{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2022:7E6MSA6GOP56H4R35CPJZVR5Y4","short_pith_number":"pith:7E6MSA6G","schema_version":"1.0","canonical_sha256":"f93cc903c673fbe3f23be89e9cd63dc73522ee659d1be64d9c036b6d1b663679","source":{"kind":"arxiv","id":"2201.05596","version":2},"attestation_state":"computed","paper":{"title":"DeepSpeed-MoE: Advancing Mixture-of-Experts Inference and Training to Power Next-Generation AI Scale","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.DC"],"primary_cat":"cs.LG","authors_text":"Ammar Ahmad Awan, Conglong Li, Jeff Rasley, Minjia Zhang, Reza Yazdani Aminabadi, Samyam Rajbhandari, Yuxiong He, Zhewei Yao","submitted_at":"2022-01-14T18:36:04Z","abstract_excerpt":"As the training of giant dense models hits the boundary on the availability and capability of the hardware resources today, Mixture-of-Experts (MoE) models become one of the most promising model architectures due to their significant training cost reduction compared to a quality-equivalent dense model. Its training cost saving is demonstrated from encoder-decoder models (prior works) to a 5x saving for auto-aggressive language models (this work along with parallel explorations). However, due to the much larger model size and unique architecture, how to provide fast MoE model inference remains "},"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":"2201.05596","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2022-01-14T18:36:04Z","cross_cats_sorted":["cs.AI","cs.DC"],"title_canon_sha256":"ed31414ea77fa05eac6534ba9c9c692e7f88f562db4f3d680df4bf8babfaa815","abstract_canon_sha256":"e124d610aa10f5cfc9feb7a5466a6dd31584ae449c79cdcfe578a00715466216"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T04:42:33.312739Z","signature_b64":"BJaiLzm+3qvarFgNjSOcPbiqsfDsG/ph/zyGP3Mo4o5XTIESEvQYmkUiGvCt8j/NoIrarl8/Rnx3PC/huNAkAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f93cc903c673fbe3f23be89e9cd63dc73522ee659d1be64d9c036b6d1b663679","last_reissued_at":"2026-07-05T04:42:33.312193Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T04:42:33.312193Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"DeepSpeed-MoE: Advancing Mixture-of-Experts Inference and Training to Power Next-Generation AI Scale","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.DC"],"primary_cat":"cs.LG","authors_text":"Ammar Ahmad Awan, Conglong Li, Jeff Rasley, Minjia Zhang, Reza Yazdani Aminabadi, Samyam Rajbhandari, Yuxiong He, Zhewei Yao","submitted_at":"2022-01-14T18:36:04Z","abstract_excerpt":"As the training of giant dense models hits the boundary on the availability and capability of the hardware resources today, Mixture-of-Experts (MoE) models become one of the most promising model architectures due to their significant training cost reduction compared to a quality-equivalent dense model. Its training cost saving is demonstrated from encoder-decoder models (prior works) to a 5x saving for auto-aggressive language models (this work along with parallel explorations). However, due to the much larger model size and unique architecture, how to provide fast MoE model inference remains "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2201.05596","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/2201.05596/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":"2201.05596","created_at":"2026-07-05T04:42:33.312252+00:00"},{"alias_kind":"arxiv_version","alias_value":"2201.05596v2","created_at":"2026-07-05T04:42:33.312252+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2201.05596","created_at":"2026-07-05T04:42:33.312252+00:00"},{"alias_kind":"pith_short_12","alias_value":"7E6MSA6GOP56","created_at":"2026-07-05T04:42:33.312252+00:00"},{"alias_kind":"pith_short_16","alias_value":"7E6MSA6GOP56H4R3","created_at":"2026-07-05T04:42:33.312252+00:00"},{"alias_kind":"pith_short_8","alias_value":"7E6MSA6G","created_at":"2026-07-05T04:42:33.312252+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":12,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2607.01789","citing_title":"EPnG: Adaptive Expert Prune-and-Grow for Parameter-Efficient MoE Fine-tuning","ref_index":23,"is_internal_anchor":false},{"citing_arxiv_id":"2606.07019","citing_title":"PCCL: Process Group-Aware Scalable and Generic Collective Algorithm Synthesizer","ref_index":44,"is_internal_anchor":false},{"citing_arxiv_id":"2606.00275","citing_title":"Hyperbolic and Evidence-Prioritized Experts for Large Vision-Language Models","ref_index":40,"is_internal_anchor":false},{"citing_arxiv_id":"2506.12119","citing_title":"Mixture-of-Experts Can Surpass Dense LLMs Under Strictly Equal Resource","ref_index":30,"is_internal_anchor":false},{"citing_arxiv_id":"2605.20179","citing_title":"TIDE: Efficient and Lossless MoE Diffusion LLM Inference with I/O-aware Expert Offload","ref_index":26,"is_internal_anchor":false},{"citing_arxiv_id":"2605.10670","citing_title":"Surviving Partial Rank Failures in Wide Expert-Parallel MoE Inference","ref_index":23,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08962","citing_title":"MegaScale-Omni: A Hyper-Scale, Workload-Resilient System for MultiModal LLM Training in Production","ref_index":44,"is_internal_anchor":false},{"citing_arxiv_id":"2604.23150","citing_title":"Scaling Multi-Node Mixture-of-Experts Inference Using Expert Activation Patterns","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06665","citing_title":"UniPool: A Globally Shared Expert Pool for Mixture-of-Experts","ref_index":39,"is_internal_anchor":false},{"citing_arxiv_id":"2604.07173","citing_title":"InfiniLoRA: Disaggregated Multi-LoRA Serving for Large Language Models","ref_index":30,"is_internal_anchor":false},{"citing_arxiv_id":"2605.02641","citing_title":"Mamoda2.5: Enhancing Unified Multimodal Model with DiT-MoE","ref_index":43,"is_internal_anchor":false},{"citing_arxiv_id":"2605.05049","citing_title":"Piper: Efficient Large-Scale MoE Training via Resource Modeling and Pipelined Hybrid Parallelism","ref_index":16,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/7E6MSA6GOP56H4R35CPJZVR5Y4","json":"https://pith.science/pith/7E6MSA6GOP56H4R35CPJZVR5Y4.json","graph_json":"https://pith.science/api/pith-number/7E6MSA6GOP56H4R35CPJZVR5Y4/graph.json","events_json":"https://pith.science/api/pith-number/7E6MSA6GOP56H4R35CPJZVR5Y4/events.json","paper":"https://pith.science/paper/7E6MSA6G"},"agent_actions":{"view_html":"https://pith.science/pith/7E6MSA6GOP56H4R35CPJZVR5Y4","download_json":"https://pith.science/pith/7E6MSA6GOP56H4R35CPJZVR5Y4.json","view_paper":"https://pith.science/paper/7E6MSA6G","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2201.05596&json=true","fetch_graph":"https://pith.science/api/pith-number/7E6MSA6GOP56H4R35CPJZVR5Y4/graph.json","fetch_events":"https://pith.science/api/pith-number/7E6MSA6GOP56H4R35CPJZVR5Y4/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/7E6MSA6GOP56H4R35CPJZVR5Y4/action/timestamp_anchor","attest_storage":"https://pith.science/pith/7E6MSA6GOP56H4R35CPJZVR5Y4/action/storage_attestation","attest_author":"https://pith.science/pith/7E6MSA6GOP56H4R35CPJZVR5Y4/action/author_attestation","sign_citation":"https://pith.science/pith/7E6MSA6GOP56H4R35CPJZVR5Y4/action/citation_signature","submit_replication":"https://pith.science/pith/7E6MSA6GOP56H4R35CPJZVR5Y4/action/replication_record"}},"created_at":"2026-07-05T04:42:33.312252+00:00","updated_at":"2026-07-05T04:42:33.312252+00:00"}