{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:RW5XONPRALUDJDUGBESKZJT2PU","short_pith_number":"pith:RW5XONPR","schema_version":"1.0","canonical_sha256":"8dbb7735f102e8348e860924aca67a7d08187c5ca7fa346374c42620a351bd02","source":{"kind":"arxiv","id":"2406.06563","version":1},"attestation_state":"computed","paper":{"title":"Skywork-MoE: A Deep Dive into Training Techniques for Mixture-of-Experts Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Biye Li, Bo Zhu, Cheng cheng, Han Fang, Jianhao Zhang, Liang Zeng, Liang Zhao, Peng Cheng, Rui Hu, Shuicheng Yan, Tianwen Wei, Weiwei L\\\"u, Xiaokun Wang, Xiaoyu Zhang, Yahui Zhou, Yutuan Ma","submitted_at":"2024-06-03T03:58:41Z","abstract_excerpt":"In this technical report, we introduce the training methodologies implemented in the development of Skywork-MoE, a high-performance mixture-of-experts (MoE) large language model (LLM) with 146 billion parameters and 16 experts. It is initialized from the pre-existing dense checkpoints of our Skywork-13B model. We explore the comparative effectiveness of upcycling versus training from scratch initializations. Our findings suggest that the choice between these two approaches should consider both the performance of the existing dense checkpoints and the MoE training budget. We highlight two innov"},"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":"2406.06563","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-06-03T03:58:41Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"19be60bf4420a9da3c11bd67d528d09e86c9af4e75b575b31331cdc395641525","abstract_canon_sha256":"70e6a03e57a3b2cb107beef0010d4d9d631c63401d32aa1285188fe444594fa1"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:29:47.524159Z","signature_b64":"dFGZVznwsekFi5V0aWqubTB1yPPl/STRXXQ0G7Uj6xFY8d2YTfkaUayK5vYvtzlFMwMR3dlJ8l/gtPRrxBbGCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8dbb7735f102e8348e860924aca67a7d08187c5ca7fa346374c42620a351bd02","last_reissued_at":"2026-07-05T08:29:47.523700Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:29:47.523700Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Skywork-MoE: A Deep Dive into Training Techniques for Mixture-of-Experts Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Biye Li, Bo Zhu, Cheng cheng, Han Fang, Jianhao Zhang, Liang Zeng, Liang Zhao, Peng Cheng, Rui Hu, Shuicheng Yan, Tianwen Wei, Weiwei L\\\"u, Xiaokun Wang, Xiaoyu Zhang, Yahui Zhou, Yutuan Ma","submitted_at":"2024-06-03T03:58:41Z","abstract_excerpt":"In this technical report, we introduce the training methodologies implemented in the development of Skywork-MoE, a high-performance mixture-of-experts (MoE) large language model (LLM) with 146 billion parameters and 16 experts. It is initialized from the pre-existing dense checkpoints of our Skywork-13B model. We explore the comparative effectiveness of upcycling versus training from scratch initializations. Our findings suggest that the choice between these two approaches should consider both the performance of the existing dense checkpoints and the MoE training budget. We highlight two innov"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.06563","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/2406.06563/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":"2406.06563","created_at":"2026-07-05T08:29:47.523758+00:00"},{"alias_kind":"arxiv_version","alias_value":"2406.06563v1","created_at":"2026-07-05T08:29:47.523758+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.06563","created_at":"2026-07-05T08:29:47.523758+00:00"},{"alias_kind":"pith_short_12","alias_value":"RW5XONPRALUD","created_at":"2026-07-05T08:29:47.523758+00:00"},{"alias_kind":"pith_short_16","alias_value":"RW5XONPRALUDJDUG","created_at":"2026-07-05T08:29:47.523758+00:00"},{"alias_kind":"pith_short_8","alias_value":"RW5XONPR","created_at":"2026-07-05T08:29:47.523758+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":11,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.25821","citing_title":"SARA: Unlocking Multilingual Knowledge in Mixture-of-Experts via Semantically Anchored Routing Alignment","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2606.26287","citing_title":"GeMoE: Gating Entropy is All You Need for Uncertainty-aware Adaptive Routing in MoE-based Large Vision-Language Models","ref_index":54,"is_internal_anchor":false},{"citing_arxiv_id":"2606.01062","citing_title":"DAG-MoE: From Simple Mixture to Structural Aggregation in Mixture-of-Experts","ref_index":23,"is_internal_anchor":false},{"citing_arxiv_id":"2409.04777","citing_title":"Optimization Hyper-parameter Laws for Large Language Models","ref_index":36,"is_internal_anchor":false},{"citing_arxiv_id":"2506.12119","citing_title":"Mixture-of-Experts Can Surpass Dense LLMs Under Strictly Equal Resource","ref_index":38,"is_internal_anchor":false},{"citing_arxiv_id":"2510.08008","citing_title":"Beyond Sunk Costs: Boosting LLM Pre-training Efficiency via Orthogonal Growth of Mixture-of-Experts","ref_index":24,"is_internal_anchor":false},{"citing_arxiv_id":"2605.20891","citing_title":"HDMoE: A Hierarchical Decoupling-Fusion Mixture-of-Experts Framework for Multimodal Cancer Survival Prediction","ref_index":42,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06501","citing_title":"Cubit: Token Mixer with Kernel Ridge Regression","ref_index":84,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08292","citing_title":"Hierarchical Mixture-of-Experts with Two-Stage Optimization","ref_index":39,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06501","citing_title":"Cubit: Token Mixer with Kernel Ridge Regression","ref_index":84,"is_internal_anchor":false},{"citing_arxiv_id":"2605.04058","citing_title":"MP-ISMoE: Mixed-Precision Interactive Side Mixture-of-Experts for Efficient Transfer Learning","ref_index":157,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/RW5XONPRALUDJDUGBESKZJT2PU","json":"https://pith.science/pith/RW5XONPRALUDJDUGBESKZJT2PU.json","graph_json":"https://pith.science/api/pith-number/RW5XONPRALUDJDUGBESKZJT2PU/graph.json","events_json":"https://pith.science/api/pith-number/RW5XONPRALUDJDUGBESKZJT2PU/events.json","paper":"https://pith.science/paper/RW5XONPR"},"agent_actions":{"view_html":"https://pith.science/pith/RW5XONPRALUDJDUGBESKZJT2PU","download_json":"https://pith.science/pith/RW5XONPRALUDJDUGBESKZJT2PU.json","view_paper":"https://pith.science/paper/RW5XONPR","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2406.06563&json=true","fetch_graph":"https://pith.science/api/pith-number/RW5XONPRALUDJDUGBESKZJT2PU/graph.json","fetch_events":"https://pith.science/api/pith-number/RW5XONPRALUDJDUGBESKZJT2PU/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/RW5XONPRALUDJDUGBESKZJT2PU/action/timestamp_anchor","attest_storage":"https://pith.science/pith/RW5XONPRALUDJDUGBESKZJT2PU/action/storage_attestation","attest_author":"https://pith.science/pith/RW5XONPRALUDJDUGBESKZJT2PU/action/author_attestation","sign_citation":"https://pith.science/pith/RW5XONPRALUDJDUGBESKZJT2PU/action/citation_signature","submit_replication":"https://pith.science/pith/RW5XONPRALUDJDUGBESKZJT2PU/action/replication_record"}},"created_at":"2026-07-05T08:29:47.523758+00:00","updated_at":"2026-07-05T08:29:47.523758+00:00"}