{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:PFUDG73R3WYXK47GXVKCGSEMBW","short_pith_number":"pith:PFUDG73R","schema_version":"1.0","canonical_sha256":"7968337f71ddb17573e6bd5423488c0db65900e368dc889d0a9c41be8d370c77","source":{"kind":"arxiv","id":"2311.08105","version":3},"attestation_state":"computed","paper":{"title":"DiLoCo: Distributed Low-Communication Training of Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Adhiguna Kuncoro, Andrei A. Rusu, Arthur Douillard, Arthur Szlam, Jiajun Shen, Marc'Aurelio Ranzato, Qixuan Feng, Rachita Chhaparia, Yani Donchev","submitted_at":"2023-11-14T12:05:45Z","abstract_excerpt":"Large language models (LLM) have become a critical component in many applications of machine learning. However, standard approaches to training LLM require a large number of tightly interconnected accelerators, with devices exchanging gradients and other intermediate states at each optimization step. While it is difficult to build and maintain a single computing cluster hosting many accelerators, it might be easier to find several computing clusters each hosting a smaller number of devices. In this work, we propose a distributed optimization algorithm, Distributed Low-Communication (DiLoCo), t"},"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":"2311.08105","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2023-11-14T12:05:45Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"74992aa76a4a3372071cffecde14de83dfaa3fd8512416edf4942b81b58e1914","abstract_canon_sha256":"58c3a7914493e4661672f97e3bb4e93ababbe6455be76af9c8c0fd402cdc0697"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:09:54.367867Z","signature_b64":"j2M8Rz9NCMNp9he5PVODN9MZpiDag5qbWU+dMB/pZmKcJ57QOok+CUkKxz9ez5TIXH0tH/cppgSuNEbLKageAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"7968337f71ddb17573e6bd5423488c0db65900e368dc889d0a9c41be8d370c77","last_reissued_at":"2026-07-05T09:09:54.367288Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:09:54.367288Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"DiLoCo: Distributed Low-Communication Training of Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Adhiguna Kuncoro, Andrei A. Rusu, Arthur Douillard, Arthur Szlam, Jiajun Shen, Marc'Aurelio Ranzato, Qixuan Feng, Rachita Chhaparia, Yani Donchev","submitted_at":"2023-11-14T12:05:45Z","abstract_excerpt":"Large language models (LLM) have become a critical component in many applications of machine learning. However, standard approaches to training LLM require a large number of tightly interconnected accelerators, with devices exchanging gradients and other intermediate states at each optimization step. While it is difficult to build and maintain a single computing cluster hosting many accelerators, it might be easier to find several computing clusters each hosting a smaller number of devices. In this work, we propose a distributed optimization algorithm, Distributed Low-Communication (DiLoCo), t"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2311.08105","kind":"arxiv","version":3},"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/2311.08105/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":"2311.08105","created_at":"2026-07-05T09:09:54.367360+00:00"},{"alias_kind":"arxiv_version","alias_value":"2311.08105v3","created_at":"2026-07-05T09:09:54.367360+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2311.08105","created_at":"2026-07-05T09:09:54.367360+00:00"},{"alias_kind":"pith_short_12","alias_value":"PFUDG73R3WYX","created_at":"2026-07-05T09:09:54.367360+00:00"},{"alias_kind":"pith_short_16","alias_value":"PFUDG73R3WYXK47G","created_at":"2026-07-05T09:09:54.367360+00:00"},{"alias_kind":"pith_short_8","alias_value":"PFUDG73R","created_at":"2026-07-05T09:09:54.367360+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":16,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.24942","citing_title":"Quantum-Resilient Decentralized AI Economies: Proof-of-Useful-Work and Post-Quantum Security","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2606.11081","citing_title":"Unifying Local Communications and Local Updates for LLM Pretraining","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2606.05484","citing_title":"Learned Subspace Compression for Communication-Efficient Pipeline Parallelism","ref_index":12,"is_internal_anchor":false},{"citing_arxiv_id":"2606.02958","citing_title":"Echelon: Auditable Aggregate-Only Language-Model Adaptation Across Privacy Boundaries","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2606.01128","citing_title":"Local MixVR: Breaking the Communication-Sample Dependence in Distributed Learning","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2605.24326","citing_title":"ScaleAcross Explorer: Exploring Communication Optimization for Scale-Across AI Model Training","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2605.28585","citing_title":"Outer-Momentum Restarting in High-Dimensional Two-Phase Optimization","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2606.19025","citing_title":"FoMoE: Breaking the Full-Replica Barrier with a Federation of MoEs","ref_index":97,"is_internal_anchor":false},{"citing_arxiv_id":"2605.19095","citing_title":"ScheduleFree+: Scaling Learning-Rate-Free & Schedule-Free Learning to Large Language Models","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2507.06542","citing_title":"On the Surprising Effectiveness of a Single Global Merging in Decentralized Learning","ref_index":16,"is_internal_anchor":false},{"citing_arxiv_id":"2601.21972","citing_title":"Learning Decentralized LLM Collaboration with Multi-Agent Actor Critic","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2605.11172","citing_title":"Optimistic Dual Averaging Unifies Modern Optimizers","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08871","citing_title":"Rennala MVR: Improved Time Complexity for Parallel Stochastic Optimization via Momentum-Based Variance Reduction","ref_index":48,"is_internal_anchor":false},{"citing_arxiv_id":"2605.09126","citing_title":"Cosine-Gated Adam-Decay: Drop-In Staleness-Aware Outer Optimization for Decoupled DiLoCo","ref_index":2,"is_internal_anchor":false},{"citing_arxiv_id":"2604.24708","citing_title":"Scalable Hyperparameter-Divergent Ensemble Training with Automatic Learning Rate Exploration for Large Models","ref_index":5,"is_internal_anchor":false},{"citing_arxiv_id":"2604.21072","citing_title":"Distributed Generative Inference of LLM at Internet Scales with Multi-Dimensional Communication Optimization","ref_index":10,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/PFUDG73R3WYXK47GXVKCGSEMBW","json":"https://pith.science/pith/PFUDG73R3WYXK47GXVKCGSEMBW.json","graph_json":"https://pith.science/api/pith-number/PFUDG73R3WYXK47GXVKCGSEMBW/graph.json","events_json":"https://pith.science/api/pith-number/PFUDG73R3WYXK47GXVKCGSEMBW/events.json","paper":"https://pith.science/paper/PFUDG73R"},"agent_actions":{"view_html":"https://pith.science/pith/PFUDG73R3WYXK47GXVKCGSEMBW","download_json":"https://pith.science/pith/PFUDG73R3WYXK47GXVKCGSEMBW.json","view_paper":"https://pith.science/paper/PFUDG73R","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2311.08105&json=true","fetch_graph":"https://pith.science/api/pith-number/PFUDG73R3WYXK47GXVKCGSEMBW/graph.json","fetch_events":"https://pith.science/api/pith-number/PFUDG73R3WYXK47GXVKCGSEMBW/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/PFUDG73R3WYXK47GXVKCGSEMBW/action/timestamp_anchor","attest_storage":"https://pith.science/pith/PFUDG73R3WYXK47GXVKCGSEMBW/action/storage_attestation","attest_author":"https://pith.science/pith/PFUDG73R3WYXK47GXVKCGSEMBW/action/author_attestation","sign_citation":"https://pith.science/pith/PFUDG73R3WYXK47GXVKCGSEMBW/action/citation_signature","submit_replication":"https://pith.science/pith/PFUDG73R3WYXK47GXVKCGSEMBW/action/replication_record"}},"created_at":"2026-07-05T09:09:54.367360+00:00","updated_at":"2026-07-05T09:09:54.367360+00:00"}