{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:AWCZRZQ4QCF2K4O2MUR2KYFJ4P","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"61deb671ab72436b4fe9934b639cef553965b827b86a5f6c7be13ed84a2aa7c3","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2025-03-07T08:36:05Z","title_canon_sha256":"2aedac95cbbedd1322221f81bf6fbba2b5fd14fda6c12942a05f73b48664f4f8"},"schema_version":"1.0","source":{"id":"2503.05236","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2503.05236","created_at":"2026-05-18T03:15:18Z"},{"alias_kind":"arxiv_version","alias_value":"2503.05236v2","created_at":"2026-05-18T03:15:18Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2503.05236","created_at":"2026-05-18T03:15:18Z"},{"alias_kind":"pith_short_12","alias_value":"AWCZRZQ4QCF2","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"AWCZRZQ4QCF2K4O2","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"AWCZRZQ4","created_at":"2026-05-18T12:33:37Z"}],"graph_snapshots":[{"event_id":"sha256:22335ef39eb445a23219cb2a0e07a34d98ce9ae3d70a497d30e43d435ea1c108","target":"graph","created_at":"2026-05-18T03:15:18Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"jointly learning to assess diverse visual tasks yields substantial mutual benefits... achieving consistent improvements across each domain."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"The large-scale human preference dataset accurately represents human judgments across tasks and the two-stage filtering strategy produces high-quality, unbiased preference pairs without introducing selection artifacts."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"UnifiedReward is the first unified reward model that jointly assesses multimodal understanding and generation to provide better preference signals for aligning vision models via DPO."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"A single reward model trained jointly on image and video tasks improves preference alignment for both understanding and generation."}],"snapshot_sha256":"a04782d1d631e82f96d7da719f6f7958ce15ad318815067e1e7a0c81fc365900"},"formal_canon":{"evidence_count":2,"snapshot_sha256":"bc8e6693c3ba327a9134bd49e0e50656b6f9543d25273089ee312b7696bb487f"},"paper":{"abstract_excerpt":"Recent advances in human preference alignment have significantly improved multimodal generation and understanding. A key approach is to train reward models that provide supervision signals for preference optimization. However, existing reward models are often task-specific, limiting their adaptability across diverse visual applications. We also argue that a reward model that jointly learning to assess multiple vision tasks may foster a synergistic effect, where improved image understanding enhances image generation assessment, and refined image evaluation benefits video assessment through bett","authors_text":"Cheng Jin, Hao Li, Jiaqi Wang, Yibin Wang, Yuhang Zang","cross_cats":[],"headline":"A single reward model trained jointly on image and video tasks improves preference alignment for both understanding and generation.","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2025-03-07T08:36:05Z","title":"Unified Reward Model for Multimodal Understanding and Generation"},"references":{"count":64,"internal_anchors":21,"resolved_work":64,"sample":[{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":1,"title":"Diffusion model alignment using direct preference optimization","work_id":"3bbf25f4-3cec-4668-a9d1-8ddccee8afd8","year":2024},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":2,"title":"Videodpo: Omni-preference alignment for video diffusion generation","work_id":"47508379-7568-448f-b406-1b4e9bbb181b","year":2024},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":4,"title":"Lift: Leveraging human feedback for text-to-video model alignment","work_id":"ab339bba-4351-45db-841d-de086f9c0d67","year":2024},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":5,"title":"Llava-critic: Learning to evaluate multimodal models","work_id":"2d8a9688-7f2b-4224-8107-31c7945ce2cd","year":2024},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":6,"title":"Internlm-xcomposer2.5-reward: A simple yet effective multi-modal reward model","work_id":"499855fc-7d44-43d3-ab3a-1038d2ccd0bc","year":2025}],"snapshot_sha256":"651a51160f69de7cdb1a714c615a241622a86798965a893fba208eafa2cf1838"},"source":{"id":"2503.05236","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-14T00:39:50.045426Z","id":"27542910-c752-4be9-9a1f-1c30da435b53","model_set":{"reader":"grok-4.3"},"one_line_summary":"UnifiedReward is the first unified reward model that jointly assesses multimodal understanding and generation to provide better preference signals for aligning vision models via DPO.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"A single reward model trained jointly on image and video tasks improves preference alignment for both understanding and generation.","strongest_claim":"jointly learning to assess diverse visual tasks yields substantial mutual benefits... achieving consistent improvements across each domain.","weakest_assumption":"The large-scale human preference dataset accurately represents human judgments across tasks and the two-stage filtering strategy produces high-quality, unbiased preference pairs without introducing selection artifacts."}},"verdict_id":"27542910-c752-4be9-9a1f-1c30da435b53"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:473c9c2b0d0024c33d588e902341e4b2330cc8f6133d1f6abe0df8e3cde3a959","target":"record","created_at":"2026-05-18T03:15:18Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"61deb671ab72436b4fe9934b639cef553965b827b86a5f6c7be13ed84a2aa7c3","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2025-03-07T08:36:05Z","title_canon_sha256":"2aedac95cbbedd1322221f81bf6fbba2b5fd14fda6c12942a05f73b48664f4f8"},"schema_version":"1.0","source":{"id":"2503.05236","kind":"arxiv","version":2}},"canonical_sha256":"058598e61c808ba571da6523a560a9e3c7a30acc1708721a83ab8864b033787e","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"058598e61c808ba571da6523a560a9e3c7a30acc1708721a83ab8864b033787e","first_computed_at":"2026-05-18T03:15:18.210251Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-18T03:15:18.210251Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"/9h7O5VyX2IpkoNvNl3rlAAqk03k8MHcWeQzQx+FU3MYSp6aPg/iiFeoibjsCw1yYlkrS01+HNw96Be7lIPDBw==","signature_status":"signed_v1","signed_at":"2026-05-18T03:15:18.211161Z","signed_message":"canonical_sha256_bytes"},"source_id":"2503.05236","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:473c9c2b0d0024c33d588e902341e4b2330cc8f6133d1f6abe0df8e3cde3a959","sha256:22335ef39eb445a23219cb2a0e07a34d98ce9ae3d70a497d30e43d435ea1c108"],"state_sha256":"617f5a4bfb8ec5abaa6dea3a593559a95adbb69f128ce132fe98a568028d07bd"}