{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:PLGJ4ZXBYN2T6XY52S5IAK7EBU","short_pith_number":"pith:PLGJ4ZXB","schema_version":"1.0","canonical_sha256":"7acc9e66e1c3753f5f1dd4ba802be40d2fbb170a8a90535901838811f8f087ce","source":{"kind":"arxiv","id":"2103.08493","version":2},"attestation_state":"computed","paper":{"title":"How Many Data Points is a Prompt Worth?","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Alexander M. Rush, Teven Le Scao","submitted_at":"2021-03-15T16:10:23Z","abstract_excerpt":"When fine-tuning pretrained models for classification, researchers either use a generic model head or a task-specific prompt for prediction. Proponents of prompting have argued that prompts provide a method for injecting task-specific guidance, which is beneficial in low-data regimes. We aim to quantify this benefit through rigorous testing of prompts in a fair setting: comparing prompted and head-based fine-tuning in equal conditions across many tasks and data sizes. By controlling for many sources of advantage, we find that prompting does indeed provide a benefit, and that this benefit can b"},"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":"2103.08493","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2021-03-15T16:10:23Z","cross_cats_sorted":[],"title_canon_sha256":"fb3c3633ea16f6862b9f025098c91d412c89427ed1277028d069f196cdefa2e9","abstract_canon_sha256":"c03bf25ad0ee0efd09633f69028085de47974f032604447e72db00ec464c94a1"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T02:29:46.468718Z","signature_b64":"5Q0entYMsyYd1eUTz4Osdnn/SkaiCijAMs1MG3iwU5bAOC9mhdTf/tSD3J5EbNRe+Y6Ur94h1rVVFgGwXWqRCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"7acc9e66e1c3753f5f1dd4ba802be40d2fbb170a8a90535901838811f8f087ce","last_reissued_at":"2026-07-05T02:29:46.468280Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T02:29:46.468280Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"How Many Data Points is a Prompt Worth?","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Alexander M. Rush, Teven Le Scao","submitted_at":"2021-03-15T16:10:23Z","abstract_excerpt":"When fine-tuning pretrained models for classification, researchers either use a generic model head or a task-specific prompt for prediction. Proponents of prompting have argued that prompts provide a method for injecting task-specific guidance, which is beneficial in low-data regimes. We aim to quantify this benefit through rigorous testing of prompts in a fair setting: comparing prompted and head-based fine-tuning in equal conditions across many tasks and data sizes. By controlling for many sources of advantage, we find that prompting does indeed provide a benefit, and that this benefit can b"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2103.08493","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/2103.08493/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":"2103.08493","created_at":"2026-07-05T02:29:46.468342+00:00"},{"alias_kind":"arxiv_version","alias_value":"2103.08493v2","created_at":"2026-07-05T02:29:46.468342+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2103.08493","created_at":"2026-07-05T02:29:46.468342+00:00"},{"alias_kind":"pith_short_12","alias_value":"PLGJ4ZXBYN2T","created_at":"2026-07-05T02:29:46.468342+00:00"},{"alias_kind":"pith_short_16","alias_value":"PLGJ4ZXBYN2T6XY5","created_at":"2026-07-05T02:29:46.468342+00:00"},{"alias_kind":"pith_short_8","alias_value":"PLGJ4ZXB","created_at":"2026-07-05T02:29:46.468342+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.09125","citing_title":"Unveiling Privacy Risks in Multi-modal Large Language Models: Task-specific Vulnerabilities and Mitigation Challenges","ref_index":69,"is_internal_anchor":false},{"citing_arxiv_id":"2604.27905","citing_title":"CoNewsReader: Supporting Comprehensive Understanding and Raising Critical Thoughts on Social Media News Through Comments","ref_index":95,"is_internal_anchor":false},{"citing_arxiv_id":"2604.27905","citing_title":"CoNewsReader: Supporting Comprehensive Understanding and Raising Critical Thoughts on Social Media News Through Comments","ref_index":95,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/PLGJ4ZXBYN2T6XY52S5IAK7EBU","json":"https://pith.science/pith/PLGJ4ZXBYN2T6XY52S5IAK7EBU.json","graph_json":"https://pith.science/api/pith-number/PLGJ4ZXBYN2T6XY52S5IAK7EBU/graph.json","events_json":"https://pith.science/api/pith-number/PLGJ4ZXBYN2T6XY52S5IAK7EBU/events.json","paper":"https://pith.science/paper/PLGJ4ZXB"},"agent_actions":{"view_html":"https://pith.science/pith/PLGJ4ZXBYN2T6XY52S5IAK7EBU","download_json":"https://pith.science/pith/PLGJ4ZXBYN2T6XY52S5IAK7EBU.json","view_paper":"https://pith.science/paper/PLGJ4ZXB","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2103.08493&json=true","fetch_graph":"https://pith.science/api/pith-number/PLGJ4ZXBYN2T6XY52S5IAK7EBU/graph.json","fetch_events":"https://pith.science/api/pith-number/PLGJ4ZXBYN2T6XY52S5IAK7EBU/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/PLGJ4ZXBYN2T6XY52S5IAK7EBU/action/timestamp_anchor","attest_storage":"https://pith.science/pith/PLGJ4ZXBYN2T6XY52S5IAK7EBU/action/storage_attestation","attest_author":"https://pith.science/pith/PLGJ4ZXBYN2T6XY52S5IAK7EBU/action/author_attestation","sign_citation":"https://pith.science/pith/PLGJ4ZXBYN2T6XY52S5IAK7EBU/action/citation_signature","submit_replication":"https://pith.science/pith/PLGJ4ZXBYN2T6XY52S5IAK7EBU/action/replication_record"}},"created_at":"2026-07-05T02:29:46.468342+00:00","updated_at":"2026-07-05T02:29:46.468342+00:00"}