{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:UKYXJ75SEHR6NQTFVRZRLULCLK","short_pith_number":"pith:UKYXJ75S","schema_version":"1.0","canonical_sha256":"a2b174ffb221e3e6c265ac7315d1625a85d5c93cfa9d63462144bc89ef9d0182","source":{"kind":"arxiv","id":"2406.13555","version":3},"attestation_state":"computed","paper":{"title":"BiLD: Bi-directional Logits Difference Loss for Large Language Model Distillation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Feng Zhou, Minchong Li, Xiaohui Song","submitted_at":"2024-06-19T13:44:56Z","abstract_excerpt":"In recent years, large language models (LLMs) have shown exceptional capabilities across various natural language processing (NLP) tasks. However, such impressive performance often comes with the trade-off of an increased parameter size, posing significant challenges for widespread deployment. Knowledge distillation (KD) provides a solution by transferring knowledge from a large teacher model to a smaller student model. In this paper, we explore the task-specific distillation of LLMs at the logit level. Our investigation reveals that the logits of fine-tuned LLMs exhibit a more extreme long-ta"},"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":"2406.13555","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-06-19T13:44:56Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"fc7e589672525d4632c03efb7968d50bff41b3a83b1dc500fb10fe6397f2cc5e","abstract_canon_sha256":"f1dd0614ec0b23d05f806c7823bd7f2c176627c9555e11df34f7b8473e7f85ff"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:15:44.183140Z","signature_b64":"HPLGywioW1HV0dlaozu3w0Glk/637BQ/ie6MfxCpCAt/l+gqpRphrWykJYsbJZAiyIdd0hxX8T/ot+ryFPkPCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a2b174ffb221e3e6c265ac7315d1625a85d5c93cfa9d63462144bc89ef9d0182","last_reissued_at":"2026-07-05T10:15:44.182498Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:15:44.182498Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"BiLD: Bi-directional Logits Difference Loss for Large Language Model Distillation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Feng Zhou, Minchong Li, Xiaohui Song","submitted_at":"2024-06-19T13:44:56Z","abstract_excerpt":"In recent years, large language models (LLMs) have shown exceptional capabilities across various natural language processing (NLP) tasks. However, such impressive performance often comes with the trade-off of an increased parameter size, posing significant challenges for widespread deployment. Knowledge distillation (KD) provides a solution by transferring knowledge from a large teacher model to a smaller student model. In this paper, we explore the task-specific distillation of LLMs at the logit level. Our investigation reveals that the logits of fine-tuned LLMs exhibit a more extreme long-ta"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.13555","kind":"arxiv","version":3},"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/2406.13555/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":"2406.13555","created_at":"2026-07-05T10:15:44.182556+00:00"},{"alias_kind":"arxiv_version","alias_value":"2406.13555v3","created_at":"2026-07-05T10:15:44.182556+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.13555","created_at":"2026-07-05T10:15:44.182556+00:00"},{"alias_kind":"pith_short_12","alias_value":"UKYXJ75SEHR6","created_at":"2026-07-05T10:15:44.182556+00:00"},{"alias_kind":"pith_short_16","alias_value":"UKYXJ75SEHR6NQTF","created_at":"2026-07-05T10:15:44.182556+00:00"},{"alias_kind":"pith_short_8","alias_value":"UKYXJ75S","created_at":"2026-07-05T10:15:44.182556+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2604.14629","citing_title":"Switch-KD: Visual-Switch Knowledge Distillation for Vision-Language Models","ref_index":26,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/UKYXJ75SEHR6NQTFVRZRLULCLK","json":"https://pith.science/pith/UKYXJ75SEHR6NQTFVRZRLULCLK.json","graph_json":"https://pith.science/api/pith-number/UKYXJ75SEHR6NQTFVRZRLULCLK/graph.json","events_json":"https://pith.science/api/pith-number/UKYXJ75SEHR6NQTFVRZRLULCLK/events.json","paper":"https://pith.science/paper/UKYXJ75S"},"agent_actions":{"view_html":"https://pith.science/pith/UKYXJ75SEHR6NQTFVRZRLULCLK","download_json":"https://pith.science/pith/UKYXJ75SEHR6NQTFVRZRLULCLK.json","view_paper":"https://pith.science/paper/UKYXJ75S","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2406.13555&json=true","fetch_graph":"https://pith.science/api/pith-number/UKYXJ75SEHR6NQTFVRZRLULCLK/graph.json","fetch_events":"https://pith.science/api/pith-number/UKYXJ75SEHR6NQTFVRZRLULCLK/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/UKYXJ75SEHR6NQTFVRZRLULCLK/action/timestamp_anchor","attest_storage":"https://pith.science/pith/UKYXJ75SEHR6NQTFVRZRLULCLK/action/storage_attestation","attest_author":"https://pith.science/pith/UKYXJ75SEHR6NQTFVRZRLULCLK/action/author_attestation","sign_citation":"https://pith.science/pith/UKYXJ75SEHR6NQTFVRZRLULCLK/action/citation_signature","submit_replication":"https://pith.science/pith/UKYXJ75SEHR6NQTFVRZRLULCLK/action/replication_record"}},"created_at":"2026-07-05T10:15:44.182556+00:00","updated_at":"2026-07-05T10:15:44.182556+00:00"}