{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:GS2EYIIBAILHC77E2CK2GCKWZG","short_pith_number":"pith:GS2EYIIB","schema_version":"1.0","canonical_sha256":"34b44c21010216717fe4d095a30956c991ff34a69844df3248933096d868dc0a","source":{"kind":"arxiv","id":"2306.07179","version":2},"attestation_state":"computed","paper":{"title":"Benchmarking Neural Network Training Algorithms","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Abel L. Peirson, Ankush Garg, Bilal Khan, Chandramouli Shama Sastry, Chris J. Maddison, Daniel Snider, Daniel Suo, Ehsan Amid, Frank Schneider, George E. Dahl, Juhan Bae, Justin Gilmer, Kongtao Chen, Michal Badura, Mike Rabbat, Naman Agarwal, Peter Mattson, Philipp Hennig, Priya Kasimbeg, Rakshith Vasudev, Rohan Anil, Runa Eschenhagen, Shankar Krishnan, Sourabh Medapati, Zachary Nado","submitted_at":"2023-06-12T15:21:02Z","abstract_excerpt":"Training algorithms, broadly construed, are an essential part of every deep learning pipeline. Training algorithm improvements that speed up training across a wide variety of workloads (e.g., better update rules, tuning protocols, learning rate schedules, or data selection schemes) could save time, save computational resources, and lead to better, more accurate, models. Unfortunately, as a community, we are currently unable to reliably identify training algorithm improvements, or even determine the state-of-the-art training algorithm. In this work, using concrete experiments, we argue that rea"},"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":"2306.07179","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2023-06-12T15:21:02Z","cross_cats_sorted":["stat.ML"],"title_canon_sha256":"736857e0dd5be377676b1cf5f0da9b1640ec791ef84d8906868d0289e4d65195","abstract_canon_sha256":"31cdbe592db1ff8c107f12d90443919b98274e7fb7d99f8891c3c2c4309d5666"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:23:13.980357Z","signature_b64":"16F3Lr0ZEXHYV6lnRv2Yt0M5tRLPAzaTmXdg9WXr86DPnc0NEl3+DRAWenxI/Pc0fkAnNhPhPCaup67TgISgCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"34b44c21010216717fe4d095a30956c991ff34a69844df3248933096d868dc0a","last_reissued_at":"2026-07-05T11:23:13.979815Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:23:13.979815Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Benchmarking Neural Network Training Algorithms","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Abel L. Peirson, Ankush Garg, Bilal Khan, Chandramouli Shama Sastry, Chris J. Maddison, Daniel Snider, Daniel Suo, Ehsan Amid, Frank Schneider, George E. Dahl, Juhan Bae, Justin Gilmer, Kongtao Chen, Michal Badura, Mike Rabbat, Naman Agarwal, Peter Mattson, Philipp Hennig, Priya Kasimbeg, Rakshith Vasudev, Rohan Anil, Runa Eschenhagen, Shankar Krishnan, Sourabh Medapati, Zachary Nado","submitted_at":"2023-06-12T15:21:02Z","abstract_excerpt":"Training algorithms, broadly construed, are an essential part of every deep learning pipeline. Training algorithm improvements that speed up training across a wide variety of workloads (e.g., better update rules, tuning protocols, learning rate schedules, or data selection schemes) could save time, save computational resources, and lead to better, more accurate, models. Unfortunately, as a community, we are currently unable to reliably identify training algorithm improvements, or even determine the state-of-the-art training algorithm. In this work, using concrete experiments, we argue that rea"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2306.07179","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/2306.07179/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":"2306.07179","created_at":"2026-07-05T11:23:13.979877+00:00"},{"alias_kind":"arxiv_version","alias_value":"2306.07179v2","created_at":"2026-07-05T11:23:13.979877+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2306.07179","created_at":"2026-07-05T11:23:13.979877+00:00"},{"alias_kind":"pith_short_12","alias_value":"GS2EYIIBAILH","created_at":"2026-07-05T11:23:13.979877+00:00"},{"alias_kind":"pith_short_16","alias_value":"GS2EYIIBAILHC77E","created_at":"2026-07-05T11:23:13.979877+00:00"},{"alias_kind":"pith_short_8","alias_value":"GS2EYIIB","created_at":"2026-07-05T11:23:13.979877+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":9,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.02365","citing_title":"FOAM: Frequency and Operator Error-Based Adaptive Damping Method for Reducing Staleness-Oriented Error for Shampoo","ref_index":30,"is_internal_anchor":false},{"citing_arxiv_id":"2605.18106","citing_title":"Symmetry-Compatible Principle for Optimizer Design: Embeddings, LM Heads, SwiGLU MLPs, and MoE Routers","ref_index":30,"is_internal_anchor":false},{"citing_arxiv_id":"2605.19811","citing_title":"LionMuon: Alternating Spectral and Sign Descent for Efficient Training","ref_index":2,"is_internal_anchor":false},{"citing_arxiv_id":"2605.22432","citing_title":"AMUSE: Anytime Muon with Stable Gradient Evaluation","ref_index":54,"is_internal_anchor":false},{"citing_arxiv_id":"2502.07529","citing_title":"Training Deep Learning Models with Norm-Constrained LMOs","ref_index":171,"is_internal_anchor":false},{"citing_arxiv_id":"2605.18106","citing_title":"Symmetry-Compatible Principle for Optimizer Design: Embeddings, LM Heads, SwiGLU MLPs, and MoE Routers","ref_index":29,"is_internal_anchor":false},{"citing_arxiv_id":"2605.19811","citing_title":"LionMuon: Alternating Spectral and Sign Descent for Efficient Training","ref_index":2,"is_internal_anchor":false},{"citing_arxiv_id":"2505.20628","citing_title":"Position: Adopt Constraints Over Fixed Penalties in Deep Learning","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2409.20325","citing_title":"Old Optimizer, New Norm: An Anthology","ref_index":12,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/GS2EYIIBAILHC77E2CK2GCKWZG","json":"https://pith.science/pith/GS2EYIIBAILHC77E2CK2GCKWZG.json","graph_json":"https://pith.science/api/pith-number/GS2EYIIBAILHC77E2CK2GCKWZG/graph.json","events_json":"https://pith.science/api/pith-number/GS2EYIIBAILHC77E2CK2GCKWZG/events.json","paper":"https://pith.science/paper/GS2EYIIB"},"agent_actions":{"view_html":"https://pith.science/pith/GS2EYIIBAILHC77E2CK2GCKWZG","download_json":"https://pith.science/pith/GS2EYIIBAILHC77E2CK2GCKWZG.json","view_paper":"https://pith.science/paper/GS2EYIIB","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2306.07179&json=true","fetch_graph":"https://pith.science/api/pith-number/GS2EYIIBAILHC77E2CK2GCKWZG/graph.json","fetch_events":"https://pith.science/api/pith-number/GS2EYIIBAILHC77E2CK2GCKWZG/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/GS2EYIIBAILHC77E2CK2GCKWZG/action/timestamp_anchor","attest_storage":"https://pith.science/pith/GS2EYIIBAILHC77E2CK2GCKWZG/action/storage_attestation","attest_author":"https://pith.science/pith/GS2EYIIBAILHC77E2CK2GCKWZG/action/author_attestation","sign_citation":"https://pith.science/pith/GS2EYIIBAILHC77E2CK2GCKWZG/action/citation_signature","submit_replication":"https://pith.science/pith/GS2EYIIBAILHC77E2CK2GCKWZG/action/replication_record"}},"created_at":"2026-07-05T11:23:13.979877+00:00","updated_at":"2026-07-05T11:23:13.979877+00:00"}