{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:GSBN4NSPLUJIVPHKOHTBUPHZCF","short_pith_number":"pith:GSBN4NSP","schema_version":"1.0","canonical_sha256":"3482de364f5d128abcea71e61a3cf911585167b70cc26181614dc8ee6e56c7cb","source":{"kind":"arxiv","id":"2412.01199","version":1},"attestation_state":"computed","paper":{"title":"TinyFusion: Diffusion Transformers Learned Shallow","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CV","authors_text":"Gongfan Fang, Kunjun Li, Xinchao Wang, Xinyin Ma","submitted_at":"2024-12-02T07:05:39Z","abstract_excerpt":"Diffusion Transformers have demonstrated remarkable capabilities in image generation but often come with excessive parameterization, resulting in considerable inference overhead in real-world applications. In this work, we present TinyFusion, a depth pruning method designed to remove redundant layers from diffusion transformers via end-to-end learning. The core principle of our approach is to create a pruned model with high recoverability, allowing it to regain strong performance after fine-tuning. To accomplish this, we introduce a differentiable sampling technique to make pruning learnable, "},"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":"2412.01199","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2024-12-02T07:05:39Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"acf3dafdbd6adaf308ec36320a9abda49fcbc0d144c9da552035264d3c8e264e","abstract_canon_sha256":"904f5fe3ed694b35fa6d695cdcae53ee00f7ab5c9d9a6869e2335d834921ad80"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:42:55.716652Z","signature_b64":"QeF5Xpv04yMy4ZT77YnyDuoB22NKpHEUmmRRTySXqFQnTIgJQ4dtUEEC5WCgSHxWyTrtla7nLY7y9FvB36+HAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"3482de364f5d128abcea71e61a3cf911585167b70cc26181614dc8ee6e56c7cb","last_reissued_at":"2026-07-05T09:42:55.716139Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:42:55.716139Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"TinyFusion: Diffusion Transformers Learned Shallow","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CV","authors_text":"Gongfan Fang, Kunjun Li, Xinchao Wang, Xinyin Ma","submitted_at":"2024-12-02T07:05:39Z","abstract_excerpt":"Diffusion Transformers have demonstrated remarkable capabilities in image generation but often come with excessive parameterization, resulting in considerable inference overhead in real-world applications. In this work, we present TinyFusion, a depth pruning method designed to remove redundant layers from diffusion transformers via end-to-end learning. The core principle of our approach is to create a pruned model with high recoverability, allowing it to regain strong performance after fine-tuning. To accomplish this, we introduce a differentiable sampling technique to make pruning learnable, "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2412.01199","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/2412.01199/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":"2412.01199","created_at":"2026-07-05T09:42:55.716205+00:00"},{"alias_kind":"arxiv_version","alias_value":"2412.01199v1","created_at":"2026-07-05T09:42:55.716205+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2412.01199","created_at":"2026-07-05T09:42:55.716205+00:00"},{"alias_kind":"pith_short_12","alias_value":"GSBN4NSPLUJI","created_at":"2026-07-05T09:42:55.716205+00:00"},{"alias_kind":"pith_short_16","alias_value":"GSBN4NSPLUJIVPHK","created_at":"2026-07-05T09:42:55.716205+00:00"},{"alias_kind":"pith_short_8","alias_value":"GSBN4NSP","created_at":"2026-07-05T09:42:55.716205+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.05204","citing_title":"D-OPSD: On-Policy Self-Distillation for Continuously Tuning Step-Distilled Diffusion Models","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2605.05204","citing_title":"D-OPSD: On-Policy Self-Distillation for Continuously Tuning Step-Distilled Diffusion Models","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2605.05204","citing_title":"D-OPSD: On-Policy Self-Distillation for Continuously Tuning Step-Distilled Diffusion Models","ref_index":19,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/GSBN4NSPLUJIVPHKOHTBUPHZCF","json":"https://pith.science/pith/GSBN4NSPLUJIVPHKOHTBUPHZCF.json","graph_json":"https://pith.science/api/pith-number/GSBN4NSPLUJIVPHKOHTBUPHZCF/graph.json","events_json":"https://pith.science/api/pith-number/GSBN4NSPLUJIVPHKOHTBUPHZCF/events.json","paper":"https://pith.science/paper/GSBN4NSP"},"agent_actions":{"view_html":"https://pith.science/pith/GSBN4NSPLUJIVPHKOHTBUPHZCF","download_json":"https://pith.science/pith/GSBN4NSPLUJIVPHKOHTBUPHZCF.json","view_paper":"https://pith.science/paper/GSBN4NSP","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2412.01199&json=true","fetch_graph":"https://pith.science/api/pith-number/GSBN4NSPLUJIVPHKOHTBUPHZCF/graph.json","fetch_events":"https://pith.science/api/pith-number/GSBN4NSPLUJIVPHKOHTBUPHZCF/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/GSBN4NSPLUJIVPHKOHTBUPHZCF/action/timestamp_anchor","attest_storage":"https://pith.science/pith/GSBN4NSPLUJIVPHKOHTBUPHZCF/action/storage_attestation","attest_author":"https://pith.science/pith/GSBN4NSPLUJIVPHKOHTBUPHZCF/action/author_attestation","sign_citation":"https://pith.science/pith/GSBN4NSPLUJIVPHKOHTBUPHZCF/action/citation_signature","submit_replication":"https://pith.science/pith/GSBN4NSPLUJIVPHKOHTBUPHZCF/action/replication_record"}},"created_at":"2026-07-05T09:42:55.716205+00:00","updated_at":"2026-07-05T09:42:55.716205+00:00"}