{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:YEX7AFDQJTMXCMJ2HWAH6KXNCS","short_pith_number":"pith:YEX7AFDQ","schema_version":"1.0","canonical_sha256":"c12ff014704cd971313a3d807f2aed14b30bc2959960b64cd7e6143bf2226f38","source":{"kind":"arxiv","id":"2406.05322","version":1},"attestation_state":"computed","paper":{"title":"Teaching-Assistant-in-the-Loop: Improving Knowledge Distillation from Imperfect Teacher Models in Low-Budget Scenarios","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Wei Ai, Yuhang Zhou","submitted_at":"2024-06-08T02:17:43Z","abstract_excerpt":"There is increasing interest in distilling task-specific knowledge from large language models (LLM) to smaller student models. Nonetheless, LLM distillation presents a dual challenge: 1) there is a high cost associated with querying the teacher LLM, such as GPT-4, for gathering an ample number of demonstrations; 2) the teacher LLM might provide imperfect outputs with a negative impact on the student's learning process. To enhance sample efficiency within resource-constrained, imperfect teacher scenarios, we propose a three-component framework leveraging three signal types. The first signal is "},"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":"2406.05322","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-06-08T02:17:43Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"ce5fc81f7d0fea5c36a52ef2200264561ea2f3dbd7a7258a35066a643116de23","abstract_canon_sha256":"605411e8f2fb5c2e03bfbc85c774508fb858281410669014b1a4af4b06671880"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:29:03.908911Z","signature_b64":"W8O/L77Z2M5VBxloiPdUZwpFZpbG/4V1WPgvD9grpzmYbTzUbSZ9WkzMb07l8TXh5NDX6W3fqeQ0PzSA71SPDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"c12ff014704cd971313a3d807f2aed14b30bc2959960b64cd7e6143bf2226f38","last_reissued_at":"2026-07-05T08:29:03.908197Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:29:03.908197Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Teaching-Assistant-in-the-Loop: Improving Knowledge Distillation from Imperfect Teacher Models in Low-Budget Scenarios","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Wei Ai, Yuhang Zhou","submitted_at":"2024-06-08T02:17:43Z","abstract_excerpt":"There is increasing interest in distilling task-specific knowledge from large language models (LLM) to smaller student models. Nonetheless, LLM distillation presents a dual challenge: 1) there is a high cost associated with querying the teacher LLM, such as GPT-4, for gathering an ample number of demonstrations; 2) the teacher LLM might provide imperfect outputs with a negative impact on the student's learning process. To enhance sample efficiency within resource-constrained, imperfect teacher scenarios, we propose a three-component framework leveraging three signal types. The first signal is "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.05322","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/2406.05322/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":"2406.05322","created_at":"2026-07-05T08:29:03.908296+00:00"},{"alias_kind":"arxiv_version","alias_value":"2406.05322v1","created_at":"2026-07-05T08:29:03.908296+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.05322","created_at":"2026-07-05T08:29:03.908296+00:00"},{"alias_kind":"pith_short_12","alias_value":"YEX7AFDQJTMX","created_at":"2026-07-05T08:29:03.908296+00:00"},{"alias_kind":"pith_short_16","alias_value":"YEX7AFDQJTMXCMJ2","created_at":"2026-07-05T08:29:03.908296+00:00"},{"alias_kind":"pith_short_8","alias_value":"YEX7AFDQ","created_at":"2026-07-05T08:29:03.908296+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.01476","citing_title":"OmniOPD: Logit-Free On-Policy Distillation via Speculative Verification","ref_index":37,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/YEX7AFDQJTMXCMJ2HWAH6KXNCS","json":"https://pith.science/pith/YEX7AFDQJTMXCMJ2HWAH6KXNCS.json","graph_json":"https://pith.science/api/pith-number/YEX7AFDQJTMXCMJ2HWAH6KXNCS/graph.json","events_json":"https://pith.science/api/pith-number/YEX7AFDQJTMXCMJ2HWAH6KXNCS/events.json","paper":"https://pith.science/paper/YEX7AFDQ"},"agent_actions":{"view_html":"https://pith.science/pith/YEX7AFDQJTMXCMJ2HWAH6KXNCS","download_json":"https://pith.science/pith/YEX7AFDQJTMXCMJ2HWAH6KXNCS.json","view_paper":"https://pith.science/paper/YEX7AFDQ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2406.05322&json=true","fetch_graph":"https://pith.science/api/pith-number/YEX7AFDQJTMXCMJ2HWAH6KXNCS/graph.json","fetch_events":"https://pith.science/api/pith-number/YEX7AFDQJTMXCMJ2HWAH6KXNCS/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/YEX7AFDQJTMXCMJ2HWAH6KXNCS/action/timestamp_anchor","attest_storage":"https://pith.science/pith/YEX7AFDQJTMXCMJ2HWAH6KXNCS/action/storage_attestation","attest_author":"https://pith.science/pith/YEX7AFDQJTMXCMJ2HWAH6KXNCS/action/author_attestation","sign_citation":"https://pith.science/pith/YEX7AFDQJTMXCMJ2HWAH6KXNCS/action/citation_signature","submit_replication":"https://pith.science/pith/YEX7AFDQJTMXCMJ2HWAH6KXNCS/action/replication_record"}},"created_at":"2026-07-05T08:29:03.908296+00:00","updated_at":"2026-07-05T08:29:03.908296+00:00"}