{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:32VBHWKAHC63CZA2QXX2YV5T2L","short_pith_number":"pith:32VBHWKA","schema_version":"1.0","canonical_sha256":"deaa13d94038bdb1641a85efac57b3d2ccef2df2cff9ab933f0e6908ce444d27","source":{"kind":"arxiv","id":"2404.15045","version":1},"attestation_state":"computed","paper":{"title":"Multi-Head Mixture-of-Experts","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Furu Wei, Shaohan Huang, Wenhui Wang, Xun Wu","submitted_at":"2024-04-23T13:47:09Z","abstract_excerpt":"Sparse Mixtures of Experts (SMoE) scales model capacity without significant increases in training and inference costs, but exhibits the following two issues: (1) Low expert activation, where only a small subset of experts are activated for optimization. (2) Lacking fine-grained analytical capabilities for multiple semantic concepts within individual tokens. We propose Multi-Head Mixture-of-Experts (MH-MoE), which employs a multi-head mechanism to split each token into multiple sub-tokens. These sub-tokens are then assigned to and processed by a diverse set of experts in parallel, and seamlessl"},"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":"2404.15045","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-04-23T13:47:09Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"ec56aea08cbed81c81424d285544960544b50f5e3c7136ef39e984965fe05328","abstract_canon_sha256":"cf71efedba4762f5de7377d390302e28d7396655dfb645d9b578ecb551b6050e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:11:14.460982Z","signature_b64":"FtPSUj3Y9xtgtWj34t4UTPNaXvH/DjdKZgL71TQ3ZivjlH1jVN3OQgnUgDB8sioH277XJ8/2gjOaSeFYlJAZDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"deaa13d94038bdb1641a85efac57b3d2ccef2df2cff9ab933f0e6908ce444d27","last_reissued_at":"2026-07-05T08:11:14.460493Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:11:14.460493Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Multi-Head Mixture-of-Experts","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Furu Wei, Shaohan Huang, Wenhui Wang, Xun Wu","submitted_at":"2024-04-23T13:47:09Z","abstract_excerpt":"Sparse Mixtures of Experts (SMoE) scales model capacity without significant increases in training and inference costs, but exhibits the following two issues: (1) Low expert activation, where only a small subset of experts are activated for optimization. (2) Lacking fine-grained analytical capabilities for multiple semantic concepts within individual tokens. We propose Multi-Head Mixture-of-Experts (MH-MoE), which employs a multi-head mechanism to split each token into multiple sub-tokens. These sub-tokens are then assigned to and processed by a diverse set of experts in parallel, and seamlessl"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2404.15045","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/2404.15045/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":"2404.15045","created_at":"2026-07-05T08:11:14.460553+00:00"},{"alias_kind":"arxiv_version","alias_value":"2404.15045v1","created_at":"2026-07-05T08:11:14.460553+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2404.15045","created_at":"2026-07-05T08:11:14.460553+00:00"},{"alias_kind":"pith_short_12","alias_value":"32VBHWKAHC63","created_at":"2026-07-05T08:11:14.460553+00:00"},{"alias_kind":"pith_short_16","alias_value":"32VBHWKAHC63CZA2","created_at":"2026-07-05T08:11:14.460553+00:00"},{"alias_kind":"pith_short_8","alias_value":"32VBHWKA","created_at":"2026-07-05T08:11:14.460553+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2502.00281","citing_title":"Sigmoid Self-Attention has Lower Sample Complexity than Softmax Self-Attention: A Mixture-of-Experts Perspective","ref_index":48,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/32VBHWKAHC63CZA2QXX2YV5T2L","json":"https://pith.science/pith/32VBHWKAHC63CZA2QXX2YV5T2L.json","graph_json":"https://pith.science/api/pith-number/32VBHWKAHC63CZA2QXX2YV5T2L/graph.json","events_json":"https://pith.science/api/pith-number/32VBHWKAHC63CZA2QXX2YV5T2L/events.json","paper":"https://pith.science/paper/32VBHWKA"},"agent_actions":{"view_html":"https://pith.science/pith/32VBHWKAHC63CZA2QXX2YV5T2L","download_json":"https://pith.science/pith/32VBHWKAHC63CZA2QXX2YV5T2L.json","view_paper":"https://pith.science/paper/32VBHWKA","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2404.15045&json=true","fetch_graph":"https://pith.science/api/pith-number/32VBHWKAHC63CZA2QXX2YV5T2L/graph.json","fetch_events":"https://pith.science/api/pith-number/32VBHWKAHC63CZA2QXX2YV5T2L/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/32VBHWKAHC63CZA2QXX2YV5T2L/action/timestamp_anchor","attest_storage":"https://pith.science/pith/32VBHWKAHC63CZA2QXX2YV5T2L/action/storage_attestation","attest_author":"https://pith.science/pith/32VBHWKAHC63CZA2QXX2YV5T2L/action/author_attestation","sign_citation":"https://pith.science/pith/32VBHWKAHC63CZA2QXX2YV5T2L/action/citation_signature","submit_replication":"https://pith.science/pith/32VBHWKAHC63CZA2QXX2YV5T2L/action/replication_record"}},"created_at":"2026-07-05T08:11:14.460553+00:00","updated_at":"2026-07-05T08:11:14.460553+00:00"}