{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:6WDVKKSFBK4XGFSUC5SCV4BXMY","short_pith_number":"pith:6WDVKKSF","canonical_record":{"source":{"id":"2605.13111","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2026-05-13T07:23:02Z","cross_cats_sorted":[],"title_canon_sha256":"ce189f10a6b255df28ffdde68e05f4dfca67e591ce6a0c610c34cf63fb17e514","abstract_canon_sha256":"75348fc516cdb834c97cc03a6a77d15ec404b5a3b0f3efe1bd8c934986d90c1c"},"schema_version":"1.0"},"canonical_sha256":"f587552a450ab973165417642af0376610e9f47f9fca2987620064cce90f9923","source":{"kind":"arxiv","id":"2605.13111","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.13111","created_at":"2026-05-18T03:08:58Z"},{"alias_kind":"arxiv_version","alias_value":"2605.13111v1","created_at":"2026-05-18T03:08:58Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.13111","created_at":"2026-05-18T03:08:58Z"},{"alias_kind":"pith_short_12","alias_value":"6WDVKKSFBK4X","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"6WDVKKSFBK4XGFSU","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"6WDVKKSF","created_at":"2026-05-18T12:33:37Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:6WDVKKSFBK4XGFSUC5SCV4BXMY","target":"record","payload":{"canonical_record":{"source":{"id":"2605.13111","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2026-05-13T07:23:02Z","cross_cats_sorted":[],"title_canon_sha256":"ce189f10a6b255df28ffdde68e05f4dfca67e591ce6a0c610c34cf63fb17e514","abstract_canon_sha256":"75348fc516cdb834c97cc03a6a77d15ec404b5a3b0f3efe1bd8c934986d90c1c"},"schema_version":"1.0"},"canonical_sha256":"f587552a450ab973165417642af0376610e9f47f9fca2987620064cce90f9923","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-18T03:08:58.115900Z","signature_b64":"lbEOQAr1gnrIRUbCXhhje0/MjhCWzZSJ0UHNJbNO1/aCNNRMzLgII2DXgns3+mJKNP36U+f431hsxryCcMFHDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f587552a450ab973165417642af0376610e9f47f9fca2987620064cce90f9923","last_reissued_at":"2026-05-18T03:08:58.115445Z","signature_status":"signed_v1","first_computed_at":"2026-05-18T03:08:58.115445Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2605.13111","source_version":1,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-18T03:08:58Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"12Ud3ItZkGRdw5VXjPNmOzsOhuNYjSPkfxIgzqz2aaWGIwaZdPsGXhxpLxzu+PWdV+UbN5RtZrspJ6WDYFPKDw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-02T11:28:48.807056Z"},"content_sha256":"6a671af55efe727b9a60c2c0a9df5ac0b6ad2f6ab4464ef5be312ce6f553ba43","schema_version":"1.0","event_id":"sha256:6a671af55efe727b9a60c2c0a9df5ac0b6ad2f6ab4464ef5be312ce6f553ba43"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:6WDVKKSFBK4XGFSUC5SCV4BXMY","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Pyramid Forcing: Head-Aware Pyramid KV Cache Policy for High-Quality Long Video Generation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Pyramid Forcing assigns different KV cache lengths to three attention head types to reduce error accumulation in long autoregressive video generation.","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Guojie Luo, Jiawei Yang, Jiayi Luo, Jiayu Chen, Junbei Tang, Maoliang Li, Wenbiao Zhao, Xiang Chen, Zihao Zheng","submitted_at":"2026-05-13T07:23:02Z","abstract_excerpt":"Autoregressive video generation enables streaming and open-ended long video synthesis, but still suffers from long-term degradation caused by accumulated errors. Existing KVCache strategies usually apply unified historical-frame retention, implicitly assuming homogeneous historical dependencies across attention heads. We revisit historical-frame attention and reveal three distinct head types: Anchor Heads require broad long-range context, Wave Heads exhibit periodic temporal dependencies, and Veil Heads focus on initial and adjacent frames. Based on this finding, we propose Pyramid Forcing, a "},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Pyramid Forcing consistently improves long-horizon generation quality on VBench-Long, increasing the 60-second Self Forcing score from 77.87 to 81.21 while enhancing motion dynamics, visual fidelity, and semantic consistency.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The three head types (Anchor, Wave, Veil) are stable across different models and datasets and can be reliably identified offline without retraining or runtime overhead.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Pyramid Forcing classifies attention heads into Anchor, Wave, and Veil types and applies type-specific KV cache policies to improve long-horizon autoregressive video generation quality.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Pyramid Forcing assigns different KV cache lengths to three attention head types to reduce error accumulation in long autoregressive video generation.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"0034745f686999cf8d3a57fd914a790d63029398a276cab7a25fe2884107ad44"},"source":{"id":"2605.13111","kind":"arxiv","version":1},"verdict":{"id":"a47233f5-e04b-434f-bff1-a76a4301e02d","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-14T19:45:28.528497Z","strongest_claim":"Pyramid Forcing consistently improves long-horizon generation quality on VBench-Long, increasing the 60-second Self Forcing score from 77.87 to 81.21 while enhancing motion dynamics, visual fidelity, and semantic consistency.","one_line_summary":"Pyramid Forcing classifies attention heads into Anchor, Wave, and Veil types and applies type-specific KV cache policies to improve long-horizon autoregressive video generation quality.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The three head types (Anchor, Wave, Veil) are stable across different models and datasets and can be reliably identified offline without retraining or runtime overhead.","pith_extraction_headline":"Pyramid Forcing assigns different KV cache lengths to three attention head types to reduce error accumulation in long autoregressive video generation."},"references":{"count":37,"sample":[{"doi":"","year":2025,"title":"Self Forcing: Bridging the Train-Test Gap in Autoregressive Video Diffusion","work_id":"53e58ef9-7932-4b83-b757-34ac14db3e0f","ref_index":1,"cited_arxiv_id":"2506.08009","is_internal_anchor":true},{"doi":"","year":2025,"title":"MAGI-1: Autoregressive Video Generation at Scale","work_id":"25e8bd3d-e51c-43ae-8126-4ea6ecdb3321","ref_index":2,"cited_arxiv_id":"2505.13211","is_internal_anchor":true},{"doi":"","year":2026,"title":"Causal forcing: Autoregressive diffusion distillation done right for high-quality real- time interactive video generation","work_id":"04f67f5b-e79a-4ad9-8e90-763e5e54bd3d","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"Scalable diffusion models with transformers","work_id":"3e203719-f1d3-4517-959c-98e6121f3e23","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2025,"title":"Wan: Open and Advanced Large-Scale Video Generative Models","work_id":"ad3ebc3b-4224-46c9-b61d-bcf135da0a7c","ref_index":5,"cited_arxiv_id":"2503.20314","is_internal_anchor":true}],"resolved_work":37,"snapshot_sha256":"0a92017860eae31d2d678829bc712e5e95c682d41cfe1ca05185fbb2b7ccdfd3","internal_anchors":10},"formal_canon":{"evidence_count":2,"snapshot_sha256":"cdf72e11bb671562156e1ed3f09472f4931f15d2709f30fe7a0712cb7d21041a"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"verdict_id":"a47233f5-e04b-434f-bff1-a76a4301e02d"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-18T03:08:58Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"xA/ovVdT50oGcjpCEPnJwnlAzWCkiP/UMnyU9DNW30wn9yPU2jos3No6fur8oryxOyF3XRARTCTqaz9Oc2yGAA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-02T11:28:48.808386Z"},"content_sha256":"a1874c2eb8b2ff50be846ca9462bdaada700b3334d55754f93be18d32d757e30","schema_version":"1.0","event_id":"sha256:a1874c2eb8b2ff50be846ca9462bdaada700b3334d55754f93be18d32d757e30"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/6WDVKKSFBK4XGFSUC5SCV4BXMY/bundle.json","state_url":"https://pith.science/pith/6WDVKKSFBK4XGFSUC5SCV4BXMY/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/6WDVKKSFBK4XGFSUC5SCV4BXMY/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-08-02T11:28:48Z","links":{"resolver":"https://pith.science/pith/6WDVKKSFBK4XGFSUC5SCV4BXMY","bundle":"https://pith.science/pith/6WDVKKSFBK4XGFSUC5SCV4BXMY/bundle.json","state":"https://pith.science/pith/6WDVKKSFBK4XGFSUC5SCV4BXMY/state.json","well_known_bundle":"https://pith.science/.well-known/pith/6WDVKKSFBK4XGFSUC5SCV4BXMY/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:6WDVKKSFBK4XGFSUC5SCV4BXMY","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":"75348fc516cdb834c97cc03a6a77d15ec404b5a3b0f3efe1bd8c934986d90c1c","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2026-05-13T07:23:02Z","title_canon_sha256":"ce189f10a6b255df28ffdde68e05f4dfca67e591ce6a0c610c34cf63fb17e514"},"schema_version":"1.0","source":{"id":"2605.13111","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.13111","created_at":"2026-05-18T03:08:58Z"},{"alias_kind":"arxiv_version","alias_value":"2605.13111v1","created_at":"2026-05-18T03:08:58Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.13111","created_at":"2026-05-18T03:08:58Z"},{"alias_kind":"pith_short_12","alias_value":"6WDVKKSFBK4X","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"6WDVKKSFBK4XGFSU","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"6WDVKKSF","created_at":"2026-05-18T12:33:37Z"}],"graph_snapshots":[{"event_id":"sha256:a1874c2eb8b2ff50be846ca9462bdaada700b3334d55754f93be18d32d757e30","target":"graph","created_at":"2026-05-18T03:08:58Z","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":"Pyramid Forcing consistently improves long-horizon generation quality on VBench-Long, increasing the 60-second Self Forcing score from 77.87 to 81.21 while enhancing motion dynamics, visual fidelity, and semantic consistency."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"The three head types (Anchor, Wave, Veil) are stable across different models and datasets and can be reliably identified offline without retraining or runtime overhead."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"Pyramid Forcing classifies attention heads into Anchor, Wave, and Veil types and applies type-specific KV cache policies to improve long-horizon autoregressive video generation quality."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Pyramid Forcing assigns different KV cache lengths to three attention head types to reduce error accumulation in long autoregressive video generation."}],"snapshot_sha256":"0034745f686999cf8d3a57fd914a790d63029398a276cab7a25fe2884107ad44"},"formal_canon":{"evidence_count":2,"snapshot_sha256":"cdf72e11bb671562156e1ed3f09472f4931f15d2709f30fe7a0712cb7d21041a"},"paper":{"abstract_excerpt":"Autoregressive video generation enables streaming and open-ended long video synthesis, but still suffers from long-term degradation caused by accumulated errors. Existing KVCache strategies usually apply unified historical-frame retention, implicitly assuming homogeneous historical dependencies across attention heads. We revisit historical-frame attention and reveal three distinct head types: Anchor Heads require broad long-range context, Wave Heads exhibit periodic temporal dependencies, and Veil Heads focus on initial and adjacent frames. Based on this finding, we propose Pyramid Forcing, a ","authors_text":"Guojie Luo, Jiawei Yang, Jiayi Luo, Jiayu Chen, Junbei Tang, Maoliang Li, Wenbiao Zhao, Xiang Chen, Zihao Zheng","cross_cats":[],"headline":"Pyramid Forcing assigns different KV cache lengths to three attention head types to reduce error accumulation in long autoregressive video generation.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2026-05-13T07:23:02Z","title":"Pyramid Forcing: Head-Aware Pyramid KV Cache Policy for High-Quality Long Video Generation"},"references":{"count":37,"internal_anchors":10,"resolved_work":37,"sample":[{"cited_arxiv_id":"2506.08009","doi":"","is_internal_anchor":true,"ref_index":1,"title":"Self Forcing: Bridging the Train-Test Gap in Autoregressive Video Diffusion","work_id":"53e58ef9-7932-4b83-b757-34ac14db3e0f","year":2025},{"cited_arxiv_id":"2505.13211","doi":"","is_internal_anchor":true,"ref_index":2,"title":"MAGI-1: Autoregressive Video Generation at Scale","work_id":"25e8bd3d-e51c-43ae-8126-4ea6ecdb3321","year":2025},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":3,"title":"Causal forcing: Autoregressive diffusion distillation done right for high-quality real- time interactive video generation","work_id":"04f67f5b-e79a-4ad9-8e90-763e5e54bd3d","year":2026},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":4,"title":"Scalable diffusion models with transformers","work_id":"3e203719-f1d3-4517-959c-98e6121f3e23","year":2023},{"cited_arxiv_id":"2503.20314","doi":"","is_internal_anchor":true,"ref_index":5,"title":"Wan: Open and Advanced Large-Scale Video Generative Models","work_id":"ad3ebc3b-4224-46c9-b61d-bcf135da0a7c","year":2025}],"snapshot_sha256":"0a92017860eae31d2d678829bc712e5e95c682d41cfe1ca05185fbb2b7ccdfd3"},"source":{"id":"2605.13111","kind":"arxiv","version":1},"verdict":{"created_at":"2026-05-14T19:45:28.528497Z","id":"a47233f5-e04b-434f-bff1-a76a4301e02d","model_set":{"reader":"grok-4.3"},"one_line_summary":"Pyramid Forcing classifies attention heads into Anchor, Wave, and Veil types and applies type-specific KV cache policies to improve long-horizon autoregressive video generation quality.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Pyramid Forcing assigns different KV cache lengths to three attention head types to reduce error accumulation in long autoregressive video generation.","strongest_claim":"Pyramid Forcing consistently improves long-horizon generation quality on VBench-Long, increasing the 60-second Self Forcing score from 77.87 to 81.21 while enhancing motion dynamics, visual fidelity, and semantic consistency.","weakest_assumption":"The three head types (Anchor, Wave, Veil) are stable across different models and datasets and can be reliably identified offline without retraining or runtime overhead."}},"verdict_id":"a47233f5-e04b-434f-bff1-a76a4301e02d"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:6a671af55efe727b9a60c2c0a9df5ac0b6ad2f6ab4464ef5be312ce6f553ba43","target":"record","created_at":"2026-05-18T03:08:58Z","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":"75348fc516cdb834c97cc03a6a77d15ec404b5a3b0f3efe1bd8c934986d90c1c","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2026-05-13T07:23:02Z","title_canon_sha256":"ce189f10a6b255df28ffdde68e05f4dfca67e591ce6a0c610c34cf63fb17e514"},"schema_version":"1.0","source":{"id":"2605.13111","kind":"arxiv","version":1}},"canonical_sha256":"f587552a450ab973165417642af0376610e9f47f9fca2987620064cce90f9923","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"f587552a450ab973165417642af0376610e9f47f9fca2987620064cce90f9923","first_computed_at":"2026-05-18T03:08:58.115445Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-18T03:08:58.115445Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"lbEOQAr1gnrIRUbCXhhje0/MjhCWzZSJ0UHNJbNO1/aCNNRMzLgII2DXgns3+mJKNP36U+f431hsxryCcMFHDw==","signature_status":"signed_v1","signed_at":"2026-05-18T03:08:58.115900Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.13111","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:6a671af55efe727b9a60c2c0a9df5ac0b6ad2f6ab4464ef5be312ce6f553ba43","sha256:a1874c2eb8b2ff50be846ca9462bdaada700b3334d55754f93be18d32d757e30"],"state_sha256":"1a3da123589284b0b15ebc85005436e2ce6e8344208fae54e22c60e4abeace3d"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"ilBepa7SeXLuJpXa3wLtlqBGtg1qPG+wIfZXftXAqYY9/w+oNdvlGJy8VX80DnRlRIGs0GNxClfziS8+9LpiCw==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-02T11:28:48.816666Z","bundle_sha256":"06903cf6e92e0d25a31bc96203baaa44576780dd6635bdccbde5626e1f9a609f"}}