{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:PZAATPBRCXYHMJOHHT5AO4XVF6","short_pith_number":"pith:PZAATPBR","schema_version":"1.0","canonical_sha256":"7e4009bc3115f07625c73cfa0772f52faf90d0f91ee2fd3e100338b1bcd6b534","source":{"kind":"arxiv","id":"2606.06920","version":1},"attestation_state":"computed","paper":{"title":"The Fine-Tuning Trap: Evaluating Negative Transfer and the Role of PEFT in Sub-1B Mathematical Reasoning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Chun Tao, Rahul Nair","submitted_at":"2026-06-05T05:34:13Z","abstract_excerpt":"Deploying Small Language Models (SLMs) on edge devices requires efficient fine-tuning strategies that adapt models to new tasks without degrading their general capabilities. In this study, we benchmark five sub-1B models (135M-1B) on mathematical reasoning tasks and uncover a critical vulnerability: Full Fine-Tuning (Full FT) actively harms performance in models under 300M parameters, often dropping accuracy below zero-shot baselines. This \"negative transfer\" makes Parameter-Efficient Fine-Tuning (PEFT) not just an efficiency preference, but a stability requirement. We find that while Low-Rank"},"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":"2606.06920","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-06-05T05:34:13Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"361365a76a4a4f6a9d43f2185ec671672f6d7eb5228bec553959702fffbedba7","abstract_canon_sha256":"9f104ff040a38c6b43a8c54146c396db0264355f7708d407db6e79f8fe9adf3f"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-08T01:04:35.974080Z","signature_b64":"KAdTY5e7Cxg3qFD9I/5GcoYQp8vny5S2y0O/Dqp17Wd8o+cGiwdduW4uDJ+619upzTG14EIEuaJYm8wOiOnTCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"7e4009bc3115f07625c73cfa0772f52faf90d0f91ee2fd3e100338b1bcd6b534","last_reissued_at":"2026-06-08T01:04:35.973342Z","signature_status":"signed_v1","first_computed_at":"2026-06-08T01:04:35.973342Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"The Fine-Tuning Trap: Evaluating Negative Transfer and the Role of PEFT in Sub-1B Mathematical Reasoning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Chun Tao, Rahul Nair","submitted_at":"2026-06-05T05:34:13Z","abstract_excerpt":"Deploying Small Language Models (SLMs) on edge devices requires efficient fine-tuning strategies that adapt models to new tasks without degrading their general capabilities. In this study, we benchmark five sub-1B models (135M-1B) on mathematical reasoning tasks and uncover a critical vulnerability: Full Fine-Tuning (Full FT) actively harms performance in models under 300M parameters, often dropping accuracy below zero-shot baselines. This \"negative transfer\" makes Parameter-Efficient Fine-Tuning (PEFT) not just an efficiency preference, but a stability requirement. We find that while Low-Rank"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.06920","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/2606.06920/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":"2606.06920","created_at":"2026-06-08T01:04:35.973455+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.06920v1","created_at":"2026-06-08T01:04:35.973455+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.06920","created_at":"2026-06-08T01:04:35.973455+00:00"},{"alias_kind":"pith_short_12","alias_value":"PZAATPBRCXYH","created_at":"2026-06-08T01:04:35.973455+00:00"},{"alias_kind":"pith_short_16","alias_value":"PZAATPBRCXYHMJOH","created_at":"2026-06-08T01:04:35.973455+00:00"},{"alias_kind":"pith_short_8","alias_value":"PZAATPBR","created_at":"2026-06-08T01:04:35.973455+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/PZAATPBRCXYHMJOHHT5AO4XVF6","json":"https://pith.science/pith/PZAATPBRCXYHMJOHHT5AO4XVF6.json","graph_json":"https://pith.science/api/pith-number/PZAATPBRCXYHMJOHHT5AO4XVF6/graph.json","events_json":"https://pith.science/api/pith-number/PZAATPBRCXYHMJOHHT5AO4XVF6/events.json","paper":"https://pith.science/paper/PZAATPBR"},"agent_actions":{"view_html":"https://pith.science/pith/PZAATPBRCXYHMJOHHT5AO4XVF6","download_json":"https://pith.science/pith/PZAATPBRCXYHMJOHHT5AO4XVF6.json","view_paper":"https://pith.science/paper/PZAATPBR","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.06920&json=true","fetch_graph":"https://pith.science/api/pith-number/PZAATPBRCXYHMJOHHT5AO4XVF6/graph.json","fetch_events":"https://pith.science/api/pith-number/PZAATPBRCXYHMJOHHT5AO4XVF6/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/PZAATPBRCXYHMJOHHT5AO4XVF6/action/timestamp_anchor","attest_storage":"https://pith.science/pith/PZAATPBRCXYHMJOHHT5AO4XVF6/action/storage_attestation","attest_author":"https://pith.science/pith/PZAATPBRCXYHMJOHHT5AO4XVF6/action/author_attestation","sign_citation":"https://pith.science/pith/PZAATPBRCXYHMJOHHT5AO4XVF6/action/citation_signature","submit_replication":"https://pith.science/pith/PZAATPBRCXYHMJOHHT5AO4XVF6/action/replication_record"}},"created_at":"2026-06-08T01:04:35.973455+00:00","updated_at":"2026-06-08T01:04:35.973455+00:00"}