{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2608.03708", "kind": "paper", "url": "https://arxiv.org/abs/2608.03708"}, {"id": "gh:i2-multimedia-lab/multicompose", "kind": "repo", "url": "https://github.com/I2-Multimedia-Lab/MultiCompose"}], "categories": ["General AI", "Vision & 3D", "cs.CV"], "description": {"en": "MSP-Bench evaluates multi-subject personalized text-to-image generation. It scores identity fidelity, attribute binding accuracy, and attribute misalignment through a dual-pathway protocol."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2608.03708", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2608.03708", "title": "MSP-Bench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "key": "claire-radar:2608.03708", "modality": null, "name": "MSP-Bench", "openness": {"code_license": null, "data_license": null, "evidence": [], "status": "unknown"}, "provenance": {"admission_policy_version": "2026-09-09.1", "confidence": "Medium", "crawl_bundle": "claire_radar_2026-10-14", "crawled_at": "2026-09-25T04:40:22Z", "data_status": "primary-source-indexed", "display_eligible": "true", "origin_record_id": "bm_msp-bench_4b910289", "origin_source": "arxiv", "recognition_confidence": "0.55", "record_sha256": "46e063dc77fba913ccc851847ea40e3fd77e2a3209091f81bfb4d414992a23c1", "relation": "evaluates_only", "review_model": "claude-code-default", "review_state": "ai-reviewed", "reviewed_at": "2026-09-24T20:55:29.497960Z", "source_url": "https://arxiv.org/abs/2608.03708"}, "publisher": null, "released": "2026-08-04", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2608.03708", "source_url": "https://arxiv.org/abs/2608.03708"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2608-03708", "source": "claire_radar", "source_benchmark_id": "2608.03708", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Unknown", "applicationDomains": ["General AI"], "area": "Vision & 3D", "attentionForecast": {"confidence": "Low", "horizon": "7d", "reason": "Niche personalization topic, and the benchmark ships inside a method repository.", "score": 30}, "capabilities": [], "confidence": "Medium", "construction": "Unknown", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-09-24T20:55:29.497960Z", "inputHash": "e7b83216d20201f36254b2313715caa6f32efb762501399a88e430aeb35e94ac", "model": "claude-code-default"}, "curation": {"admissionPolicyVersion": "2026-09-09.1", "canonicalNameEvidence": "We further introduce MSP-Bench, a benchmark", "canonicalNameSource": "abstract", "decisionReason": "The benchmark is named, has explicit metrics (ID, BIND, MIS), uses a defined protocol, and comes with a public code link.", "evaluationEvidence": {"reuse": "Code is available at https://github.com/I2-Multimedia-Lab/MultiCompose", "scoring": "identity fidelity (ID), attribute binding accuracy (BIND), and attribute misalignment (MIS) through a dual-pathway protocol", "task": "jointly evaluates identity fidelity (ID), attribute binding accuracy (BIND), and attribute misalignment (MIS)"}, "model": "claude-code-default", "reviewedAt": "2026-09-24T20:55:29.497960Z", "state": "ai-reviewed"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "MSP-Bench evaluates multi-subject personalized text-to-image generation. It scores identity fidelity, attribute binding accuracy, and attribute misalignment through a dual-pathway protocol.", "displayEligible": true, "domainCuration": {"method": "rules-v1", "state": "auto"}, "evaluationMode": "public_reusable", "evidence": {"reasonCodes": ["exact named benchmark artifact released in abstract", "public artifact URL", "named entity is a model, method, framework, or dataset used in evaluation"], "snippet": "We further introduce MSP-Bench, a benchmark that jointly evaluates identity fidelity (ID), attribute binding accuracy (BIND), and attribute misalignment (MIS) through a dual-pathway protocol."}, "familyId": "bmf_8c7e23649c9b", "firstSeenAt": "2026-08-19", "heat": null, "id": "bm_msp-bench_4b910289", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": "https://github.com/I2-Multimedia-Lab/MultiCompose", "data": null, "paper": "https://arxiv.org/abs/2608.03708", "pdf": "https://arxiv.org/pdf/2608.03708", "project": null, "report": "https://arxiv.org/abs/2608.03708"}, "metrics": [], "motivation": "Text-to-image diffusion models enable personalization of specific visual concepts from a small number of reference images.", "name": "MSP-Bench", "oneLine": "MSP-Bench evaluates multi-subject personalized text-to-image generation. It scores identity fidelity, attribute binding accuracy, and attribute misalignment through a dual-pathway protocol.", "paperTitle": "MultiCompose: Multi-Concept Personalized Composition with Per-Subject Attribute Binding", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2608.03708", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-08-19T11:05:13.395391Z"}, "publications": [], "publishers": [{"name": "I2-Multimedia-Lab", "organizationType": "academic-lab", "role": "benchmark-publisher", "sourceUrl": "https://github.com/I2-Multimedia-Lab/MultiCompose"}], "readiness": "Runnable", "recognitionConfidence": 0.55, "relation": "evaluates_only", "releasedAt": "2026-08-04", "reviewLifecycle": {"state": "pending"}, "source": {"authors": ["Ruirui Zhang", "Zhengkai Zhao", "Pan Gao"], "categories": ["cs.CV"], "id": "2608.03708", "title": "MultiCompose: Multi-Concept Personalized Composition with Per-Subject Attribute Binding", "type": "arxiv", "url": "https://arxiv.org/abs/2608.03708"}, "sourceUpdatedAt": "2026-08-04T14:13:18Z", "topics": ["cs.CV"], "venueAttempts": [], "whyItMatters": "It targets identity loss and attribute misbinding when several personalized subjects share one image, failure modes the authors say conventional metrics miss."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
