{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:7H3MJSG5BCB4B53Y47DGJ7T4YX","short_pith_number":"pith:7H3MJSG5","schema_version":"1.0","canonical_sha256":"f9f6c4c8dd0883c0f778e7c664fe7cc5de653f41d593862fbbdccee0a89eef72","source":{"kind":"arxiv","id":"2505.13380","version":1},"attestation_state":"computed","paper":{"title":"CompeteSMoE -- Statistically Guaranteed Mixture of Experts Training via Competition","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.AI","authors_text":"Huy Nguyen, Nam V. Nguyen, Nhat Ho, Quang Pham, Savitha Ramasamy, Van Nguyen","submitted_at":"2025-05-19T17:24:26Z","abstract_excerpt":"Sparse mixture of experts (SMoE) offers an appealing solution to scale up the model complexity beyond the mean of increasing the network's depth or width. However, we argue that effective SMoE training remains challenging because of the suboptimal routing process where experts that perform computation do not directly contribute to the routing process. In this work, we propose competition, a novel mechanism to route tokens to experts with the highest neural response. Theoretically, we show that the competition mechanism enjoys a better sample efficiency than the traditional softmax routing. Fur"},"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":"2505.13380","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.AI","submitted_at":"2025-05-19T17:24:26Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"c03ed8e583c47f6f0771eb78d5d400326c53edd86d1e6cb29652adfe7cef23fa","abstract_canon_sha256":"b26298cc48e2109c7b7658831d9f092e5ec793dbd562da1621f100868a577be3"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:05:29.233379Z","signature_b64":"7+0cbtMLZms6UFJvE7784QfgWdKfJRe9F4BYQlFEwbRcH7QwO7MVwg9pkcIK9ZNGpxInYrX20G52mMCd09YWBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f9f6c4c8dd0883c0f778e7c664fe7cc5de653f41d593862fbbdccee0a89eef72","last_reissued_at":"2026-07-05T11:05:29.232883Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:05:29.232883Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"CompeteSMoE -- Statistically Guaranteed Mixture of Experts Training via Competition","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.AI","authors_text":"Huy Nguyen, Nam V. Nguyen, Nhat Ho, Quang Pham, Savitha Ramasamy, Van Nguyen","submitted_at":"2025-05-19T17:24:26Z","abstract_excerpt":"Sparse mixture of experts (SMoE) offers an appealing solution to scale up the model complexity beyond the mean of increasing the network's depth or width. However, we argue that effective SMoE training remains challenging because of the suboptimal routing process where experts that perform computation do not directly contribute to the routing process. In this work, we propose competition, a novel mechanism to route tokens to experts with the highest neural response. Theoretically, we show that the competition mechanism enjoys a better sample efficiency than the traditional softmax routing. Fur"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.13380","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/2505.13380/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":"2505.13380","created_at":"2026-07-05T11:05:29.232940+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.13380v1","created_at":"2026-07-05T11:05:29.232940+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.13380","created_at":"2026-07-05T11:05:29.232940+00:00"},{"alias_kind":"pith_short_12","alias_value":"7H3MJSG5BCB4","created_at":"2026-07-05T11:05:29.232940+00:00"},{"alias_kind":"pith_short_16","alias_value":"7H3MJSG5BCB4B53Y","created_at":"2026-07-05T11:05:29.232940+00:00"},{"alias_kind":"pith_short_8","alias_value":"7H3MJSG5","created_at":"2026-07-05T11:05:29.232940+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/7H3MJSG5BCB4B53Y47DGJ7T4YX","json":"https://pith.science/pith/7H3MJSG5BCB4B53Y47DGJ7T4YX.json","graph_json":"https://pith.science/api/pith-number/7H3MJSG5BCB4B53Y47DGJ7T4YX/graph.json","events_json":"https://pith.science/api/pith-number/7H3MJSG5BCB4B53Y47DGJ7T4YX/events.json","paper":"https://pith.science/paper/7H3MJSG5"},"agent_actions":{"view_html":"https://pith.science/pith/7H3MJSG5BCB4B53Y47DGJ7T4YX","download_json":"https://pith.science/pith/7H3MJSG5BCB4B53Y47DGJ7T4YX.json","view_paper":"https://pith.science/paper/7H3MJSG5","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.13380&json=true","fetch_graph":"https://pith.science/api/pith-number/7H3MJSG5BCB4B53Y47DGJ7T4YX/graph.json","fetch_events":"https://pith.science/api/pith-number/7H3MJSG5BCB4B53Y47DGJ7T4YX/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/7H3MJSG5BCB4B53Y47DGJ7T4YX/action/timestamp_anchor","attest_storage":"https://pith.science/pith/7H3MJSG5BCB4B53Y47DGJ7T4YX/action/storage_attestation","attest_author":"https://pith.science/pith/7H3MJSG5BCB4B53Y47DGJ7T4YX/action/author_attestation","sign_citation":"https://pith.science/pith/7H3MJSG5BCB4B53Y47DGJ7T4YX/action/citation_signature","submit_replication":"https://pith.science/pith/7H3MJSG5BCB4B53Y47DGJ7T4YX/action/replication_record"}},"created_at":"2026-07-05T11:05:29.232940+00:00","updated_at":"2026-07-05T11:05:29.232940+00:00"}