{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:INIGVSIUJVSX35FTGP57N7GPOX","short_pith_number":"pith:INIGVSIU","schema_version":"1.0","canonical_sha256":"43506ac9144d657df4b333fbf6fccf75cf24aa149d905c231fbcaec3188033ec","source":{"kind":"arxiv","id":"2109.02008","version":3},"attestation_state":"computed","paper":{"title":"Cross-token Modeling with Conditional Computation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CV"],"primary_cat":"cs.LG","authors_text":"Fuzhao Xue, Yang You, Yuxuan Lou, Zangwei Zheng","submitted_at":"2021-09-05T06:43:08Z","abstract_excerpt":"Mixture-of-Experts (MoE), a conditional computation architecture, achieved promising performance by scaling local module (i.e. feed-forward network) of transformer. However, scaling the cross-token module (i.e. self-attention) is challenging due to the unstable training. This work proposes Sparse-MLP, an all-MLP model which applies sparsely-activated MLPs to cross-token modeling. Specifically, in each Sparse block of our all-MLP model, we apply two stages of MoE layers: one with MLP experts mixing information within channels along image patch dimension, the other with MLP experts mixing inform"},"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":"2109.02008","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2021-09-05T06:43:08Z","cross_cats_sorted":["cs.AI","cs.CV"],"title_canon_sha256":"8bf424a2b4b3faf1a2a254bebc72eb0d7628ae3eb3437f1b24afe9706f5e4027","abstract_canon_sha256":"d46d8f138c21aae9f9a03b901af052e3d4e327c8d238a489a7a37efbbc425d1f"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T03:48:19.027551Z","signature_b64":"idn18ReFo/pzyCWPOI4jqRiXS5C/bg/NrLL9ar+vJhW3S7fhYIgxzXC6PZX2PBFdRP6GrZd2H1O1RHtdNhiiBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"43506ac9144d657df4b333fbf6fccf75cf24aa149d905c231fbcaec3188033ec","last_reissued_at":"2026-07-05T03:48:19.027042Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T03:48:19.027042Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Cross-token Modeling with Conditional Computation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CV"],"primary_cat":"cs.LG","authors_text":"Fuzhao Xue, Yang You, Yuxuan Lou, Zangwei Zheng","submitted_at":"2021-09-05T06:43:08Z","abstract_excerpt":"Mixture-of-Experts (MoE), a conditional computation architecture, achieved promising performance by scaling local module (i.e. feed-forward network) of transformer. However, scaling the cross-token module (i.e. self-attention) is challenging due to the unstable training. This work proposes Sparse-MLP, an all-MLP model which applies sparsely-activated MLPs to cross-token modeling. Specifically, in each Sparse block of our all-MLP model, we apply two stages of MoE layers: one with MLP experts mixing information within channels along image patch dimension, the other with MLP experts mixing inform"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2109.02008","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/2109.02008/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":"2109.02008","created_at":"2026-07-05T03:48:19.027105+00:00"},{"alias_kind":"arxiv_version","alias_value":"2109.02008v3","created_at":"2026-07-05T03:48:19.027105+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2109.02008","created_at":"2026-07-05T03:48:19.027105+00:00"},{"alias_kind":"pith_short_12","alias_value":"INIGVSIUJVSX","created_at":"2026-07-05T03:48:19.027105+00:00"},{"alias_kind":"pith_short_16","alias_value":"INIGVSIUJVSX35FT","created_at":"2026-07-05T03:48:19.027105+00:00"},{"alias_kind":"pith_short_8","alias_value":"INIGVSIU","created_at":"2026-07-05T03:48:19.027105+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.18878","citing_title":"Prognostic Value of Lung Ultrasound Biomarkers for Readmission Risk in Congestive Heart Failure: A Pilot Data-Driven Analysis","ref_index":241,"is_internal_anchor":false},{"citing_arxiv_id":"2602.03204","citing_title":"Sparsity is Combinatorial Depth: Quantifying MoE Expressivity via Tropical Geometry","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2602.12708","citing_title":"Mixture of Predefined Experts: Maximizing Data Usage on Vertical Federated Learning","ref_index":29,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/INIGVSIUJVSX35FTGP57N7GPOX","json":"https://pith.science/pith/INIGVSIUJVSX35FTGP57N7GPOX.json","graph_json":"https://pith.science/api/pith-number/INIGVSIUJVSX35FTGP57N7GPOX/graph.json","events_json":"https://pith.science/api/pith-number/INIGVSIUJVSX35FTGP57N7GPOX/events.json","paper":"https://pith.science/paper/INIGVSIU"},"agent_actions":{"view_html":"https://pith.science/pith/INIGVSIUJVSX35FTGP57N7GPOX","download_json":"https://pith.science/pith/INIGVSIUJVSX35FTGP57N7GPOX.json","view_paper":"https://pith.science/paper/INIGVSIU","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2109.02008&json=true","fetch_graph":"https://pith.science/api/pith-number/INIGVSIUJVSX35FTGP57N7GPOX/graph.json","fetch_events":"https://pith.science/api/pith-number/INIGVSIUJVSX35FTGP57N7GPOX/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/INIGVSIUJVSX35FTGP57N7GPOX/action/timestamp_anchor","attest_storage":"https://pith.science/pith/INIGVSIUJVSX35FTGP57N7GPOX/action/storage_attestation","attest_author":"https://pith.science/pith/INIGVSIUJVSX35FTGP57N7GPOX/action/author_attestation","sign_citation":"https://pith.science/pith/INIGVSIUJVSX35FTGP57N7GPOX/action/citation_signature","submit_replication":"https://pith.science/pith/INIGVSIUJVSX35FTGP57N7GPOX/action/replication_record"}},"created_at":"2026-07-05T03:48:19.027105+00:00","updated_at":"2026-07-05T03:48:19.027105+00:00"}