{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:X72AYRLPIFP2UVYVUQLNZSTJUW","short_pith_number":"pith:X72AYRLP","schema_version":"1.0","canonical_sha256":"bff40c456f415faa5715a416dcca69a58f8517e58edb5dbb00fcc891a5172832","source":{"kind":"arxiv","id":"2506.12119","version":2},"attestation_state":"computed","paper":{"title":"Mixture-of-Experts Can Surpass Dense LLMs Under Strictly Equal Resource","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Daxin Jiang, Haocheng Zhang, Houyi Li, Ka Man Lo, Shijie Xuyang, Shuigeng Zhou, Wenzhen Zheng, Xiangyu Zhang, Zhao Li, Ziqi Wang","submitted_at":"2025-06-13T17:59:05Z","abstract_excerpt":"Mixture-of-Experts (MoE) language models dramatically expand model capacity and achieve remarkable performance without increasing per-token compute. However, can MoEs surpass dense architectures under strictly equal resource constraints -- that is, when the total parameter count, training compute, and data budget are identical? This question remains under-explored despite its significant practical value and potential. In this paper, we propose a novel perspective and methodological framework to study this question thoroughly. First, we comprehensively investigate the architecture of MoEs and a"},"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":"2506.12119","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-06-13T17:59:05Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"c8d05a6d1a6abc684516b768f88f8bbc31e1321191376167f229e9fcbd881786","abstract_canon_sha256":"b036db061cb57c216e9138d78a89a575fe4cb1ae247a93116e45a94dbadeef2a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T00:02:51.436871Z","signature_b64":"j7rLgjTmz6X1kXyNDwbTYKohG1dwDYcNftW+ebpouHIWu4gIVDXFEb1B/8DC2xe5M5sKiCycKXvHpVLJwCf6DA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"bff40c456f415faa5715a416dcca69a58f8517e58edb5dbb00fcc891a5172832","last_reissued_at":"2026-05-20T00:02:51.436046Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T00:02:51.436046Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Mixture-of-Experts Can Surpass Dense LLMs Under Strictly Equal Resource","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Daxin Jiang, Haocheng Zhang, Houyi Li, Ka Man Lo, Shijie Xuyang, Shuigeng Zhou, Wenzhen Zheng, Xiangyu Zhang, Zhao Li, Ziqi Wang","submitted_at":"2025-06-13T17:59:05Z","abstract_excerpt":"Mixture-of-Experts (MoE) language models dramatically expand model capacity and achieve remarkable performance without increasing per-token compute. However, can MoEs surpass dense architectures under strictly equal resource constraints -- that is, when the total parameter count, training compute, and data budget are identical? This question remains under-explored despite its significant practical value and potential. In this paper, we propose a novel perspective and methodological framework to study this question thoroughly. First, we comprehensively investigate the architecture of MoEs and a"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.12119","kind":"arxiv","version":2},"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/2506.12119/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":"2506.12119","created_at":"2026-05-20T00:02:51.436182+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.12119v2","created_at":"2026-05-20T00:02:51.436182+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.12119","created_at":"2026-05-20T00:02:51.436182+00:00"},{"alias_kind":"pith_short_12","alias_value":"X72AYRLPIFP2","created_at":"2026-05-20T00:02:51.436182+00:00"},{"alias_kind":"pith_short_16","alias_value":"X72AYRLPIFP2UVYV","created_at":"2026-05-20T00:02:51.436182+00:00"},{"alias_kind":"pith_short_8","alias_value":"X72AYRLP","created_at":"2026-05-20T00:02:51.436182+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":3,"sample":[{"citing_arxiv_id":"2605.10933","citing_title":"DECO: Sparse Mixture-of-Experts with Dense-Comparable Performance on End-Side Devices","ref_index":177,"is_internal_anchor":true},{"citing_arxiv_id":"2605.10933","citing_title":"DECO: Sparse Mixture-of-Experts with Dense-Comparable Performance on End-Side Devices","ref_index":177,"is_internal_anchor":true},{"citing_arxiv_id":"2605.10933","citing_title":"DECO: Sparse Mixture-of-Experts with Dense-Comparable Performance on End-Side Devices","ref_index":177,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/X72AYRLPIFP2UVYVUQLNZSTJUW","json":"https://pith.science/pith/X72AYRLPIFP2UVYVUQLNZSTJUW.json","graph_json":"https://pith.science/api/pith-number/X72AYRLPIFP2UVYVUQLNZSTJUW/graph.json","events_json":"https://pith.science/api/pith-number/X72AYRLPIFP2UVYVUQLNZSTJUW/events.json","paper":"https://pith.science/paper/X72AYRLP"},"agent_actions":{"view_html":"https://pith.science/pith/X72AYRLPIFP2UVYVUQLNZSTJUW","download_json":"https://pith.science/pith/X72AYRLPIFP2UVYVUQLNZSTJUW.json","view_paper":"https://pith.science/paper/X72AYRLP","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.12119&json=true","fetch_graph":"https://pith.science/api/pith-number/X72AYRLPIFP2UVYVUQLNZSTJUW/graph.json","fetch_events":"https://pith.science/api/pith-number/X72AYRLPIFP2UVYVUQLNZSTJUW/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/X72AYRLPIFP2UVYVUQLNZSTJUW/action/timestamp_anchor","attest_storage":"https://pith.science/pith/X72AYRLPIFP2UVYVUQLNZSTJUW/action/storage_attestation","attest_author":"https://pith.science/pith/X72AYRLPIFP2UVYVUQLNZSTJUW/action/author_attestation","sign_citation":"https://pith.science/pith/X72AYRLPIFP2UVYVUQLNZSTJUW/action/citation_signature","submit_replication":"https://pith.science/pith/X72AYRLPIFP2UVYVUQLNZSTJUW/action/replication_record"}},"created_at":"2026-05-20T00:02:51.436182+00:00","updated_at":"2026-05-20T00:02:51.436182+00:00"}