{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:6UWFNICGVXZ6O3DW2WW6VOJTUA","short_pith_number":"pith:6UWFNICG","schema_version":"1.0","canonical_sha256":"f52c56a046adf3e76c76d5adeab933a03512f24542f0bd26ac59a4f2999a0aea","source":{"kind":"arxiv","id":"2406.08660","version":2},"attestation_state":"computed","paper":{"title":"Fine-Tuned 'Small' LLMs (Still) Significantly Outperform Zero-Shot Generative AI Models in Text Classification","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Marco Martini, Martin Juan Jos\\'e Bucher","submitted_at":"2024-06-12T21:46:13Z","abstract_excerpt":"Generative AI offers a simple, prompt-based alternative to fine-tuning smaller BERT-style LLMs for text classification tasks. This promises to eliminate the need for manually labeled training data and task-specific model training. However, it remains an open question whether tools like ChatGPT can deliver on this promise. In this paper, we show that smaller, fine-tuned LLMs (still) consistently and significantly outperform larger, zero-shot prompted models in text classification. We compare three major generative AI models (ChatGPT with GPT-3.5/GPT-4 and Claude Opus) with several fine-tuned LL"},"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":"2406.08660","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-06-12T21:46:13Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"dce087072db83dac797e49f02f3e42f85dcb0a32d8fa1052eb11372efcc4b829","abstract_canon_sha256":"e6faf2a7867d778c1bec95679ca6037e12d094eb6068e164c81af0715f3d655d"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:56:05.213800Z","signature_b64":"6u2FpDJNFZzE+nAAWCb859mjLIDRXgrZdwJc1FqvAQUgo7PKDeTTuyM+sV0jbJgjDRDbwC6ziupmhTlpinwlDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f52c56a046adf3e76c76d5adeab933a03512f24542f0bd26ac59a4f2999a0aea","last_reissued_at":"2026-07-05T08:56:05.213324Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:56:05.213324Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Fine-Tuned 'Small' LLMs (Still) Significantly Outperform Zero-Shot Generative AI Models in Text Classification","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Marco Martini, Martin Juan Jos\\'e Bucher","submitted_at":"2024-06-12T21:46:13Z","abstract_excerpt":"Generative AI offers a simple, prompt-based alternative to fine-tuning smaller BERT-style LLMs for text classification tasks. This promises to eliminate the need for manually labeled training data and task-specific model training. However, it remains an open question whether tools like ChatGPT can deliver on this promise. In this paper, we show that smaller, fine-tuned LLMs (still) consistently and significantly outperform larger, zero-shot prompted models in text classification. We compare three major generative AI models (ChatGPT with GPT-3.5/GPT-4 and Claude Opus) with several fine-tuned LL"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.08660","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/2406.08660/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":"2406.08660","created_at":"2026-07-05T08:56:05.213384+00:00"},{"alias_kind":"arxiv_version","alias_value":"2406.08660v2","created_at":"2026-07-05T08:56:05.213384+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.08660","created_at":"2026-07-05T08:56:05.213384+00:00"},{"alias_kind":"pith_short_12","alias_value":"6UWFNICGVXZ6","created_at":"2026-07-05T08:56:05.213384+00:00"},{"alias_kind":"pith_short_16","alias_value":"6UWFNICGVXZ6O3DW","created_at":"2026-07-05T08:56:05.213384+00:00"},{"alias_kind":"pith_short_8","alias_value":"6UWFNICG","created_at":"2026-07-05T08:56:05.213384+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":10,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.26489","citing_title":"Comparing BERT Sentence-Pair Classification and Few-Shot LLM Prompting for Detecting Threat and Solution Framing in German Climate News","ref_index":35,"is_internal_anchor":false},{"citing_arxiv_id":"2607.02407","citing_title":"Text-Driven 3D Indoor Scene Synthesis in Non-Manhattan Environments","ref_index":69,"is_internal_anchor":false},{"citing_arxiv_id":"2606.11854","citing_title":"Fine-tuning Multi-modal LLMs with ART: Art-based Reinforcement Training","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2606.10210","citing_title":"AnnotateThis: Analyzing a human-LLM system for annotating social media data with the concept of climate change mitigation pessimism","ref_index":45,"is_internal_anchor":false},{"citing_arxiv_id":"2606.07479","citing_title":"Supervision versus Demonstration-Based In-Context Learning for Multiword Expression Classification","ref_index":163,"is_internal_anchor":false},{"citing_arxiv_id":"2606.00462","citing_title":"Short-form Text Rewriting with Phi Silica","ref_index":1,"is_internal_anchor":false},{"citing_arxiv_id":"2512.05525","citing_title":"Poodle: Seamlessly Scaling Down Large Language Models with Just-in-Time Model Replacement","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2603.12021","citing_title":"Just Use XML: Revisiting Joint Translation and Label Projection","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2604.27143","citing_title":"Enhancing Linux Privilege Escalation Attack Capabilities of Local LLM Agents","ref_index":2,"is_internal_anchor":false},{"citing_arxiv_id":"2604.17725","citing_title":"RePrompT: Recurrent Prompt Tuning for Integrating Structured EHR Encoders with Large Language Models","ref_index":120,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/6UWFNICGVXZ6O3DW2WW6VOJTUA","json":"https://pith.science/pith/6UWFNICGVXZ6O3DW2WW6VOJTUA.json","graph_json":"https://pith.science/api/pith-number/6UWFNICGVXZ6O3DW2WW6VOJTUA/graph.json","events_json":"https://pith.science/api/pith-number/6UWFNICGVXZ6O3DW2WW6VOJTUA/events.json","paper":"https://pith.science/paper/6UWFNICG"},"agent_actions":{"view_html":"https://pith.science/pith/6UWFNICGVXZ6O3DW2WW6VOJTUA","download_json":"https://pith.science/pith/6UWFNICGVXZ6O3DW2WW6VOJTUA.json","view_paper":"https://pith.science/paper/6UWFNICG","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2406.08660&json=true","fetch_graph":"https://pith.science/api/pith-number/6UWFNICGVXZ6O3DW2WW6VOJTUA/graph.json","fetch_events":"https://pith.science/api/pith-number/6UWFNICGVXZ6O3DW2WW6VOJTUA/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/6UWFNICGVXZ6O3DW2WW6VOJTUA/action/timestamp_anchor","attest_storage":"https://pith.science/pith/6UWFNICGVXZ6O3DW2WW6VOJTUA/action/storage_attestation","attest_author":"https://pith.science/pith/6UWFNICGVXZ6O3DW2WW6VOJTUA/action/author_attestation","sign_citation":"https://pith.science/pith/6UWFNICGVXZ6O3DW2WW6VOJTUA/action/citation_signature","submit_replication":"https://pith.science/pith/6UWFNICGVXZ6O3DW2WW6VOJTUA/action/replication_record"}},"created_at":"2026-07-05T08:56:05.213384+00:00","updated_at":"2026-07-05T08:56:05.213384+00:00"}