{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2018:7IWIZNLYFUIJ4SZAZ2LQCJPTS2","short_pith_number":"pith:7IWIZNLY","schema_version":"1.0","canonical_sha256":"fa2c8cb5782d109e4b20ce970125f396854b119082a4c72c81c3c3e8ba6c042f","source":{"kind":"arxiv","id":"1811.01088","version":2},"attestation_state":"computed","paper":{"title":"Sentence Encoders on STILTs: Supplementary Training on Intermediate Labeled-data Tasks","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Jason Phang, Samuel R. Bowman, Thibault F\\'evry","submitted_at":"2018-11-02T21:04:24Z","abstract_excerpt":"Pretraining sentence encoders with language modeling and related unsupervised tasks has recently been shown to be very effective for language understanding tasks. By supplementing language model-style pretraining with further training on data-rich supervised tasks, such as natural language inference, we obtain additional performance improvements on the GLUE benchmark. Applying supplementary training on BERT (Devlin et al., 2018), we attain a GLUE score of 81.8---the state of the art (as of 02/24/2019) and a 1.4 point improvement over BERT. We also observe reduced variance across random restart"},"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":"1811.01088","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2018-11-02T21:04:24Z","cross_cats_sorted":[],"title_canon_sha256":"4ea0ab34f32f1d43029ac61d5447ffe171c23a179d6a4850c28726eda1dd5411","abstract_canon_sha256":"e18761c853c3b921e4144eb1c486183e765e763de36d04a604e6d76bdd2f3d89"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-17T23:52:30.048358Z","signature_b64":"buukDQ8/N+WhGmLG1YO9UUxhFKXfWCMotePt3KQQuCL/Zq+nzZHpIbB4ryzJxUlafaRn9AqJzy1KDH3COWSrDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"fa2c8cb5782d109e4b20ce970125f396854b119082a4c72c81c3c3e8ba6c042f","last_reissued_at":"2026-05-17T23:52:30.047917Z","signature_status":"signed_v1","first_computed_at":"2026-05-17T23:52:30.047917Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Sentence Encoders on STILTs: Supplementary Training on Intermediate Labeled-data Tasks","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Jason Phang, Samuel R. Bowman, Thibault F\\'evry","submitted_at":"2018-11-02T21:04:24Z","abstract_excerpt":"Pretraining sentence encoders with language modeling and related unsupervised tasks has recently been shown to be very effective for language understanding tasks. By supplementing language model-style pretraining with further training on data-rich supervised tasks, such as natural language inference, we obtain additional performance improvements on the GLUE benchmark. Applying supplementary training on BERT (Devlin et al., 2018), we attain a GLUE score of 81.8---the state of the art (as of 02/24/2019) and a 1.4 point improvement over BERT. We also observe reduced variance across random restart"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1811.01088","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":""},"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":"1811.01088","created_at":"2026-05-17T23:52:30.047978+00:00"},{"alias_kind":"arxiv_version","alias_value":"1811.01088v2","created_at":"2026-05-17T23:52:30.047978+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1811.01088","created_at":"2026-05-17T23:52:30.047978+00:00"},{"alias_kind":"pith_short_12","alias_value":"7IWIZNLYFUIJ","created_at":"2026-05-18T12:32:11.075285+00:00"},{"alias_kind":"pith_short_16","alias_value":"7IWIZNLYFUIJ4SZA","created_at":"2026-05-18T12:32:11.075285+00:00"},{"alias_kind":"pith_short_8","alias_value":"7IWIZNLY","created_at":"2026-05-18T12:32:11.075285+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":9,"internal_anchor_count":4,"sample":[{"citing_arxiv_id":"1907.02581","citing_title":"Transfer Learning for Risk Classification of Social Media Posts: Model Evaluation Study","ref_index":36,"is_internal_anchor":true},{"citing_arxiv_id":"2405.13071","citing_title":"A Novel Method for News Article Event-Based Embedding","ref_index":34,"is_internal_anchor":true},{"citing_arxiv_id":"2003.10555","citing_title":"ELECTRA: Pre-training Text Encoders as Discriminators Rather Than Generators","ref_index":8,"is_internal_anchor":true},{"citing_arxiv_id":"1905.00537","citing_title":"SuperGLUE: A Stickier Benchmark for General-Purpose Language Understanding Systems","ref_index":132,"is_internal_anchor":true},{"citing_arxiv_id":"1904.09728","citing_title":"SocialIQA: Commonsense Reasoning about Social Interactions","ref_index":24,"is_internal_anchor":false},{"citing_arxiv_id":"1905.10044","citing_title":"BoolQ: Exploring the Surprising Difficulty of Natural Yes/No Questions","ref_index":25,"is_internal_anchor":false},{"citing_arxiv_id":"1910.10683","citing_title":"Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer","ref_index":54,"is_internal_anchor":false},{"citing_arxiv_id":"2205.01068","citing_title":"OPT: Open Pre-trained Transformer Language Models","ref_index":73,"is_internal_anchor":false},{"citing_arxiv_id":"2005.14165","citing_title":"Language Models are Few-Shot Learners","ref_index":59,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/7IWIZNLYFUIJ4SZAZ2LQCJPTS2","json":"https://pith.science/pith/7IWIZNLYFUIJ4SZAZ2LQCJPTS2.json","graph_json":"https://pith.science/api/pith-number/7IWIZNLYFUIJ4SZAZ2LQCJPTS2/graph.json","events_json":"https://pith.science/api/pith-number/7IWIZNLYFUIJ4SZAZ2LQCJPTS2/events.json","paper":"https://pith.science/paper/7IWIZNLY"},"agent_actions":{"view_html":"https://pith.science/pith/7IWIZNLYFUIJ4SZAZ2LQCJPTS2","download_json":"https://pith.science/pith/7IWIZNLYFUIJ4SZAZ2LQCJPTS2.json","view_paper":"https://pith.science/paper/7IWIZNLY","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=1811.01088&json=true","fetch_graph":"https://pith.science/api/pith-number/7IWIZNLYFUIJ4SZAZ2LQCJPTS2/graph.json","fetch_events":"https://pith.science/api/pith-number/7IWIZNLYFUIJ4SZAZ2LQCJPTS2/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/7IWIZNLYFUIJ4SZAZ2LQCJPTS2/action/timestamp_anchor","attest_storage":"https://pith.science/pith/7IWIZNLYFUIJ4SZAZ2LQCJPTS2/action/storage_attestation","attest_author":"https://pith.science/pith/7IWIZNLYFUIJ4SZAZ2LQCJPTS2/action/author_attestation","sign_citation":"https://pith.science/pith/7IWIZNLYFUIJ4SZAZ2LQCJPTS2/action/citation_signature","submit_replication":"https://pith.science/pith/7IWIZNLYFUIJ4SZAZ2LQCJPTS2/action/replication_record"}},"created_at":"2026-05-17T23:52:30.047978+00:00","updated_at":"2026-05-17T23:52:30.047978+00:00"}