{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:HDD6G3Q276LPRQCE6OS727O4U4","short_pith_number":"pith:HDD6G3Q2","schema_version":"1.0","canonical_sha256":"38c7e36e1aff96f8c044f3a5fd7ddca718cd528c7df4576a5981b0354111fa3e","source":{"kind":"arxiv","id":"2503.21979","version":2},"attestation_state":"computed","paper":{"title":"Harmonizing Visual Representations for Unified Multimodal Understanding and Generation","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Chen Change Loy, Lumin Xu, Qingyi Tao, Sheng Jin, Size Wu, Wei Li, Wentao Liu, Wenwei Zhang, Zhonghua Wu","submitted_at":"2025-03-27T20:50:38Z","abstract_excerpt":"Unifying visual understanding and generation within a single multimodal framework remains a significant challenge, as the two inherently heterogeneous tasks require representations at different levels of granularity. Current approaches that utilize vector quantization (VQ) or variational autoencoders (VAE) for unified visual representation prioritize intrinsic imagery features over semantics, compromising understanding performance. In this work, we take inspiration from masked image modelling (MIM) that learns rich semantics via a mask-and-reconstruct pre-training and its successful extension "},"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":"2503.21979","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by-sa/4.0/","primary_cat":"cs.CV","submitted_at":"2025-03-27T20:50:38Z","cross_cats_sorted":[],"title_canon_sha256":"7028085c202708841c945b388e1cf6babd75c605e9798c5f0f3f71cf187f0352","abstract_canon_sha256":"b8cfb125576452fc1e7881a93981df0bb7c437450f6fcd734528ba53cd73721a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:52:21.621535Z","signature_b64":"6zyEnfRPyhlil+HtFlHo25zh/HFBUNes1TtfFyo52y3pRgR3yKT/rcJsniV9WhOrAOXbdN9hhMqbxT2uMZdxAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"38c7e36e1aff96f8c044f3a5fd7ddca718cd528c7df4576a5981b0354111fa3e","last_reissued_at":"2026-07-05T10:52:21.621026Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:52:21.621026Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Harmonizing Visual Representations for Unified Multimodal Understanding and Generation","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Chen Change Loy, Lumin Xu, Qingyi Tao, Sheng Jin, Size Wu, Wei Li, Wentao Liu, Wenwei Zhang, Zhonghua Wu","submitted_at":"2025-03-27T20:50:38Z","abstract_excerpt":"Unifying visual understanding and generation within a single multimodal framework remains a significant challenge, as the two inherently heterogeneous tasks require representations at different levels of granularity. Current approaches that utilize vector quantization (VQ) or variational autoencoders (VAE) for unified visual representation prioritize intrinsic imagery features over semantics, compromising understanding performance. In this work, we take inspiration from masked image modelling (MIM) that learns rich semantics via a mask-and-reconstruct pre-training and its successful extension "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2503.21979","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/2503.21979/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":"2503.21979","created_at":"2026-07-05T10:52:21.621086+00:00"},{"alias_kind":"arxiv_version","alias_value":"2503.21979v2","created_at":"2026-07-05T10:52:21.621086+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2503.21979","created_at":"2026-07-05T10:52:21.621086+00:00"},{"alias_kind":"pith_short_12","alias_value":"HDD6G3Q276LP","created_at":"2026-07-05T10:52:21.621086+00:00"},{"alias_kind":"pith_short_16","alias_value":"HDD6G3Q276LPRQCE","created_at":"2026-07-05T10:52:21.621086+00:00"},{"alias_kind":"pith_short_8","alias_value":"HDD6G3Q2","created_at":"2026-07-05T10:52:21.621086+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":12,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.24333","citing_title":"UniTranslator: A Unified Multi-modal Framework for End-to-end In-Image Machine Translation","ref_index":48,"is_internal_anchor":false},{"citing_arxiv_id":"2606.23041","citing_title":"SPAR: Semantic-Pixel Self-Alignment and Adaptive Routing for Unified Multimodal Models","ref_index":55,"is_internal_anchor":false},{"citing_arxiv_id":"2606.23041","citing_title":"SPAR: Semantic-Pixel Self-Alignment and Adaptive Routing for Unified Multimodal Models","ref_index":62,"is_internal_anchor":false},{"citing_arxiv_id":"2606.13289","citing_title":"HYDRA-X: Native Unified Multimodal Models with Holistic Visual Tokenizers","ref_index":145,"is_internal_anchor":false},{"citing_arxiv_id":"2605.18115","citing_title":"WinTok: A Win-Win Hybrid Tokenizer via Decomposing Visual Understanding and Generation with Transferable Tokens","ref_index":92,"is_internal_anchor":false},{"citing_arxiv_id":"2605.19227","citing_title":"Token by Token, Compromised: Backdoor Vulnerabilities in Unified Autoregressive Models","ref_index":79,"is_internal_anchor":false},{"citing_arxiv_id":"2604.03893","citing_title":"FeynmanBench: Benchmarking Multimodal LLMs on Diagrammatic Physics Reasoning","ref_index":50,"is_internal_anchor":false},{"citing_arxiv_id":"2605.12305","citing_title":"Images in Sentences: Scaling Interleaved Instructions for Unified Visual Generation","ref_index":43,"is_internal_anchor":false},{"citing_arxiv_id":"2604.19954","citing_title":"Camera Control for Text-to-Image Generation via Learning Viewpoint Tokens","ref_index":42,"is_internal_anchor":false},{"citing_arxiv_id":"2604.10949","citing_title":"Pseudo-Unification: Entropy Probing Reveals Divergent Information Patterns in Unified Multimodal Models","ref_index":66,"is_internal_anchor":false},{"citing_arxiv_id":"2604.06966","citing_title":"MAR-GRPO: Stabilized GRPO for AR-diffusion Hybrid Image Generation","ref_index":33,"is_internal_anchor":false},{"citing_arxiv_id":"2604.21904","citing_title":"UniGenDet: A Unified Generative-Discriminative Framework for Co-Evolutionary Image Generation and Generated Image Detection","ref_index":63,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/HDD6G3Q276LPRQCE6OS727O4U4","json":"https://pith.science/pith/HDD6G3Q276LPRQCE6OS727O4U4.json","graph_json":"https://pith.science/api/pith-number/HDD6G3Q276LPRQCE6OS727O4U4/graph.json","events_json":"https://pith.science/api/pith-number/HDD6G3Q276LPRQCE6OS727O4U4/events.json","paper":"https://pith.science/paper/HDD6G3Q2"},"agent_actions":{"view_html":"https://pith.science/pith/HDD6G3Q276LPRQCE6OS727O4U4","download_json":"https://pith.science/pith/HDD6G3Q276LPRQCE6OS727O4U4.json","view_paper":"https://pith.science/paper/HDD6G3Q2","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2503.21979&json=true","fetch_graph":"https://pith.science/api/pith-number/HDD6G3Q276LPRQCE6OS727O4U4/graph.json","fetch_events":"https://pith.science/api/pith-number/HDD6G3Q276LPRQCE6OS727O4U4/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/HDD6G3Q276LPRQCE6OS727O4U4/action/timestamp_anchor","attest_storage":"https://pith.science/pith/HDD6G3Q276LPRQCE6OS727O4U4/action/storage_attestation","attest_author":"https://pith.science/pith/HDD6G3Q276LPRQCE6OS727O4U4/action/author_attestation","sign_citation":"https://pith.science/pith/HDD6G3Q276LPRQCE6OS727O4U4/action/citation_signature","submit_replication":"https://pith.science/pith/HDD6G3Q276LPRQCE6OS727O4U4/action/replication_record"}},"created_at":"2026-07-05T10:52:21.621086+00:00","updated_at":"2026-07-05T10:52:21.621086+00:00"}