{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2019:T7IKL3GLAXLBJOYXKZMF3MDWHY","short_pith_number":"pith:T7IKL3GL","schema_version":"1.0","canonical_sha256":"9fd0a5eccb05d614bb1756585db0763e3a6c805b3115a552673dc43cafa8d0c7","source":{"kind":"arxiv","id":"1907.04829","version":1},"attestation_state":"computed","paper":{"title":"BAM! Born-Again Multi-Task Networks for Natural Language Understanding","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Christopher D. Manning, Kevin Clark, Minh-Thang Luong, Quoc V. Le, Urvashi Khandelwal","submitted_at":"2019-07-10T17:14:47Z","abstract_excerpt":"It can be challenging to train multi-task neural networks that outperform or even match their single-task counterparts. To help address this, we propose using knowledge distillation where single-task models teach a multi-task model. We enhance this training with teacher annealing, a novel method that gradually transitions the model from distillation to supervised learning, helping the multi-task model surpass its single-task teachers. We evaluate our approach by multi-task fine-tuning BERT on the GLUE benchmark. Our method consistently improves over standard single-task and multi-task training"},"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":"1907.04829","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2019-07-10T17:14:47Z","cross_cats_sorted":[],"title_canon_sha256":"62e94252249c83973ddb886d192787ba0c94db747de22add975161a39a53b7f9","abstract_canon_sha256":"5caf51a454972cc2c521357ef1c8c17c63cac92de4edea6257949f2243781e41"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-17T23:40:56.570042Z","signature_b64":"PrPIx31YPJbbX1T7pm7XI4w291IYXpTGrw7lAYngoX/IroRog270UmyYpH0bxY99rM0j2COKTX22m0rlK5x+Dw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"9fd0a5eccb05d614bb1756585db0763e3a6c805b3115a552673dc43cafa8d0c7","last_reissued_at":"2026-05-17T23:40:56.569407Z","signature_status":"signed_v1","first_computed_at":"2026-05-17T23:40:56.569407Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"BAM! Born-Again Multi-Task Networks for Natural Language Understanding","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Christopher D. Manning, Kevin Clark, Minh-Thang Luong, Quoc V. Le, Urvashi Khandelwal","submitted_at":"2019-07-10T17:14:47Z","abstract_excerpt":"It can be challenging to train multi-task neural networks that outperform or even match their single-task counterparts. To help address this, we propose using knowledge distillation where single-task models teach a multi-task model. We enhance this training with teacher annealing, a novel method that gradually transitions the model from distillation to supervised learning, helping the multi-task model surpass its single-task teachers. We evaluate our approach by multi-task fine-tuning BERT on the GLUE benchmark. Our method consistently improves over standard single-task and multi-task training"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1907.04829","kind":"arxiv","version":1},"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":"1907.04829","created_at":"2026-05-17T23:40:56.569531+00:00"},{"alias_kind":"arxiv_version","alias_value":"1907.04829v1","created_at":"2026-05-17T23:40:56.569531+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1907.04829","created_at":"2026-05-17T23:40:56.569531+00:00"},{"alias_kind":"pith_short_12","alias_value":"T7IKL3GLAXLB","created_at":"2026-05-18T12:33:27.125529+00:00"},{"alias_kind":"pith_short_16","alias_value":"T7IKL3GLAXLBJOYX","created_at":"2026-05-18T12:33:27.125529+00:00"},{"alias_kind":"pith_short_8","alias_value":"T7IKL3GL","created_at":"2026-05-18T12:33:27.125529+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"1905.00537","citing_title":"SuperGLUE: A Stickier Benchmark for General-Purpose Language Understanding Systems","ref_index":92,"is_internal_anchor":true},{"citing_arxiv_id":"1909.11942","citing_title":"ALBERT: A Lite BERT for Self-supervised Learning of Language Representations","ref_index":5,"is_internal_anchor":false},{"citing_arxiv_id":"2205.01068","citing_title":"OPT: Open Pre-trained Transformer Language Models","ref_index":14,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/T7IKL3GLAXLBJOYXKZMF3MDWHY","json":"https://pith.science/pith/T7IKL3GLAXLBJOYXKZMF3MDWHY.json","graph_json":"https://pith.science/api/pith-number/T7IKL3GLAXLBJOYXKZMF3MDWHY/graph.json","events_json":"https://pith.science/api/pith-number/T7IKL3GLAXLBJOYXKZMF3MDWHY/events.json","paper":"https://pith.science/paper/T7IKL3GL"},"agent_actions":{"view_html":"https://pith.science/pith/T7IKL3GLAXLBJOYXKZMF3MDWHY","download_json":"https://pith.science/pith/T7IKL3GLAXLBJOYXKZMF3MDWHY.json","view_paper":"https://pith.science/paper/T7IKL3GL","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=1907.04829&json=true","fetch_graph":"https://pith.science/api/pith-number/T7IKL3GLAXLBJOYXKZMF3MDWHY/graph.json","fetch_events":"https://pith.science/api/pith-number/T7IKL3GLAXLBJOYXKZMF3MDWHY/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/T7IKL3GLAXLBJOYXKZMF3MDWHY/action/timestamp_anchor","attest_storage":"https://pith.science/pith/T7IKL3GLAXLBJOYXKZMF3MDWHY/action/storage_attestation","attest_author":"https://pith.science/pith/T7IKL3GLAXLBJOYXKZMF3MDWHY/action/author_attestation","sign_citation":"https://pith.science/pith/T7IKL3GLAXLBJOYXKZMF3MDWHY/action/citation_signature","submit_replication":"https://pith.science/pith/T7IKL3GLAXLBJOYXKZMF3MDWHY/action/replication_record"}},"created_at":"2026-05-17T23:40:56.569531+00:00","updated_at":"2026-05-17T23:40:56.569531+00:00"}