{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:X55NRAQWKZICPN7FTFCW4AKPRQ","short_pith_number":"pith:X55NRAQW","schema_version":"1.0","canonical_sha256":"bf7ad88216565027b7e599456e014f8c2f7fa63d8996bb6d73bf5ce1ec756ade","source":{"kind":"arxiv","id":"2112.06905","version":2},"attestation_state":"computed","paper":{"title":"GLaM: Efficient Scaling of Language Models with Mixture-of-Experts","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Adams Wei Yu, Andrew M. Dai, Barret Zoph, Claire Cui, Dmitry Lepikhin, Kathleen Meier-Hellstern, Kellie Webster, Kevin Robinson, Kun Zhang, Liam Fedus, Lucas Dixon, Maarten Bosma, Marie Pellat, Maxim Krikun, Nan Du, Orhan Firat, Quoc V Le, Simon Tong, Tao Wang, Toju Duke, Yanping Huang, Yanqi Zhou, Yonghui Wu, Yuanzhong Xu, Yu Emma Wang, Zhifeng Chen, Zongwei Zhou","submitted_at":"2021-12-13T18:58:19Z","abstract_excerpt":"Scaling language models with more data, compute and parameters has driven significant progress in natural language processing. For example, thanks to scaling, GPT-3 was able to achieve strong results on in-context learning tasks. However, training these large dense models requires significant amounts of computing resources. In this paper, we propose and develop a family of language models named GLaM (Generalist Language Model), which uses a sparsely activated mixture-of-experts architecture to scale the model capacity while also incurring substantially less training cost compared to dense vari"},"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":"2112.06905","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2021-12-13T18:58:19Z","cross_cats_sorted":[],"title_canon_sha256":"1ea822aed4319341b85fc4c057b5182f6531c8966a07e8610cd0522d7fa51b38","abstract_canon_sha256":"93a133889a92f47646ab0e1d08f1045e03f91da64b6fb4708d1419209b3809e8"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T04:45:14.339528Z","signature_b64":"GtEPL3UUUdbEygWX7fuEU0h2xDpQMIzAFP1jjHMZF8B4jzuUmyppGFGyJ1wNmfXzLA0+nM7SYxeyhjv7MvE5DQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"bf7ad88216565027b7e599456e014f8c2f7fa63d8996bb6d73bf5ce1ec756ade","last_reissued_at":"2026-07-05T04:45:14.339106Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T04:45:14.339106Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"GLaM: Efficient Scaling of Language Models with Mixture-of-Experts","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Adams Wei Yu, Andrew M. Dai, Barret Zoph, Claire Cui, Dmitry Lepikhin, Kathleen Meier-Hellstern, Kellie Webster, Kevin Robinson, Kun Zhang, Liam Fedus, Lucas Dixon, Maarten Bosma, Marie Pellat, Maxim Krikun, Nan Du, Orhan Firat, Quoc V Le, Simon Tong, Tao Wang, Toju Duke, Yanping Huang, Yanqi Zhou, Yonghui Wu, Yuanzhong Xu, Yu Emma Wang, Zhifeng Chen, Zongwei Zhou","submitted_at":"2021-12-13T18:58:19Z","abstract_excerpt":"Scaling language models with more data, compute and parameters has driven significant progress in natural language processing. For example, thanks to scaling, GPT-3 was able to achieve strong results on in-context learning tasks. However, training these large dense models requires significant amounts of computing resources. In this paper, we propose and develop a family of language models named GLaM (Generalist Language Model), which uses a sparsely activated mixture-of-experts architecture to scale the model capacity while also incurring substantially less training cost compared to dense vari"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2112.06905","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/2112.06905/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":"2112.06905","created_at":"2026-07-05T04:45:14.339162+00:00"},{"alias_kind":"arxiv_version","alias_value":"2112.06905v2","created_at":"2026-07-05T04:45:14.339162+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2112.06905","created_at":"2026-07-05T04:45:14.339162+00:00"},{"alias_kind":"pith_short_12","alias_value":"X55NRAQWKZIC","created_at":"2026-07-05T04:45:14.339162+00:00"},{"alias_kind":"pith_short_16","alias_value":"X55NRAQWKZICPN7F","created_at":"2026-07-05T04:45:14.339162+00:00"},{"alias_kind":"pith_short_8","alias_value":"X55NRAQW","created_at":"2026-07-05T04:45:14.339162+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":13,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.22541","citing_title":"ASAP: A Disaggregated and Asynchronous Inference System for MoE Prefill","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2606.00275","citing_title":"Hyperbolic and Evidence-Prioritized Experts for Large Vision-Language Models","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2507.06261","citing_title":"Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities","ref_index":20,"is_internal_anchor":false},{"citing_arxiv_id":"2207.14255","citing_title":"Efficient Training of Language Models to Fill in the Middle","ref_index":107,"is_internal_anchor":false},{"citing_arxiv_id":"2605.11800","citing_title":"ROMER: Expert Replacement and Router Calibration for Robust MoE LLMs on Analog Compute-in-Memory Systems","ref_index":40,"is_internal_anchor":false},{"citing_arxiv_id":"2309.14509","citing_title":"DeepSpeed Ulysses: System Optimizations for Enabling Training of Extreme Long Sequence Transformer Models","ref_index":169,"is_internal_anchor":false},{"citing_arxiv_id":"2305.10403","citing_title":"PaLM 2 Technical Report","ref_index":41,"is_internal_anchor":false},{"citing_arxiv_id":"2206.10789","citing_title":"Scaling Autoregressive Models for Content-Rich Text-to-Image Generation","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2604.19835","citing_title":"Expert Upcycling: Shifting the Compute-Efficient Frontier of Mixture-of-Experts","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2206.07682","citing_title":"Emergent Abilities of Large Language Models","ref_index":24,"is_internal_anchor":false},{"citing_arxiv_id":"2204.02311","citing_title":"PaLM: Scaling Language Modeling with Pathways","ref_index":39,"is_internal_anchor":false},{"citing_arxiv_id":"2604.19835","citing_title":"Expert Upcycling: Shifting the Compute-Efficient Frontier of Mixture-of-Experts","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2605.02300","citing_title":"A Meta Reinforcement Learning Approach to Goals-Based Wealth Management","ref_index":262,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/X55NRAQWKZICPN7FTFCW4AKPRQ","json":"https://pith.science/pith/X55NRAQWKZICPN7FTFCW4AKPRQ.json","graph_json":"https://pith.science/api/pith-number/X55NRAQWKZICPN7FTFCW4AKPRQ/graph.json","events_json":"https://pith.science/api/pith-number/X55NRAQWKZICPN7FTFCW4AKPRQ/events.json","paper":"https://pith.science/paper/X55NRAQW"},"agent_actions":{"view_html":"https://pith.science/pith/X55NRAQWKZICPN7FTFCW4AKPRQ","download_json":"https://pith.science/pith/X55NRAQWKZICPN7FTFCW4AKPRQ.json","view_paper":"https://pith.science/paper/X55NRAQW","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2112.06905&json=true","fetch_graph":"https://pith.science/api/pith-number/X55NRAQWKZICPN7FTFCW4AKPRQ/graph.json","fetch_events":"https://pith.science/api/pith-number/X55NRAQWKZICPN7FTFCW4AKPRQ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/X55NRAQWKZICPN7FTFCW4AKPRQ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/X55NRAQWKZICPN7FTFCW4AKPRQ/action/storage_attestation","attest_author":"https://pith.science/pith/X55NRAQWKZICPN7FTFCW4AKPRQ/action/author_attestation","sign_citation":"https://pith.science/pith/X55NRAQWKZICPN7FTFCW4AKPRQ/action/citation_signature","submit_replication":"https://pith.science/pith/X55NRAQWKZICPN7FTFCW4AKPRQ/action/replication_record"}},"created_at":"2026-07-05T04:45:14.339162+00:00","updated_at":"2026-07-05T04:45:14.339162+00:00"}