotic manipulation."}, "familyId": "bmf_e8d56608be68", "firstSeenAt": "2026-09-09", "heat": null, "id": "bm_memobench_dd60fcde", "indexedAt": "2026-09-09T06:18:29.794093Z", "industrySectors": ["Robotics"], "links": {"code": "https://github.com/Collab-Gen/MEMOBench", "data": null, "paper": "https://arxiv.org/abs/2609.07047", "pdf": "https://arxiv.org/pdf/2609.07047", "project": null, "report": "https://arxiv.org/abs/2609.07047"}, "metrics": [], "motivation": "Robotic manipulation often requires acting on information that is no longer visible, yet Vision-Language-Action policies are usually evaluated when the current observation largely determines the next action.", "name": "MEMOBench", "oneLine": "Evaluates robotic manipulation policies on memory fidelity via 30 history-dependent tasks with annotated checkpoints measuring storage, update, and compression operations.", "paperTitle": "MEMOBench: A Process Level Memory Benchmark for Robotic Manipulation", "primaryDomain": "Robotics & Embodied AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2609.07047", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-09-09T06:18:29.794093Z"}, "publications": [], "ranking": {"30d": {"confidence": "High", "coverage": 0.85, "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": 30, "historyDate": "2026-09-05", "rank": null, "score": null, "windowDays": 30}, "level": {"components": {"githubStars": {"normalization": "log-max", "normalized": 0.20440210918121815, "value": 3}, "hfDatasetDownloads": {"normalization": "log-max", "normalized": null, "value": null}, "hfPaperUpvotes": {"normalization": "log-max", "normalized": 0.12906235291407148, "value": 1}}, "confidence": "High", "coverage": 0.85, "rank": 122, "score": 23}, "method": "attention-ranking-v13", "rank": 122, "score": 23}, "90d": {"confidence": "Medium", "coverage": 0.7, "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": 0.18504537701418716, "value": 3}, "hfDatasetDownloads": {"normalization": "log-max", "normalized": null, "value": null}, "hfPaperUpvotes": {"normalization": "log-max", "normalized": 0.12247537038455211, "value": 1}}, "confidence": "Medium", "coverage": 0.7, "rank": 350, "score": 27}, "method": "attention-ranking-v13", "rank": 350, "score": 27}}, "readiness": "Runnable", "recognitionConfidence": 1.0, "relation": "introduces", "releasedAt": "2026-09-07", "source": {"authors": ["Haiyang Sun", "Haoxiao Wang", "Junming Chen", "Weicheng Fang", "Zihao Su", "Jingkun Yi", "Wenyou Yi", "Hao Chen", "Zhou Zhao"], "categories": ["cs.RO", "cs.AI"], "id": "2609.07047", "title": "MEMOBench: A Process Level Memory Benchmark for Robotic Manipulation", "type": "arxiv", "url": "https://arxiv.org/abs/2609.07047"}, "sourceUpdatedAt": "2026-09-07T05:06:20Z", "topics": ["Robotics"], "venueAttempts": [], "whyItMatters": "Provides process-level memory diagnostics beyond final success, enabling fine-grained analysis of forgetting in VLA policies."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
