{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:5VTS4UTVINN6JIIHH7JPEIJXRT","short_pith_number":"pith:5VTS4UTV","schema_version":"1.0","canonical_sha256":"ed672e5275435be4a1073fd2f221378cc8d29d28c14255c7ae32c199ed445881","source":{"kind":"arxiv","id":"2407.11633","version":3},"attestation_state":"computed","paper":{"title":"Scaling Diffusion Transformers to 16 Billion Parameters","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Changqian Yu, Debang Li, Junshi Huang, Mingyuan Fan, Zhengcong Fei","submitted_at":"2024-07-16T11:55:23Z","abstract_excerpt":"In this paper, we present DiT-MoE, a sparse version of the diffusion Transformer, that is scalable and competitive with dense networks while exhibiting highly optimized inference. The DiT-MoE includes two simple designs: shared expert routing and expert-level balance loss, thereby capturing common knowledge and reducing redundancy among the different routed experts. When applied to conditional image generation, a deep analysis of experts specialization gains some interesting observations: (i) Expert selection shows preference with spatial position and denoising time step, while insensitive wit"},"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":"2407.11633","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2024-07-16T11:55:23Z","cross_cats_sorted":[],"title_canon_sha256":"c42f3b7e9dcaa1eb264f56b8ef4e2cc48eb5865f3ca39f4129561b4c7cfb2950","abstract_canon_sha256":"5363e8d682e6cde2d0477bfa76ec966542a253deb5c642ef700bae9e60ad1089"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:04:38.106503Z","signature_b64":"0LFUv3kD6WQeekdZ0SiAUPMPk5l4ndOzzBY+RkAUKoiZLrvyLUd7IjR4AsojM7TL4T5IHCo5yORW9qV6ZSWECw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ed672e5275435be4a1073fd2f221378cc8d29d28c14255c7ae32c199ed445881","last_reissued_at":"2026-07-05T09:04:38.106078Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:04:38.106078Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Scaling Diffusion Transformers to 16 Billion Parameters","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Changqian Yu, Debang Li, Junshi Huang, Mingyuan Fan, Zhengcong Fei","submitted_at":"2024-07-16T11:55:23Z","abstract_excerpt":"In this paper, we present DiT-MoE, a sparse version of the diffusion Transformer, that is scalable and competitive with dense networks while exhibiting highly optimized inference. The DiT-MoE includes two simple designs: shared expert routing and expert-level balance loss, thereby capturing common knowledge and reducing redundancy among the different routed experts. When applied to conditional image generation, a deep analysis of experts specialization gains some interesting observations: (i) Expert selection shows preference with spatial position and denoising time step, while insensitive wit"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2407.11633","kind":"arxiv","version":3},"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/2407.11633/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":"2407.11633","created_at":"2026-07-05T09:04:38.106135+00:00"},{"alias_kind":"arxiv_version","alias_value":"2407.11633v3","created_at":"2026-07-05T09:04:38.106135+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2407.11633","created_at":"2026-07-05T09:04:38.106135+00:00"},{"alias_kind":"pith_short_12","alias_value":"5VTS4UTVINN6","created_at":"2026-07-05T09:04:38.106135+00:00"},{"alias_kind":"pith_short_16","alias_value":"5VTS4UTVINN6JIIH","created_at":"2026-07-05T09:04:38.106135+00:00"},{"alias_kind":"pith_short_8","alias_value":"5VTS4UTV","created_at":"2026-07-05T09:04:38.106135+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":15,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.26938","citing_title":"Focusing on What Matters: Saliency-Harnessing Accurate Routing for Diffusion MoE","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2606.23626","citing_title":"DiT-Reward: Generative Representations for Text-to-Image Reward Modeling","ref_index":87,"is_internal_anchor":false},{"citing_arxiv_id":"2606.21868","citing_title":"WiSP: A Working-Set View of Mixture-of-Experts Serving on Extremely Low-Resource Hardware","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2606.21033","citing_title":"MoECodec: Image Compression for joint human and machine perception via Mixture-of-Experts","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2606.02090","citing_title":"FocusDiT: Masking Queries in Diffusion Transformers for Fine-grained Image Generation","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2606.31991","citing_title":"Amplifying Membership Signal Through Chained Regeneration","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2605.30045","citing_title":"DualEraser: Joint Video Object and Effect Removal via Balanced Text-Mask Guidance and Decoupled Locator-Preserver","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2605.19378","citing_title":"Sparse Mixture-of-Experts Routing in Visual Diffusion Transformers:Diagnosis, Boundary Calibration and Evolutionary Roadmap from Routing Collapse to Selective Deadlock","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2502.06608","citing_title":"TripoSG: High-Fidelity 3D Shape Synthesis using Large-Scale Rectified Flow Models","ref_index":130,"is_internal_anchor":false},{"citing_arxiv_id":"2512.21788","citing_title":"InstructMoLE: Instruction-Guided Mixture of Low-rank Experts for Multi-Conditional Image Generation","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2603.10079","citing_title":"Large Spikes in Stochastic Gradient Descent: A Large-Deviations View","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08712","citing_title":"From Articulated Kinematics to Routed Visual Control for Action-Conditioned Surgical Video Generation","ref_index":17,"is_internal_anchor":false},{"citing_arxiv_id":"2604.19636","citing_title":"CoInteract: Physically-Consistent Human-Object Interaction Video Synthesis via Spatially-Structured Co-Generation","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2604.07210","citing_title":"VersaVogue: Visual Expert Orchestration and Preference Alignment for Unified Fashion Synthesis","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2605.02641","citing_title":"Mamoda2.5: Enhancing Unified Multimodal Model with DiT-MoE","ref_index":12,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/5VTS4UTVINN6JIIHH7JPEIJXRT","json":"https://pith.science/pith/5VTS4UTVINN6JIIHH7JPEIJXRT.json","graph_json":"https://pith.science/api/pith-number/5VTS4UTVINN6JIIHH7JPEIJXRT/graph.json","events_json":"https://pith.science/api/pith-number/5VTS4UTVINN6JIIHH7JPEIJXRT/events.json","paper":"https://pith.science/paper/5VTS4UTV"},"agent_actions":{"view_html":"https://pith.science/pith/5VTS4UTVINN6JIIHH7JPEIJXRT","download_json":"https://pith.science/pith/5VTS4UTVINN6JIIHH7JPEIJXRT.json","view_paper":"https://pith.science/paper/5VTS4UTV","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2407.11633&json=true","fetch_graph":"https://pith.science/api/pith-number/5VTS4UTVINN6JIIHH7JPEIJXRT/graph.json","fetch_events":"https://pith.science/api/pith-number/5VTS4UTVINN6JIIHH7JPEIJXRT/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/5VTS4UTVINN6JIIHH7JPEIJXRT/action/timestamp_anchor","attest_storage":"https://pith.science/pith/5VTS4UTVINN6JIIHH7JPEIJXRT/action/storage_attestation","attest_author":"https://pith.science/pith/5VTS4UTVINN6JIIHH7JPEIJXRT/action/author_attestation","sign_citation":"https://pith.science/pith/5VTS4UTVINN6JIIHH7JPEIJXRT/action/citation_signature","submit_replication":"https://pith.science/pith/5VTS4UTVINN6JIIHH7JPEIJXRT/action/replication_record"}},"created_at":"2026-07-05T09:04:38.106135+00:00","updated_at":"2026-07-05T09:04:38.106135+00:00"}