{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:WSTCFYENRLJ6EM47ZVMAZ3Y2UU","short_pith_number":"pith:WSTCFYEN","schema_version":"1.0","canonical_sha256":"b4a622e08d8ad3e2339fcd580cef1aa52cc41e85df5141ea25b6e5c275bf8ee2","source":{"kind":"arxiv","id":"2508.03789","version":2},"attestation_state":"computed","paper":{"title":"HPSv3: Towards Wide-Spectrum Human Preference Score","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Hongsheng Li, Keqiang Sun, Xiaoshi Wu, Yuhang Ma, Yunhao Shui","submitted_at":"2025-08-05T17:17:13Z","abstract_excerpt":"Evaluating text-to-image generation models requires alignment with human perception, yet existing human-centric metrics are constrained by limited data coverage, suboptimal feature extraction, and inefficient loss functions. To address these challenges, we introduce Human Preference Score v3 (HPSv3). (1) We release HPDv3, the first wide-spectrum human preference dataset integrating 1.08M text-image pairs and 1.17M annotated pairwise comparisons from state-of-the-art generative models and low to high-quality real-world images. (2) We introduce a VLM-based preference model trained using an uncer"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2508.03789","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2025-08-05T17:17:13Z","cross_cats_sorted":[],"title_canon_sha256":"ec184bbfaea4a70a90021a52faac5d5b58d09724e9a5646ca12fe9ab04a8769d","abstract_canon_sha256":"fddea1cc1ce041f450495f745eb31b6ca805954cff31f2bb65ed5a24bd392e10"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:57:38.428913Z","signature_b64":"l7fHTat41szPsXMZYB8Ay2pQCGLnAbjzYNbvbvBEbuIWXv6uELu88T6eh3MvLALyE/6TKdk0MLMJsD7Vir2BAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b4a622e08d8ad3e2339fcd580cef1aa52cc41e85df5141ea25b6e5c275bf8ee2","last_reissued_at":"2026-07-05T11:57:38.428447Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:57:38.428447Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"HPSv3: Towards Wide-Spectrum Human Preference Score","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Hongsheng Li, Keqiang Sun, Xiaoshi Wu, Yuhang Ma, Yunhao Shui","submitted_at":"2025-08-05T17:17:13Z","abstract_excerpt":"Evaluating text-to-image generation models requires alignment with human perception, yet existing human-centric metrics are constrained by limited data coverage, suboptimal feature extraction, and inefficient loss functions. To address these challenges, we introduce Human Preference Score v3 (HPSv3). (1) We release HPDv3, the first wide-spectrum human preference dataset integrating 1.08M text-image pairs and 1.17M annotated pairwise comparisons from state-of-the-art generative models and low to high-quality real-world images. (2) We introduce a VLM-based preference model trained using an uncer"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2508.03789","kind":"arxiv","version":2},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2508.03789/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2508.03789","created_at":"2026-07-05T11:57:38.428507+00:00"},{"alias_kind":"arxiv_version","alias_value":"2508.03789v2","created_at":"2026-07-05T11:57:38.428507+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2508.03789","created_at":"2026-07-05T11:57:38.428507+00:00"},{"alias_kind":"pith_short_12","alias_value":"WSTCFYENRLJ6","created_at":"2026-07-05T11:57:38.428507+00:00"},{"alias_kind":"pith_short_16","alias_value":"WSTCFYENRLJ6EM47","created_at":"2026-07-05T11:57:38.428507+00:00"},{"alias_kind":"pith_short_8","alias_value":"WSTCFYEN","created_at":"2026-07-05T11:57:38.428507+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":19,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.26930","citing_title":"PortraitGen: Exemplar-Driven GRPO with Dual-Reward Guidance for Photorealistic Portrait Generation","ref_index":28,"is_internal_anchor":false},{"citing_arxiv_id":"2607.02220","citing_title":"DetailAnywhere: Fashion Detail Generation via Cross-Modal Feature Alignment Distillation","ref_index":70,"is_internal_anchor":false},{"citing_arxiv_id":"2606.09076","citing_title":"Z-Reward: Beyond Scalar Rewards by Internalizing Reasoning into Score Distributions","ref_index":36,"is_internal_anchor":false},{"citing_arxiv_id":"2606.03216","citing_title":"Follow-Your-Preference++: Rethinking Preference Alignment for Image Inpainting","ref_index":44,"is_internal_anchor":false},{"citing_arxiv_id":"2606.01985","citing_title":"MT-EditFlow: Reinforcement Learning for Multi-Turn Image Editing with Flow Matching","ref_index":33,"is_internal_anchor":false},{"citing_arxiv_id":"2606.32020","citing_title":"Cross-Space Distillation: Teaching One-Step Students with Modern Diffusion Teachers","ref_index":28,"is_internal_anchor":false},{"citing_arxiv_id":"2605.19320","citing_title":"TextAlign: Preference Alignment for Text Rendering with Hierarchical Rewards","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2605.25759","citing_title":"Towards Anatomically Plausible Human Image Generation via Synthetic Localized Preferences","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2605.25874","citing_title":"WBench: A Comprehensive Multi-turn Benchmark for Interactive Video World Model Evaluation","ref_index":65,"is_internal_anchor":false},{"citing_arxiv_id":"2510.21583","citing_title":"Principled RL for Flow Matching Emerges from the Chunk-level Policy Optimization","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2604.27505","citing_title":"Leveraging Verifier-Based Reinforcement Learning in Image Editing","ref_index":39,"is_internal_anchor":false},{"citing_arxiv_id":"2605.19320","citing_title":"TextAlign: Preference Alignment for Text Rendering with Hierarchical Rewards","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2605.16842","citing_title":"Sketch Then Paint: Hierarchical Reinforcement Learning for Diffusion Multi-Modal Large Language Models","ref_index":35,"is_internal_anchor":false},{"citing_arxiv_id":"2512.01236","citing_title":"PSR: Scaling Multi-Subject Personalized Image Generation with Pairwise Subject-Consistency Rewards","ref_index":20,"is_internal_anchor":false},{"citing_arxiv_id":"2605.14278","citing_title":"KVPO: ODE-Native GRPO for Autoregressive Video Alignment via KV Semantic Exploration","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2604.27505","citing_title":"Leveraging Verifier-Based Reinforcement Learning in Image Editing","ref_index":39,"is_internal_anchor":false},{"citing_arxiv_id":"2604.24953","citing_title":"ViPO: Visual Preference Optimization at Scale","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2604.18258","citing_title":"Long-Text-to-Image Generation via Compositional Prompt Decomposition","ref_index":50,"is_internal_anchor":false},{"citing_arxiv_id":"2604.09850","citing_title":"Training-Free Object-Background Compositional T2I via Dynamic Spatial Guidance and Multi-Path Pruning","ref_index":31,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/WSTCFYENRLJ6EM47ZVMAZ3Y2UU","json":"https://pith.science/pith/WSTCFYENRLJ6EM47ZVMAZ3Y2UU.json","graph_json":"https://pith.science/api/pith-number/WSTCFYENRLJ6EM47ZVMAZ3Y2UU/graph.json","events_json":"https://pith.science/api/pith-number/WSTCFYENRLJ6EM47ZVMAZ3Y2UU/events.json","paper":"https://pith.science/paper/WSTCFYEN"},"agent_actions":{"view_html":"https://pith.science/pith/WSTCFYENRLJ6EM47ZVMAZ3Y2UU","download_json":"https://pith.science/pith/WSTCFYENRLJ6EM47ZVMAZ3Y2UU.json","view_paper":"https://pith.science/paper/WSTCFYEN","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2508.03789&json=true","fetch_graph":"https://pith.science/api/pith-number/WSTCFYENRLJ6EM47ZVMAZ3Y2UU/graph.json","fetch_events":"https://pith.science/api/pith-number/WSTCFYENRLJ6EM47ZVMAZ3Y2UU/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/WSTCFYENRLJ6EM47ZVMAZ3Y2UU/action/timestamp_anchor","attest_storage":"https://pith.science/pith/WSTCFYENRLJ6EM47ZVMAZ3Y2UU/action/storage_attestation","attest_author":"https://pith.science/pith/WSTCFYENRLJ6EM47ZVMAZ3Y2UU/action/author_attestation","sign_citation":"https://pith.science/pith/WSTCFYENRLJ6EM47ZVMAZ3Y2UU/action/citation_signature","submit_replication":"https://pith.science/pith/WSTCFYENRLJ6EM47ZVMAZ3Y2UU/action/replication_record"}},"created_at":"2026-07-05T11:57:38.428507+00:00","updated_at":"2026-07-05T11:57:38.428507+00:00"}