{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:QACEXORQUVYLLNROY4H2LQ4BAV","short_pith_number":"pith:QACEXORQ","schema_version":"1.0","canonical_sha256":"80044bba30a570b5b62ec70fa5c3810543d75d00f1ed5a2b78f587e8053718e6","source":{"kind":"arxiv","id":"2410.18588","version":1},"attestation_state":"computed","paper":{"title":"Knowledge Distillation Using Frontier Open-source LLMs: Generalizability and the Role of Synthetic Data","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Anup Shirgaonkar, Nazmiye Ceren Abay, Nikhil Pandey, Tolga Aktas, Vijay Aski","submitted_at":"2024-10-24T09:37:23Z","abstract_excerpt":"Leading open-source large language models (LLMs) such as Llama-3.1-Instruct-405B are extremely capable at generating text, answering questions, and solving a variety of natural language understanding tasks. However, they incur higher inference cost and latency compared to smaller LLMs. Knowledge distillation provides a way to use outputs from these large, capable teacher models to train smaller student models which can be used for inference at lower cost and latency, while retaining comparable accuracy. We investigate the efficacy of distillation using the Llama-3.1-405B-Instruct teacher and t"},"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":"2410.18588","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.LG","submitted_at":"2024-10-24T09:37:23Z","cross_cats_sorted":[],"title_canon_sha256":"edf65d95bf284a9f2be89a1b3e87044f8e922317b66e5ec26ad141bda42d43d2","abstract_canon_sha256":"b77b506efc975ca9b1098a28c1c07affdf2561aa7be669510d8f09bdf737d947"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:25:10.836076Z","signature_b64":"dMvIS/21PDdDCl+FvoAFQAVApEHaB2ZoGg1i4abP5a7W35/zrdqmqIEouFZFCBwad1F9DyYDYikmBlQRxpxLBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"80044bba30a570b5b62ec70fa5c3810543d75d00f1ed5a2b78f587e8053718e6","last_reissued_at":"2026-07-05T09:25:10.835586Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:25:10.835586Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Knowledge Distillation Using Frontier Open-source LLMs: Generalizability and the Role of Synthetic Data","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Anup Shirgaonkar, Nazmiye Ceren Abay, Nikhil Pandey, Tolga Aktas, Vijay Aski","submitted_at":"2024-10-24T09:37:23Z","abstract_excerpt":"Leading open-source large language models (LLMs) such as Llama-3.1-Instruct-405B are extremely capable at generating text, answering questions, and solving a variety of natural language understanding tasks. However, they incur higher inference cost and latency compared to smaller LLMs. Knowledge distillation provides a way to use outputs from these large, capable teacher models to train smaller student models which can be used for inference at lower cost and latency, while retaining comparable accuracy. We investigate the efficacy of distillation using the Llama-3.1-405B-Instruct teacher and t"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.18588","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":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2410.18588/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":"2410.18588","created_at":"2026-07-05T09:25:10.835643+00:00"},{"alias_kind":"arxiv_version","alias_value":"2410.18588v1","created_at":"2026-07-05T09:25:10.835643+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.18588","created_at":"2026-07-05T09:25:10.835643+00:00"},{"alias_kind":"pith_short_12","alias_value":"QACEXORQUVYL","created_at":"2026-07-05T09:25:10.835643+00:00"},{"alias_kind":"pith_short_16","alias_value":"QACEXORQUVYLLNRO","created_at":"2026-07-05T09:25:10.835643+00:00"},{"alias_kind":"pith_short_8","alias_value":"QACEXORQ","created_at":"2026-07-05T09:25:10.835643+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.22942","citing_title":"Understanding Knowledge Distillation in Post-Training: When It Helps and When It Fails","ref_index":2,"is_internal_anchor":false},{"citing_arxiv_id":"2604.07486","citing_title":"Private Seeds, Public LLMs: Realistic and Privacy-Preserving Synthetic Data Generation","ref_index":10,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/QACEXORQUVYLLNROY4H2LQ4BAV","json":"https://pith.science/pith/QACEXORQUVYLLNROY4H2LQ4BAV.json","graph_json":"https://pith.science/api/pith-number/QACEXORQUVYLLNROY4H2LQ4BAV/graph.json","events_json":"https://pith.science/api/pith-number/QACEXORQUVYLLNROY4H2LQ4BAV/events.json","paper":"https://pith.science/paper/QACEXORQ"},"agent_actions":{"view_html":"https://pith.science/pith/QACEXORQUVYLLNROY4H2LQ4BAV","download_json":"https://pith.science/pith/QACEXORQUVYLLNROY4H2LQ4BAV.json","view_paper":"https://pith.science/paper/QACEXORQ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2410.18588&json=true","fetch_graph":"https://pith.science/api/pith-number/QACEXORQUVYLLNROY4H2LQ4BAV/graph.json","fetch_events":"https://pith.science/api/pith-number/QACEXORQUVYLLNROY4H2LQ4BAV/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/QACEXORQUVYLLNROY4H2LQ4BAV/action/timestamp_anchor","attest_storage":"https://pith.science/pith/QACEXORQUVYLLNROY4H2LQ4BAV/action/storage_attestation","attest_author":"https://pith.science/pith/QACEXORQUVYLLNROY4H2LQ4BAV/action/author_attestation","sign_citation":"https://pith.science/pith/QACEXORQUVYLLNROY4H2LQ4BAV/action/citation_signature","submit_replication":"https://pith.science/pith/QACEXORQUVYLLNROY4H2LQ4BAV/action/replication_record"}},"created_at":"2026-07-05T09:25:10.835643+00:00","updated_at":"2026-07-05T09:25:10.835643+00:00"}