{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2019:75FMDYMD3AI75FJL4GWJMHGNOO","short_pith_number":"pith:75FMDYMD","schema_version":"1.0","canonical_sha256":"ff4ac1e183d811fe952be1ac961ccd739b5014a5916c16ee8a2821787671ff04","source":{"kind":"arxiv","id":"1910.03723","version":2},"attestation_state":"computed","paper":{"title":"Knowledge Distillation from Internal Representations","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Benjamin Yao, Chenlei Guo, Gustavo Aguilar, Xing Fan, Yuan Ling, Yu Zhang","submitted_at":"2019-10-08T23:56:27Z","abstract_excerpt":"Knowledge distillation is typically conducted by training a small model (the student) to mimic a large and cumbersome model (the teacher). The idea is to compress the knowledge from the teacher by using its output probabilities as soft-labels to optimize the student. However, when the teacher is considerably large, there is no guarantee that the internal knowledge of the teacher will be transferred into the student; even if the student closely matches the soft-labels, its internal representations may be considerably different. This internal mismatch can undermine the generalization capabilitie"},"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":"1910.03723","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2019-10-08T23:56:27Z","cross_cats_sorted":[],"title_canon_sha256":"9b0091e1354f2ad3626540dfa27324bc1a76e8bbf886030f1727bfb2d2f94596","abstract_canon_sha256":"3568bd16b68c3a7e7be5107002165f792961fb9ab03ef713709440cbd9e3c2ba"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T00:33:54.236969Z","signature_b64":"BH+t1ordH0Ja7VKxhic/lKHRm3qfW4FUCVrmuojATfEisz7nKalJvYnNykhpMdElz4H9N/GpcmE1h7MBggSEBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ff4ac1e183d811fe952be1ac961ccd739b5014a5916c16ee8a2821787671ff04","last_reissued_at":"2026-07-05T00:33:54.236485Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T00:33:54.236485Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Knowledge Distillation from Internal Representations","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Benjamin Yao, Chenlei Guo, Gustavo Aguilar, Xing Fan, Yuan Ling, Yu Zhang","submitted_at":"2019-10-08T23:56:27Z","abstract_excerpt":"Knowledge distillation is typically conducted by training a small model (the student) to mimic a large and cumbersome model (the teacher). The idea is to compress the knowledge from the teacher by using its output probabilities as soft-labels to optimize the student. However, when the teacher is considerably large, there is no guarantee that the internal knowledge of the teacher will be transferred into the student; even if the student closely matches the soft-labels, its internal representations may be considerably different. This internal mismatch can undermine the generalization capabilitie"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1910.03723","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/1910.03723/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":"1910.03723","created_at":"2026-07-05T00:33:54.236544+00:00"},{"alias_kind":"arxiv_version","alias_value":"1910.03723v2","created_at":"2026-07-05T00:33:54.236544+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1910.03723","created_at":"2026-07-05T00:33:54.236544+00:00"},{"alias_kind":"pith_short_12","alias_value":"75FMDYMD3AI7","created_at":"2026-07-05T00:33:54.236544+00:00"},{"alias_kind":"pith_short_16","alias_value":"75FMDYMD3AI75FJL","created_at":"2026-07-05T00:33:54.236544+00:00"},{"alias_kind":"pith_short_8","alias_value":"75FMDYMD","created_at":"2026-07-05T00:33:54.236544+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.07721","citing_title":"Memory-Efficient Looped Transformer: Decoupling Compute from Memory in Looped Language Models","ref_index":28,"is_internal_anchor":false},{"citing_arxiv_id":"2605.07721","citing_title":"Memory-Efficient Looped Transformer: Decoupling Compute from Memory in Looped Language Models","ref_index":28,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/75FMDYMD3AI75FJL4GWJMHGNOO","json":"https://pith.science/pith/75FMDYMD3AI75FJL4GWJMHGNOO.json","graph_json":"https://pith.science/api/pith-number/75FMDYMD3AI75FJL4GWJMHGNOO/graph.json","events_json":"https://pith.science/api/pith-number/75FMDYMD3AI75FJL4GWJMHGNOO/events.json","paper":"https://pith.science/paper/75FMDYMD"},"agent_actions":{"view_html":"https://pith.science/pith/75FMDYMD3AI75FJL4GWJMHGNOO","download_json":"https://pith.science/pith/75FMDYMD3AI75FJL4GWJMHGNOO.json","view_paper":"https://pith.science/paper/75FMDYMD","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=1910.03723&json=true","fetch_graph":"https://pith.science/api/pith-number/75FMDYMD3AI75FJL4GWJMHGNOO/graph.json","fetch_events":"https://pith.science/api/pith-number/75FMDYMD3AI75FJL4GWJMHGNOO/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/75FMDYMD3AI75FJL4GWJMHGNOO/action/timestamp_anchor","attest_storage":"https://pith.science/pith/75FMDYMD3AI75FJL4GWJMHGNOO/action/storage_attestation","attest_author":"https://pith.science/pith/75FMDYMD3AI75FJL4GWJMHGNOO/action/author_attestation","sign_citation":"https://pith.science/pith/75FMDYMD3AI75FJL4GWJMHGNOO/action/citation_signature","submit_replication":"https://pith.science/pith/75FMDYMD3AI75FJL4GWJMHGNOO/action/replication_record"}},"created_at":"2026-07-05T00:33:54.236544+00:00","updated_at":"2026-07-05T00:33:54.236544+00:00"}