{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:M6FSUGMQHIG5V67XK34WTH2N6R","short_pith_number":"pith:M6FSUGMQ","schema_version":"1.0","canonical_sha256":"678b2a19903a0ddafbf756f9699f4df478254f2fb96f90e5713fc13119724e1d","source":{"kind":"arxiv","id":"2603.01875","version":3},"attestation_state":"computed","paper":{"title":"KDFlow: A User-Friendly and Efficient Knowledge Distillation Framework for Large Language Models","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Bojie Hu, Jinan Xu, Songming Zhang, Tong Zhang, Xue Zhang, Yufeng Chen","submitted_at":"2026-03-02T13:54:19Z","abstract_excerpt":"Knowledge distillation (KD) is an essential technique to compress large language models (LLMs) into smaller ones. However, despite the distinct roles of the student model and the teacher model in KD, most existing frameworks still use a homogeneous training backend (e.g., FSDP and DeepSpeed) for both models, leading to suboptimal training efficiency. In this paper, we present a novel framework for LLM distillation, termed \\textbf{KDFlow}, which features a decoupled architecture and employs SGLang for teacher inference. By bridging the training efficiency of FSDP2 and the inference efficiency o"},"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":"2603.01875","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CL","submitted_at":"2026-03-02T13:54:19Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"d34a7b8076c128e9b679da84002d88a896fe5cc4908fc4f0063131caff575731","abstract_canon_sha256":"48a65437ce1a10b5ce0839c41659eec05de930e074b8f89dbb2cb11f64f14b41"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-20T01:18:28.269908Z","signature_b64":"MBLGj8yz+maRyu+hkGPrNLsec7sDZ2sYilBWh5ZeMgXIjyYy0uzCg1y3hGRS2sdGl0YfE9rjIG0CqTWH+WirBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"678b2a19903a0ddafbf756f9699f4df478254f2fb96f90e5713fc13119724e1d","last_reissued_at":"2026-07-20T01:18:28.268847Z","signature_status":"signed_v1","first_computed_at":"2026-07-20T01:18:28.268847Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"KDFlow: A User-Friendly and Efficient Knowledge Distillation Framework for Large Language Models","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Bojie Hu, Jinan Xu, Songming Zhang, Tong Zhang, Xue Zhang, Yufeng Chen","submitted_at":"2026-03-02T13:54:19Z","abstract_excerpt":"Knowledge distillation (KD) is an essential technique to compress large language models (LLMs) into smaller ones. However, despite the distinct roles of the student model and the teacher model in KD, most existing frameworks still use a homogeneous training backend (e.g., FSDP and DeepSpeed) for both models, leading to suboptimal training efficiency. In this paper, we present a novel framework for LLM distillation, termed \\textbf{KDFlow}, which features a decoupled architecture and employs SGLang for teacher inference. By bridging the training efficiency of FSDP2 and the inference efficiency o"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2603.01875","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/2603.01875/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":"2603.01875","created_at":"2026-07-20T01:18:28.269319+00:00"},{"alias_kind":"arxiv_version","alias_value":"2603.01875v3","created_at":"2026-07-20T01:18:28.269319+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2603.01875","created_at":"2026-07-20T01:18:28.269319+00:00"},{"alias_kind":"pith_short_12","alias_value":"M6FSUGMQHIG5","created_at":"2026-07-20T01:18:28.269319+00:00"},{"alias_kind":"pith_short_16","alias_value":"M6FSUGMQHIG5V67X","created_at":"2026-07-20T01:18:28.269319+00:00"},{"alias_kind":"pith_short_8","alias_value":"M6FSUGMQ","created_at":"2026-07-20T01:18:28.269319+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":3,"sample":[{"citing_arxiv_id":"2606.24143","citing_title":"AsyncOPD: How Stale Can On-Policy Distillation Be?","ref_index":29,"is_internal_anchor":true},{"citing_arxiv_id":"2605.07711","citing_title":"SimCT: Recovering Lost Supervision for Cross-Tokenizer On-Policy Distillation","ref_index":60,"is_internal_anchor":true},{"citing_arxiv_id":"2605.07711","citing_title":"SimCT: Recovering Lost Supervision for Cross-Tokenizer On-Policy Distillation","ref_index":59,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/M6FSUGMQHIG5V67XK34WTH2N6R","json":"https://pith.science/pith/M6FSUGMQHIG5V67XK34WTH2N6R.json","graph_json":"https://pith.science/api/pith-number/M6FSUGMQHIG5V67XK34WTH2N6R/graph.json","events_json":"https://pith.science/api/pith-number/M6FSUGMQHIG5V67XK34WTH2N6R/events.json","paper":"https://pith.science/paper/M6FSUGMQ"},"agent_actions":{"view_html":"https://pith.science/pith/M6FSUGMQHIG5V67XK34WTH2N6R","download_json":"https://pith.science/pith/M6FSUGMQHIG5V67XK34WTH2N6R.json","view_paper":"https://pith.science/paper/M6FSUGMQ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2603.01875&json=true","fetch_graph":"https://pith.science/api/pith-number/M6FSUGMQHIG5V67XK34WTH2N6R/graph.json","fetch_events":"https://pith.science/api/pith-number/M6FSUGMQHIG5V67XK34WTH2N6R/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/M6FSUGMQHIG5V67XK34WTH2N6R/action/timestamp_anchor","attest_storage":"https://pith.science/pith/M6FSUGMQHIG5V67XK34WTH2N6R/action/storage_attestation","attest_author":"https://pith.science/pith/M6FSUGMQHIG5V67XK34WTH2N6R/action/author_attestation","sign_citation":"https://pith.science/pith/M6FSUGMQHIG5V67XK34WTH2N6R/action/citation_signature","submit_replication":"https://pith.science/pith/M6FSUGMQHIG5V67XK34WTH2N6R/action/replication_record"}},"created_at":"2026-07-20T01:18:28.269319+00:00","updated_at":"2026-07-20T01:18:28.269319+00:00"}