{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:WNZ4XCELIVTJNPVEVMHHBH34ZK","short_pith_number":"pith:WNZ4XCEL","schema_version":"1.0","canonical_sha256":"b373cb888b456696bea4ab0e709f7cca8ea3640a5c1db5106f65784538e53252","source":{"kind":"arxiv","id":"2402.03898","version":2},"attestation_state":"computed","paper":{"title":"DistiLLM: Towards Streamlined Distillation for Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Jongwoo Ko, Se-Young Yun, Sungnyun Kim, Tianyi Chen","submitted_at":"2024-02-06T11:10:35Z","abstract_excerpt":"Knowledge distillation (KD) is widely used for compressing a teacher model to a smaller student model, reducing its inference cost and memory footprint while preserving model capabilities. However, current KD methods for auto-regressive sequence models (e.g., large language models) suffer from missing a standardized objective function. Moreover, the recent use of student-generated outputs to address training-inference mismatches has significantly escalated computational costs. To tackle these issues, we introduce DistiLLM, a more effective and efficient KD framework for auto-regressive languag"},"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":"2402.03898","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-02-06T11:10:35Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"380678fc17a23545c1f487d6c90cc98afee2c03b716085caf0657e05317a9180","abstract_canon_sha256":"cc3612366344b214dad19c9b221607421c0438662063dafd400ed446a67e78b8"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:39:30.788249Z","signature_b64":"mh6nci/ck7ixKkZDY5lDLhaSKcOEDqxkwV+NnRPkppfuR+pFcWNznvMRtJhwrNjX50hYyNEojXAFG28A7E6RBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b373cb888b456696bea4ab0e709f7cca8ea3640a5c1db5106f65784538e53252","last_reissued_at":"2026-07-05T08:39:30.787792Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:39:30.787792Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"DistiLLM: Towards Streamlined Distillation for Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Jongwoo Ko, Se-Young Yun, Sungnyun Kim, Tianyi Chen","submitted_at":"2024-02-06T11:10:35Z","abstract_excerpt":"Knowledge distillation (KD) is widely used for compressing a teacher model to a smaller student model, reducing its inference cost and memory footprint while preserving model capabilities. However, current KD methods for auto-regressive sequence models (e.g., large language models) suffer from missing a standardized objective function. Moreover, the recent use of student-generated outputs to address training-inference mismatches has significantly escalated computational costs. To tackle these issues, we introduce DistiLLM, a more effective and efficient KD framework for auto-regressive languag"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2402.03898","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/2402.03898/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":"2402.03898","created_at":"2026-07-05T08:39:30.787860+00:00"},{"alias_kind":"arxiv_version","alias_value":"2402.03898v2","created_at":"2026-07-05T08:39:30.787860+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2402.03898","created_at":"2026-07-05T08:39:30.787860+00:00"},{"alias_kind":"pith_short_12","alias_value":"WNZ4XCELIVTJ","created_at":"2026-07-05T08:39:30.787860+00:00"},{"alias_kind":"pith_short_16","alias_value":"WNZ4XCELIVTJNPVE","created_at":"2026-07-05T08:39:30.787860+00:00"},{"alias_kind":"pith_short_8","alias_value":"WNZ4XCEL","created_at":"2026-07-05T08:39:30.787860+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":17,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.05394","citing_title":"Weak-to-Strong Generalization via Direct On-Policy Distillation","ref_index":35,"is_internal_anchor":true},{"citing_arxiv_id":"2606.18216","citing_title":"Zone of Proximal Policy Optimization: Teacher in Prompts, Not Gradients","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2606.11709","citing_title":"RLCSD: Reinforcement Learning with Contrastive On-Policy Self-Distillation","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2606.09396","citing_title":"PriFT: Prior-Support Guided Supervised Fine-Tuning","ref_index":12,"is_internal_anchor":false},{"citing_arxiv_id":"2606.29340","citing_title":"PHF: Privileged Hidden Flow for On-Policy Self-Distillation","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2605.30448","citing_title":"Bounded Behavioral Indistinguishability for Black-Box LLM Distillation","ref_index":12,"is_internal_anchor":false},{"citing_arxiv_id":"2606.07604","citing_title":"Contribution Weights: A Geometrical Analysis of Self-Attention Transformers","ref_index":117,"is_internal_anchor":false},{"citing_arxiv_id":"2605.21606","citing_title":"When Are Teacher Tokens Reliable? Position-Weighted On-Policy Self-Distillation for Reasoning","ref_index":20,"is_internal_anchor":false},{"citing_arxiv_id":"2605.21924","citing_title":"Visual-Advantage On-Policy Distillation for Vision-Language Models","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2605.17743","citing_title":"MoASE++: Mixture of Activation Sparsity Experts with Domain-Adaptive On-policy Distillation for Continual Test Time Adaptation","ref_index":63,"is_internal_anchor":false},{"citing_arxiv_id":"2605.12741","citing_title":"Learning with Rare Success but Rich Feedback via Reflection-Enhanced Self-Distillation","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2605.11260","citing_title":"Curriculum Learning-Guided Progressive Distillation in Large Language Models","ref_index":26,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08737","citing_title":"The Extrapolation Cliff in On-Policy Distillation of Near-Deterministic Structured Outputs","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2605.05940","citing_title":"Near-Policy: Accelerating On-Policy Distillation via Asynchronous Generation and Selective Packing","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2605.01374","citing_title":"MTA: Multi-Granular Trajectory Alignment for Large Language Model Distillation","ref_index":77,"is_internal_anchor":false},{"citing_arxiv_id":"2604.14629","citing_title":"Switch-KD: Visual-Switch Knowledge Distillation for Vision-Language Models","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2604.20244","citing_title":"Hybrid Policy Distillation for LLMs","ref_index":47,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/WNZ4XCELIVTJNPVEVMHHBH34ZK","json":"https://pith.science/pith/WNZ4XCELIVTJNPVEVMHHBH34ZK.json","graph_json":"https://pith.science/api/pith-number/WNZ4XCELIVTJNPVEVMHHBH34ZK/graph.json","events_json":"https://pith.science/api/pith-number/WNZ4XCELIVTJNPVEVMHHBH34ZK/events.json","paper":"https://pith.science/paper/WNZ4XCEL"},"agent_actions":{"view_html":"https://pith.science/pith/WNZ4XCELIVTJNPVEVMHHBH34ZK","download_json":"https://pith.science/pith/WNZ4XCELIVTJNPVEVMHHBH34ZK.json","view_paper":"https://pith.science/paper/WNZ4XCEL","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2402.03898&json=true","fetch_graph":"https://pith.science/api/pith-number/WNZ4XCELIVTJNPVEVMHHBH34ZK/graph.json","fetch_events":"https://pith.science/api/pith-number/WNZ4XCELIVTJNPVEVMHHBH34ZK/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/WNZ4XCELIVTJNPVEVMHHBH34ZK/action/timestamp_anchor","attest_storage":"https://pith.science/pith/WNZ4XCELIVTJNPVEVMHHBH34ZK/action/storage_attestation","attest_author":"https://pith.science/pith/WNZ4XCELIVTJNPVEVMHHBH34ZK/action/author_attestation","sign_citation":"https://pith.science/pith/WNZ4XCELIVTJNPVEVMHHBH34ZK/action/citation_signature","submit_replication":"https://pith.science/pith/WNZ4XCELIVTJNPVEVMHHBH34ZK/action/replication_record"}},"created_at":"2026-07-05T08:39:30.787860+00:00","updated_at":"2026-07-05T08:39:30.787860+00:00"}