{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:HVAI6H44KNGDDLXRKDRQ5NECC6","short_pith_number":"pith:HVAI6H44","schema_version":"1.0","canonical_sha256":"3d408f1f9c534c31aef150e30eb48217bc188fe40b120b4b8ae7ca0d88fdb78b","source":{"kind":"arxiv","id":"2605.31604","version":1},"attestation_state":"computed","paper":{"title":"Representation Forcing for Bottleneck-Free Unified Multimodal Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Ceyuan Yang, Fei Xiao, Fuyun Wang, Hao He, Haoqi Fan, Qi Zhao, Shuai Wang, Xihui Liu, Yang Zhao, Youliang Zhang, Yuqing Wang, Zhijie Lin, Zihan Ding","submitted_at":"2026-05-29T17:59:55Z","abstract_excerpt":"Unified multimodal models (UMMs) aim to handle perception and generation in a single model. Yet existing UMMs still rely on a frozen, separately pretrained VAE for image generation, imposing a structural bottleneck. Naively removing it introduces a quality gap, as the model must learn both high-level structure and low-level details from raw pixels. In this paper, we propose Representation Forcing (RF), a technique that closes this gap by making representation prediction a native capability of the model. Concretely, RF forces the decoder to autoregressively predict visual representations as int"},"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":"2605.31604","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2026-05-29T17:59:55Z","cross_cats_sorted":[],"title_canon_sha256":"cc37dddce517d400c59ae3d527145c2c56709220a09c8e63ad833f15525a0900","abstract_canon_sha256":"54e655310881ba91bc8d8419729e537bc4c98217e2c55f143a002c3f44f6fb95"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-01T02:04:15.435206Z","signature_b64":"ELu5ix5Niz47QTLSGinT49sjHHgBrgm7SppL0466U1ZR42RcpoeVtqXi+hLy5JD8qbvSmhOVOorG5WQMl2h4Aw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"3d408f1f9c534c31aef150e30eb48217bc188fe40b120b4b8ae7ca0d88fdb78b","last_reissued_at":"2026-06-01T02:04:15.434448Z","signature_status":"signed_v1","first_computed_at":"2026-06-01T02:04:15.434448Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Representation Forcing for Bottleneck-Free Unified Multimodal Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Ceyuan Yang, Fei Xiao, Fuyun Wang, Hao He, Haoqi Fan, Qi Zhao, Shuai Wang, Xihui Liu, Yang Zhao, Youliang Zhang, Yuqing Wang, Zhijie Lin, Zihan Ding","submitted_at":"2026-05-29T17:59:55Z","abstract_excerpt":"Unified multimodal models (UMMs) aim to handle perception and generation in a single model. Yet existing UMMs still rely on a frozen, separately pretrained VAE for image generation, imposing a structural bottleneck. Naively removing it introduces a quality gap, as the model must learn both high-level structure and low-level details from raw pixels. In this paper, we propose Representation Forcing (RF), a technique that closes this gap by making representation prediction a native capability of the model. Concretely, RF forces the decoder to autoregressively predict visual representations as int"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.31604","kind":"arxiv","version":1},"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/2605.31604/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":"2605.31604","created_at":"2026-06-01T02:04:15.434574+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.31604v1","created_at":"2026-06-01T02:04:15.434574+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.31604","created_at":"2026-06-01T02:04:15.434574+00:00"},{"alias_kind":"pith_short_12","alias_value":"HVAI6H44KNGD","created_at":"2026-06-01T02:04:15.434574+00:00"},{"alias_kind":"pith_short_16","alias_value":"HVAI6H44KNGDDLXR","created_at":"2026-06-01T02:04:15.434574+00:00"},{"alias_kind":"pith_short_8","alias_value":"HVAI6H44","created_at":"2026-06-01T02:04:15.434574+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/HVAI6H44KNGDDLXRKDRQ5NECC6","json":"https://pith.science/pith/HVAI6H44KNGDDLXRKDRQ5NECC6.json","graph_json":"https://pith.science/api/pith-number/HVAI6H44KNGDDLXRKDRQ5NECC6/graph.json","events_json":"https://pith.science/api/pith-number/HVAI6H44KNGDDLXRKDRQ5NECC6/events.json","paper":"https://pith.science/paper/HVAI6H44"},"agent_actions":{"view_html":"https://pith.science/pith/HVAI6H44KNGDDLXRKDRQ5NECC6","download_json":"https://pith.science/pith/HVAI6H44KNGDDLXRKDRQ5NECC6.json","view_paper":"https://pith.science/paper/HVAI6H44","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.31604&json=true","fetch_graph":"https://pith.science/api/pith-number/HVAI6H44KNGDDLXRKDRQ5NECC6/graph.json","fetch_events":"https://pith.science/api/pith-number/HVAI6H44KNGDDLXRKDRQ5NECC6/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/HVAI6H44KNGDDLXRKDRQ5NECC6/action/timestamp_anchor","attest_storage":"https://pith.science/pith/HVAI6H44KNGDDLXRKDRQ5NECC6/action/storage_attestation","attest_author":"https://pith.science/pith/HVAI6H44KNGDDLXRKDRQ5NECC6/action/author_attestation","sign_citation":"https://pith.science/pith/HVAI6H44KNGDDLXRKDRQ5NECC6/action/citation_signature","submit_replication":"https://pith.science/pith/HVAI6H44KNGDDLXRKDRQ5NECC6/action/replication_record"}},"created_at":"2026-06-01T02:04:15.434574+00:00","updated_at":"2026-06-01T02:04:15.434574+00:00"}