{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:ZLR3EUYQNAMDETY2ISHJ6OFCKF","short_pith_number":"pith:ZLR3EUYQ","schema_version":"1.0","canonical_sha256":"cae3b253106818324f1a448e9f38a2515581292a48bc2fa3698f19f2af1ed762","source":{"kind":"arxiv","id":"2607.03801","version":1},"attestation_state":"computed","paper":{"title":"Punching Above Their Weight: Classification-Head Fine-Tuning of Tiny Language Models (TLMs) for Verifiable Multiple-Choice Tasks","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Bhavesh Sood, Jaromir Savelka","submitted_at":"2026-07-04T10:09:32Z","abstract_excerpt":"We define Tiny Language Models (TLMs) as models below roughly 3B parameters that fit on mainstream consumer devices. We study how to adapt them for and use them on verifiable multiple-choice tasks. We compare three LoRA-based fine-tuning paradigms (label generation, gold only, and our discriminative classification head) on a unified setup across several Qwen3 models from 0.6B to 8B and five benchmarks: HellaSwag, WinoGrande, PIQA, SciQ and ARC-C. Classification-head fine-tuning reliably outperforms label generation (+2-3%) at the 0.6B and 1.7B scales. Further, TLMs fine-tuned using the discrim"},"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":"2607.03801","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-07-04T10:09:32Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"bdd5ae64f081bcfbcb1d7878b148d337da2a89f0819e31aaf8cd12507866f94e","abstract_canon_sha256":"f4d818dd4dfc814a76eaa0c55b7ca1320c4981326e9b8b312c1c79f077ff77c3"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-07T02:18:08.334925Z","signature_b64":"BHwUvsdO7/rWLNc6E+tTt8+w2DIpHRLjtq6cfK7HplwDEpfA7VgRrx2/eygvqE048p+h0e/KDoRuNumHjl0IBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"cae3b253106818324f1a448e9f38a2515581292a48bc2fa3698f19f2af1ed762","last_reissued_at":"2026-07-07T02:18:08.334166Z","signature_status":"signed_v1","first_computed_at":"2026-07-07T02:18:08.334166Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Punching Above Their Weight: Classification-Head Fine-Tuning of Tiny Language Models (TLMs) for Verifiable Multiple-Choice Tasks","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Bhavesh Sood, Jaromir Savelka","submitted_at":"2026-07-04T10:09:32Z","abstract_excerpt":"We define Tiny Language Models (TLMs) as models below roughly 3B parameters that fit on mainstream consumer devices. We study how to adapt them for and use them on verifiable multiple-choice tasks. We compare three LoRA-based fine-tuning paradigms (label generation, gold only, and our discriminative classification head) on a unified setup across several Qwen3 models from 0.6B to 8B and five benchmarks: HellaSwag, WinoGrande, PIQA, SciQ and ARC-C. Classification-head fine-tuning reliably outperforms label generation (+2-3%) at the 0.6B and 1.7B scales. Further, TLMs fine-tuned using the discrim"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.03801","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/2607.03801/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":"2607.03801","created_at":"2026-07-07T02:18:08.334276+00:00"},{"alias_kind":"arxiv_version","alias_value":"2607.03801v1","created_at":"2026-07-07T02:18:08.334276+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.03801","created_at":"2026-07-07T02:18:08.334276+00:00"},{"alias_kind":"pith_short_12","alias_value":"ZLR3EUYQNAMD","created_at":"2026-07-07T02:18:08.334276+00:00"},{"alias_kind":"pith_short_16","alias_value":"ZLR3EUYQNAMDETY2","created_at":"2026-07-07T02:18:08.334276+00:00"},{"alias_kind":"pith_short_8","alias_value":"ZLR3EUYQ","created_at":"2026-07-07T02:18:08.334276+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/ZLR3EUYQNAMDETY2ISHJ6OFCKF","json":"https://pith.science/pith/ZLR3EUYQNAMDETY2ISHJ6OFCKF.json","graph_json":"https://pith.science/api/pith-number/ZLR3EUYQNAMDETY2ISHJ6OFCKF/graph.json","events_json":"https://pith.science/api/pith-number/ZLR3EUYQNAMDETY2ISHJ6OFCKF/events.json","paper":"https://pith.science/paper/ZLR3EUYQ"},"agent_actions":{"view_html":"https://pith.science/pith/ZLR3EUYQNAMDETY2ISHJ6OFCKF","download_json":"https://pith.science/pith/ZLR3EUYQNAMDETY2ISHJ6OFCKF.json","view_paper":"https://pith.science/paper/ZLR3EUYQ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2607.03801&json=true","fetch_graph":"https://pith.science/api/pith-number/ZLR3EUYQNAMDETY2ISHJ6OFCKF/graph.json","fetch_events":"https://pith.science/api/pith-number/ZLR3EUYQNAMDETY2ISHJ6OFCKF/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ZLR3EUYQNAMDETY2ISHJ6OFCKF/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ZLR3EUYQNAMDETY2ISHJ6OFCKF/action/storage_attestation","attest_author":"https://pith.science/pith/ZLR3EUYQNAMDETY2ISHJ6OFCKF/action/author_attestation","sign_citation":"https://pith.science/pith/ZLR3EUYQNAMDETY2ISHJ6OFCKF/action/citation_signature","submit_replication":"https://pith.science/pith/ZLR3EUYQNAMDETY2ISHJ6OFCKF/action/replication_record"}},"created_at":"2026-07-07T02:18:08.334276+00:00","updated_at":"2026-07-07T02:18:08.334276+00:00"}