{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2020:GXGTFHA647HRAR4TS65COGCHWA","short_pith_number":"pith:GXGTFHA6","schema_version":"1.0","canonical_sha256":"35cd329c1ee7cf10479397ba271847b031076cb2242220b426713fd98ea01c62","source":{"kind":"arxiv","id":"2012.15355","version":4},"attestation_state":"computed","paper":{"title":"Optimizing Deeper Transformers on Small Datasets","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Chenyang Huang, Dhruv Kumar, Jackie Chi Kit Cheung, Keyi Tang, Peng Xu, Simon J.D. Prince, Wei Yang, Wenjie Zi, Yanshuai Cao","submitted_at":"2020-12-30T22:53:49Z","abstract_excerpt":"It is a common belief that training deep transformers from scratch requires large datasets. Consequently, for small datasets, people usually use shallow and simple additional layers on top of pre-trained models during fine-tuning. This work shows that this does not always need to be the case: with proper initialization and optimization, the benefits of very deep transformers can carry over to challenging tasks with small datasets, including Text-to-SQL semantic parsing and logical reading comprehension. In particular, we successfully train $48$ layers of transformers, comprising $24$ fine-tune"},"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":"2012.15355","kind":"arxiv","version":4},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CL","submitted_at":"2020-12-30T22:53:49Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"27bed61c118b9e57faa06e26bea18db4417eb4653bf17e0c36526ef3557a6e26","abstract_canon_sha256":"ec847c4ee2b40085ffbd8a7cf28eab23958875d2d62164d5ad9efa73e9c734ef"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T02:44:35.071150Z","signature_b64":"YhvAbEWrOHKK6dJoqor3IpnT+ZFD2tNl9IujTSKHGYwhBiy9RIW+XVcbwaJTk6VIUa5keCqyOr+cItNJ3XJdDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"35cd329c1ee7cf10479397ba271847b031076cb2242220b426713fd98ea01c62","last_reissued_at":"2026-07-05T02:44:35.070669Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T02:44:35.070669Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Optimizing Deeper Transformers on Small Datasets","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Chenyang Huang, Dhruv Kumar, Jackie Chi Kit Cheung, Keyi Tang, Peng Xu, Simon J.D. Prince, Wei Yang, Wenjie Zi, Yanshuai Cao","submitted_at":"2020-12-30T22:53:49Z","abstract_excerpt":"It is a common belief that training deep transformers from scratch requires large datasets. Consequently, for small datasets, people usually use shallow and simple additional layers on top of pre-trained models during fine-tuning. This work shows that this does not always need to be the case: with proper initialization and optimization, the benefits of very deep transformers can carry over to challenging tasks with small datasets, including Text-to-SQL semantic parsing and logical reading comprehension. In particular, we successfully train $48$ layers of transformers, comprising $24$ fine-tune"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2012.15355","kind":"arxiv","version":4},"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/2012.15355/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":"2012.15355","created_at":"2026-07-05T02:44:35.070726+00:00"},{"alias_kind":"arxiv_version","alias_value":"2012.15355v4","created_at":"2026-07-05T02:44:35.070726+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2012.15355","created_at":"2026-07-05T02:44:35.070726+00:00"},{"alias_kind":"pith_short_12","alias_value":"GXGTFHA647HR","created_at":"2026-07-05T02:44:35.070726+00:00"},{"alias_kind":"pith_short_16","alias_value":"GXGTFHA647HRAR4T","created_at":"2026-07-05T02:44:35.070726+00:00"},{"alias_kind":"pith_short_8","alias_value":"GXGTFHA6","created_at":"2026-07-05T02:44:35.070726+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2306.14048","citing_title":"H$_2$O: Heavy-Hitter Oracle for Efficient Generative Inference of Large Language Models","ref_index":87,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/GXGTFHA647HRAR4TS65COGCHWA","json":"https://pith.science/pith/GXGTFHA647HRAR4TS65COGCHWA.json","graph_json":"https://pith.science/api/pith-number/GXGTFHA647HRAR4TS65COGCHWA/graph.json","events_json":"https://pith.science/api/pith-number/GXGTFHA647HRAR4TS65COGCHWA/events.json","paper":"https://pith.science/paper/GXGTFHA6"},"agent_actions":{"view_html":"https://pith.science/pith/GXGTFHA647HRAR4TS65COGCHWA","download_json":"https://pith.science/pith/GXGTFHA647HRAR4TS65COGCHWA.json","view_paper":"https://pith.science/paper/GXGTFHA6","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2012.15355&json=true","fetch_graph":"https://pith.science/api/pith-number/GXGTFHA647HRAR4TS65COGCHWA/graph.json","fetch_events":"https://pith.science/api/pith-number/GXGTFHA647HRAR4TS65COGCHWA/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/GXGTFHA647HRAR4TS65COGCHWA/action/timestamp_anchor","attest_storage":"https://pith.science/pith/GXGTFHA647HRAR4TS65COGCHWA/action/storage_attestation","attest_author":"https://pith.science/pith/GXGTFHA647HRAR4TS65COGCHWA/action/author_attestation","sign_citation":"https://pith.science/pith/GXGTFHA647HRAR4TS65COGCHWA/action/citation_signature","submit_replication":"https://pith.science/pith/GXGTFHA647HRAR4TS65COGCHWA/action/replication_record"}},"created_at":"2026-07-05T02:44:35.070726+00:00","updated_at":"2026-07-05T02:44:35.070726+00:00"}