{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2019:42SMMQ3VPN2OUCQMAWJZU2PIHQ","short_pith_number":"pith:42SMMQ3V","schema_version":"1.0","canonical_sha256":"e6a4c643757b74ea0a0c05939a69e83c275da56d4ac1d4fe20f5c0ef330e25e9","source":{"kind":"arxiv","id":"1904.09482","version":1},"attestation_state":"computed","paper":{"title":"Improving Multi-Task Deep Neural Networks via Knowledge Distillation for Natural Language Understanding","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Jianfeng Gao, Pengcheng He, Weizhu Chen, XiaoDong Liu","submitted_at":"2019-04-20T19:11:00Z","abstract_excerpt":"This paper explores the use of knowledge distillation to improve a Multi-Task Deep Neural Network (MT-DNN) (Liu et al., 2019) for learning text representations across multiple natural language understanding tasks. Although ensemble learning can improve model performance, serving an ensemble of large DNNs such as MT-DNN can be prohibitively expensive. Here we apply the knowledge distillation method (Hinton et al., 2015) in the multi-task learning setting. For each task, we train an ensemble of different MT-DNNs (teacher) that outperforms any single model, and then train a single MT-DNN (student"},"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":"1904.09482","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2019-04-20T19:11:00Z","cross_cats_sorted":[],"title_canon_sha256":"6a9e9a251acf9880b99aeee6057d8c24e42681c6fe02e8ca600eb3704f408a20","abstract_canon_sha256":"46ee743439aec50b97176b5f1a809d50c1198bc77e427ce4cabe314287ec497b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-17T23:48:04.092162Z","signature_b64":"Ip1/0F1GNezKvinCawTs6LZN97m8paHxBW0tPfdLXSFbfdiC7F49y9TS76gKRX+b1UaHBhNEYVZu3Tc9B3hTCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e6a4c643757b74ea0a0c05939a69e83c275da56d4ac1d4fe20f5c0ef330e25e9","last_reissued_at":"2026-05-17T23:48:04.091650Z","signature_status":"signed_v1","first_computed_at":"2026-05-17T23:48:04.091650Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Improving Multi-Task Deep Neural Networks via Knowledge Distillation for Natural Language Understanding","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Jianfeng Gao, Pengcheng He, Weizhu Chen, XiaoDong Liu","submitted_at":"2019-04-20T19:11:00Z","abstract_excerpt":"This paper explores the use of knowledge distillation to improve a Multi-Task Deep Neural Network (MT-DNN) (Liu et al., 2019) for learning text representations across multiple natural language understanding tasks. Although ensemble learning can improve model performance, serving an ensemble of large DNNs such as MT-DNN can be prohibitively expensive. Here we apply the knowledge distillation method (Hinton et al., 2015) in the multi-task learning setting. For each task, we train an ensemble of different MT-DNNs (teacher) that outperforms any single model, and then train a single MT-DNN (student"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1904.09482","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":"1904.09482","created_at":"2026-05-17T23:48:04.091737+00:00"},{"alias_kind":"arxiv_version","alias_value":"1904.09482v1","created_at":"2026-05-17T23:48:04.091737+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1904.09482","created_at":"2026-05-17T23:48:04.091737+00:00"},{"alias_kind":"pith_short_12","alias_value":"42SMMQ3VPN2O","created_at":"2026-05-18T12:33:10.108867+00:00"},{"alias_kind":"pith_short_16","alias_value":"42SMMQ3VPN2OUCQM","created_at":"2026-05-18T12:33:10.108867+00:00"},{"alias_kind":"pith_short_8","alias_value":"42SMMQ3V","created_at":"2026-05-18T12:33:10.108867+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":2,"sample":[{"citing_arxiv_id":"2606.30775","citing_title":"A Single Rewrite Suffices: Empirical Lessons from Production Skill Description Optimization","ref_index":75,"is_internal_anchor":true},{"citing_arxiv_id":"1905.00537","citing_title":"SuperGLUE: A Stickier Benchmark for General-Purpose Language Understanding Systems","ref_index":119,"is_internal_anchor":true},{"citing_arxiv_id":"2205.01068","citing_title":"OPT: Open Pre-trained Transformer Language Models","ref_index":71,"is_internal_anchor":false},{"citing_arxiv_id":"2005.14165","citing_title":"Language Models are Few-Shot Learners","ref_index":39,"is_internal_anchor":false},{"citing_arxiv_id":"1907.11692","citing_title":"RoBERTa: A Robustly Optimized BERT Pretraining Approach","ref_index":25,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/42SMMQ3VPN2OUCQMAWJZU2PIHQ","json":"https://pith.science/pith/42SMMQ3VPN2OUCQMAWJZU2PIHQ.json","graph_json":"https://pith.science/api/pith-number/42SMMQ3VPN2OUCQMAWJZU2PIHQ/graph.json","events_json":"https://pith.science/api/pith-number/42SMMQ3VPN2OUCQMAWJZU2PIHQ/events.json","paper":"https://pith.science/paper/42SMMQ3V"},"agent_actions":{"view_html":"https://pith.science/pith/42SMMQ3VPN2OUCQMAWJZU2PIHQ","download_json":"https://pith.science/pith/42SMMQ3VPN2OUCQMAWJZU2PIHQ.json","view_paper":"https://pith.science/paper/42SMMQ3V","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=1904.09482&json=true","fetch_graph":"https://pith.science/api/pith-number/42SMMQ3VPN2OUCQMAWJZU2PIHQ/graph.json","fetch_events":"https://pith.science/api/pith-number/42SMMQ3VPN2OUCQMAWJZU2PIHQ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/42SMMQ3VPN2OUCQMAWJZU2PIHQ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/42SMMQ3VPN2OUCQMAWJZU2PIHQ/action/storage_attestation","attest_author":"https://pith.science/pith/42SMMQ3VPN2OUCQMAWJZU2PIHQ/action/author_attestation","sign_citation":"https://pith.science/pith/42SMMQ3VPN2OUCQMAWJZU2PIHQ/action/citation_signature","submit_replication":"https://pith.science/pith/42SMMQ3VPN2OUCQMAWJZU2PIHQ/action/replication_record"}},"created_at":"2026-05-17T23:48:04.091737+00:00","updated_at":"2026-05-17T23:48:04.091737+00:00"}