{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:QZYEY5TJ26IX3MH2UEVMJELF3S","short_pith_number":"pith:QZYEY5TJ","schema_version":"1.0","canonical_sha256":"86704c7669d7917db0faa12ac49165dcaebe1325d6d31a3838231c4cdddeb13b","source":{"kind":"arxiv","id":"2501.12389","version":1},"attestation_state":"computed","paper":{"title":"Taming Teacher Forcing for Masked Autoregressive Video Generation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Deyu Zhou, Duomin Wang, Heung-Yeung Shum, Kun Yan, Lionel M. Ni, Nan Duan, Quan Sun, Runpei Dong, Xiangyu Zhang, Yuang Peng, Zheng Ge","submitted_at":"2025-01-21T18:59:31Z","abstract_excerpt":"We introduce MAGI, a hybrid video generation framework that combines masked modeling for intra-frame generation with causal modeling for next-frame generation. Our key innovation, Complete Teacher Forcing (CTF), conditions masked frames on complete observation frames rather than masked ones (namely Masked Teacher Forcing, MTF), enabling a smooth transition from token-level (patch-level) to frame-level autoregressive generation. CTF significantly outperforms MTF, achieving a +23% improvement in FVD scores on first-frame conditioned video prediction. To address issues like exposure bias, we empl"},"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":"2501.12389","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2025-01-21T18:59:31Z","cross_cats_sorted":[],"title_canon_sha256":"ae937970bb349eae2bbf18952385f75db069d83ee413b85575bbd3f59cc3821a","abstract_canon_sha256":"d85f271424c8ef13b2eea5800db1f735d9ad5ef896f31f67e6d15686e69385ce"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:03:35.566651Z","signature_b64":"Pm4Q66g1ckBLRc7LFOopDlZVrHme9J/frg/0K2kgbdDpQTpeNvUtFgZrxErDmpBMcRE741hqR4KEhat01m9UBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"86704c7669d7917db0faa12ac49165dcaebe1325d6d31a3838231c4cdddeb13b","last_reissued_at":"2026-07-05T10:03:35.566202Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:03:35.566202Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Taming Teacher Forcing for Masked Autoregressive Video Generation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Deyu Zhou, Duomin Wang, Heung-Yeung Shum, Kun Yan, Lionel M. Ni, Nan Duan, Quan Sun, Runpei Dong, Xiangyu Zhang, Yuang Peng, Zheng Ge","submitted_at":"2025-01-21T18:59:31Z","abstract_excerpt":"We introduce MAGI, a hybrid video generation framework that combines masked modeling for intra-frame generation with causal modeling for next-frame generation. Our key innovation, Complete Teacher Forcing (CTF), conditions masked frames on complete observation frames rather than masked ones (namely Masked Teacher Forcing, MTF), enabling a smooth transition from token-level (patch-level) to frame-level autoregressive generation. CTF significantly outperforms MTF, achieving a +23% improvement in FVD scores on first-frame conditioned video prediction. To address issues like exposure bias, we empl"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2501.12389","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/2501.12389/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":"2501.12389","created_at":"2026-07-05T10:03:35.566259+00:00"},{"alias_kind":"arxiv_version","alias_value":"2501.12389v1","created_at":"2026-07-05T10:03:35.566259+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2501.12389","created_at":"2026-07-05T10:03:35.566259+00:00"},{"alias_kind":"pith_short_12","alias_value":"QZYEY5TJ26IX","created_at":"2026-07-05T10:03:35.566259+00:00"},{"alias_kind":"pith_short_16","alias_value":"QZYEY5TJ26IX3MH2","created_at":"2026-07-05T10:03:35.566259+00:00"},{"alias_kind":"pith_short_8","alias_value":"QZYEY5TJ","created_at":"2026-07-05T10:03:35.566259+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.30083","citing_title":"Future Forcing: Future-aware Training-free KV Cache Policy for Autoregressive Video Generation","ref_index":60,"is_internal_anchor":false},{"citing_arxiv_id":"2502.10248","citing_title":"Step-Video-T2V Technical Report: The Practice, Challenges, and Future of Video Foundation Model","ref_index":56,"is_internal_anchor":false},{"citing_arxiv_id":"2505.05472","citing_title":"Mogao: An Omni Foundation Model for Interleaved Multi-Modal Generation","ref_index":98,"is_internal_anchor":false},{"citing_arxiv_id":"2503.19325","citing_title":"Long-Context Autoregressive Video Modeling with Next-Frame Prediction","ref_index":27,"is_internal_anchor":false},{"citing_arxiv_id":"2509.22622","citing_title":"LongLive: Real-time Interactive Long Video Generation","ref_index":51,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/QZYEY5TJ26IX3MH2UEVMJELF3S","json":"https://pith.science/pith/QZYEY5TJ26IX3MH2UEVMJELF3S.json","graph_json":"https://pith.science/api/pith-number/QZYEY5TJ26IX3MH2UEVMJELF3S/graph.json","events_json":"https://pith.science/api/pith-number/QZYEY5TJ26IX3MH2UEVMJELF3S/events.json","paper":"https://pith.science/paper/QZYEY5TJ"},"agent_actions":{"view_html":"https://pith.science/pith/QZYEY5TJ26IX3MH2UEVMJELF3S","download_json":"https://pith.science/pith/QZYEY5TJ26IX3MH2UEVMJELF3S.json","view_paper":"https://pith.science/paper/QZYEY5TJ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2501.12389&json=true","fetch_graph":"https://pith.science/api/pith-number/QZYEY5TJ26IX3MH2UEVMJELF3S/graph.json","fetch_events":"https://pith.science/api/pith-number/QZYEY5TJ26IX3MH2UEVMJELF3S/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/QZYEY5TJ26IX3MH2UEVMJELF3S/action/timestamp_anchor","attest_storage":"https://pith.science/pith/QZYEY5TJ26IX3MH2UEVMJELF3S/action/storage_attestation","attest_author":"https://pith.science/pith/QZYEY5TJ26IX3MH2UEVMJELF3S/action/author_attestation","sign_citation":"https://pith.science/pith/QZYEY5TJ26IX3MH2UEVMJELF3S/action/citation_signature","submit_replication":"https://pith.science/pith/QZYEY5TJ26IX3MH2UEVMJELF3S/action/replication_record"}},"created_at":"2026-07-05T10:03:35.566259+00:00","updated_at":"2026-07-05T10:03:35.566259+00:00"}