{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:DPCEBRV2CKVUHOX2CRFIVVL47S","short_pith_number":"pith:DPCEBRV2","schema_version":"1.0","canonical_sha256":"1bc440c6ba12ab43bafa144a8ad57cfca2d898d8ab4b9b1c259f77c4eefac471","source":{"kind":"arxiv","id":"2502.19545","version":2},"attestation_state":"computed","paper":{"title":"Winning Big with Small Models: Knowledge Distillation vs. Self-Training for Reducing Hallucination in Product QA Agents","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Ashley Lewis, Jing Liu, Kieran Parsons, Michael White, Toshiaki Koike-Akino, Ye Wang","submitted_at":"2025-02-26T20:34:58Z","abstract_excerpt":"The deployment of Large Language Models (LLMs) in customer support is constrained by hallucination (generating false information) and the high cost of proprietary models. To address these challenges, we propose a retrieval-augmented question-answering (QA) pipeline and explore how to balance human input and automation. Using a dataset of questions about a Samsung Smart TV user manual, we demonstrate that synthetic data generated by LLMs outperforms crowdsourced data in reducing hallucination in finetuned models. We also compare self-training (fine-tuning models on their own outputs) and knowle"},"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":"2502.19545","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-02-26T20:34:58Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"36d3683d98f56fd5986c3bc44228248d213e41b67303b401b91c11656bdf8938","abstract_canon_sha256":"2a805b12781b0764ff4c3b63ee89007316a6c539e1ffc6315a296ff1c6d096da"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:40:33.250207Z","signature_b64":"CQqivYyljBmE3Za0ZCHHe1uPDc5bbVSHo5hhuoXsL+NA//AHBcB0rqUbX9PD8LXTZBX8gXdth/fAXJuoOl/IDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"1bc440c6ba12ab43bafa144a8ad57cfca2d898d8ab4b9b1c259f77c4eefac471","last_reissued_at":"2026-07-05T11:40:33.249720Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:40:33.249720Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Winning Big with Small Models: Knowledge Distillation vs. Self-Training for Reducing Hallucination in Product QA Agents","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Ashley Lewis, Jing Liu, Kieran Parsons, Michael White, Toshiaki Koike-Akino, Ye Wang","submitted_at":"2025-02-26T20:34:58Z","abstract_excerpt":"The deployment of Large Language Models (LLMs) in customer support is constrained by hallucination (generating false information) and the high cost of proprietary models. To address these challenges, we propose a retrieval-augmented question-answering (QA) pipeline and explore how to balance human input and automation. Using a dataset of questions about a Samsung Smart TV user manual, we demonstrate that synthetic data generated by LLMs outperforms crowdsourced data in reducing hallucination in finetuned models. We also compare self-training (fine-tuning models on their own outputs) and knowle"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.19545","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/2502.19545/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":"2502.19545","created_at":"2026-07-05T11:40:33.249776+00:00"},{"alias_kind":"arxiv_version","alias_value":"2502.19545v2","created_at":"2026-07-05T11:40:33.249776+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.19545","created_at":"2026-07-05T11:40:33.249776+00:00"},{"alias_kind":"pith_short_12","alias_value":"DPCEBRV2CKVU","created_at":"2026-07-05T11:40:33.249776+00:00"},{"alias_kind":"pith_short_16","alias_value":"DPCEBRV2CKVUHOX2","created_at":"2026-07-05T11:40:33.249776+00:00"},{"alias_kind":"pith_short_8","alias_value":"DPCEBRV2","created_at":"2026-07-05T11:40:33.249776+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":1,"is_internal_anchor":false},{"citing_arxiv_id":"2605.15208","citing_title":"Quantization Undoes Alignment: Bias Emergence in Compressed LLMs Across Models and Precision Levels","ref_index":7,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/DPCEBRV2CKVUHOX2CRFIVVL47S","json":"https://pith.science/pith/DPCEBRV2CKVUHOX2CRFIVVL47S.json","graph_json":"https://pith.science/api/pith-number/DPCEBRV2CKVUHOX2CRFIVVL47S/graph.json","events_json":"https://pith.science/api/pith-number/DPCEBRV2CKVUHOX2CRFIVVL47S/events.json","paper":"https://pith.science/paper/DPCEBRV2"},"agent_actions":{"view_html":"https://pith.science/pith/DPCEBRV2CKVUHOX2CRFIVVL47S","download_json":"https://pith.science/pith/DPCEBRV2CKVUHOX2CRFIVVL47S.json","view_paper":"https://pith.science/paper/DPCEBRV2","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2502.19545&json=true","fetch_graph":"https://pith.science/api/pith-number/DPCEBRV2CKVUHOX2CRFIVVL47S/graph.json","fetch_events":"https://pith.science/api/pith-number/DPCEBRV2CKVUHOX2CRFIVVL47S/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/DPCEBRV2CKVUHOX2CRFIVVL47S/action/timestamp_anchor","attest_storage":"https://pith.science/pith/DPCEBRV2CKVUHOX2CRFIVVL47S/action/storage_attestation","attest_author":"https://pith.science/pith/DPCEBRV2CKVUHOX2CRFIVVL47S/action/author_attestation","sign_citation":"https://pith.science/pith/DPCEBRV2CKVUHOX2CRFIVVL47S/action/citation_signature","submit_replication":"https://pith.science/pith/DPCEBRV2CKVUHOX2CRFIVVL47S/action/replication_record"}},"created_at":"2026-07-05T11:40:33.249776+00:00","updated_at":"2026-07-05T11:40:33.249776+00:00"}