{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:ABOQVOZCRP6FUQK6P5VPPBY4LG","short_pith_number":"pith:ABOQVOZC","schema_version":"1.0","canonical_sha256":"005d0abb228bfc5a415e7f6af7871c5999988772bcc889cca3d7ce0efcea4c3c","source":{"kind":"arxiv","id":"2410.01723","version":6},"attestation_state":"computed","paper":{"title":"HarmoniCa: Harmonizing Training and Inference for Better Feature Caching in Diffusion Transformer Acceleration","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Jing Liu, Jinyang Guo, Jun Zhang, Ruihao Gong, Xianglong Liu, Xinjie Zhang, Yushi Huang, Zining Wang","submitted_at":"2024-10-02T16:34:29Z","abstract_excerpt":"Diffusion Transformers (DiTs) excel in generative tasks but face practical deployment challenges due to high inference costs. Feature caching, which stores and retrieves redundant computations, offers the potential for acceleration. Existing learning-based caching, though adaptive, overlooks the impact of the prior timestep. It also suffers from misaligned objectives--aligned predicted noise vs. high-quality images--between training and inference. These two discrepancies compromise both performance and efficiency. To this end, we harmonize training and inference with a novel learning-based cac"},"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":"2410.01723","kind":"arxiv","version":6},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2024-10-02T16:34:29Z","cross_cats_sorted":[],"title_canon_sha256":"a2ab36186761ca2649be978673a53de962326afe08f42379ede01df4dc29ff99","abstract_canon_sha256":"8494585fcb7bd024bec1f11514c8a6aea4b5252d3c9eacc004f42df837fe573f"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:13:13.362524Z","signature_b64":"PP8RdXyqwxzjoKDsTM3zNPHtdWKhLmOVsnnTPpxMKGm367f7Znk2wjuHiQHwYooKmndEBsaQSYqtxZrc4SCzBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"005d0abb228bfc5a415e7f6af7871c5999988772bcc889cca3d7ce0efcea4c3c","last_reissued_at":"2026-07-05T11:13:13.362000Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:13:13.362000Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"HarmoniCa: Harmonizing Training and Inference for Better Feature Caching in Diffusion Transformer Acceleration","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Jing Liu, Jinyang Guo, Jun Zhang, Ruihao Gong, Xianglong Liu, Xinjie Zhang, Yushi Huang, Zining Wang","submitted_at":"2024-10-02T16:34:29Z","abstract_excerpt":"Diffusion Transformers (DiTs) excel in generative tasks but face practical deployment challenges due to high inference costs. Feature caching, which stores and retrieves redundant computations, offers the potential for acceleration. Existing learning-based caching, though adaptive, overlooks the impact of the prior timestep. It also suffers from misaligned objectives--aligned predicted noise vs. high-quality images--between training and inference. These two discrepancies compromise both performance and efficiency. To this end, we harmonize training and inference with a novel learning-based cac"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.01723","kind":"arxiv","version":6},"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/2410.01723/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":"2410.01723","created_at":"2026-07-05T11:13:13.362049+00:00"},{"alias_kind":"arxiv_version","alias_value":"2410.01723v6","created_at":"2026-07-05T11:13:13.362049+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.01723","created_at":"2026-07-05T11:13:13.362049+00:00"},{"alias_kind":"pith_short_12","alias_value":"ABOQVOZCRP6F","created_at":"2026-07-05T11:13:13.362049+00:00"},{"alias_kind":"pith_short_16","alias_value":"ABOQVOZCRP6FUQK6","created_at":"2026-07-05T11:13:13.362049+00:00"},{"alias_kind":"pith_short_8","alias_value":"ABOQVOZC","created_at":"2026-07-05T11:13:13.362049+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.26769","citing_title":"ResilPhase: Plug-and-Play Phase Mapping and Noise-Resilient Macro-Trajectory Extrapolation for Diffusion Acceleration","ref_index":97,"is_internal_anchor":false},{"citing_arxiv_id":"2605.30116","citing_title":"SGMD: Score Gradient Matching Distillation for Few-Step Video Diffusion Distillation","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2604.03118","citing_title":"Salt: Self-Consistent Distribution Matching with Cache-Aware Training for Fast Video Generation","ref_index":16,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/ABOQVOZCRP6FUQK6P5VPPBY4LG","json":"https://pith.science/pith/ABOQVOZCRP6FUQK6P5VPPBY4LG.json","graph_json":"https://pith.science/api/pith-number/ABOQVOZCRP6FUQK6P5VPPBY4LG/graph.json","events_json":"https://pith.science/api/pith-number/ABOQVOZCRP6FUQK6P5VPPBY4LG/events.json","paper":"https://pith.science/paper/ABOQVOZC"},"agent_actions":{"view_html":"https://pith.science/pith/ABOQVOZCRP6FUQK6P5VPPBY4LG","download_json":"https://pith.science/pith/ABOQVOZCRP6FUQK6P5VPPBY4LG.json","view_paper":"https://pith.science/paper/ABOQVOZC","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2410.01723&json=true","fetch_graph":"https://pith.science/api/pith-number/ABOQVOZCRP6FUQK6P5VPPBY4LG/graph.json","fetch_events":"https://pith.science/api/pith-number/ABOQVOZCRP6FUQK6P5VPPBY4LG/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ABOQVOZCRP6FUQK6P5VPPBY4LG/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ABOQVOZCRP6FUQK6P5VPPBY4LG/action/storage_attestation","attest_author":"https://pith.science/pith/ABOQVOZCRP6FUQK6P5VPPBY4LG/action/author_attestation","sign_citation":"https://pith.science/pith/ABOQVOZCRP6FUQK6P5VPPBY4LG/action/citation_signature","submit_replication":"https://pith.science/pith/ABOQVOZCRP6FUQK6P5VPPBY4LG/action/replication_record"}},"created_at":"2026-07-05T11:13:13.362049+00:00","updated_at":"2026-07-05T11:13:13.362049+00:00"}