{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:437CQS27UKPFZBP6CL5EN5NAXT","short_pith_number":"pith:437CQS27","schema_version":"1.0","canonical_sha256":"e6fe284b5fa29e5c85fe12fa46f5a0bcd78d5f6c2daa5dbff00dca8af435b8d6","source":{"kind":"arxiv","id":"2108.13487","version":1},"attestation_state":"computed","paper":{"title":"Want To Reduce Labeling Cost? GPT-3 Can Help","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Chenguang Zhu, Michael Zeng, Shuohang Wang, Yang Liu, Yichong Xu","submitted_at":"2021-08-30T19:18:24Z","abstract_excerpt":"Data annotation is a time-consuming and labor-intensive process for many NLP tasks. Although there exist various methods to produce pseudo data labels, they are often task-specific and require a decent amount of labeled data to start with. Recently, the immense language model GPT-3 with 175 billion parameters has achieved tremendous improvement across many few-shot learning tasks. In this paper, we explore ways to leverage GPT-3 as a low-cost data labeler to train other models. We find that, to make the downstream model achieve the same performance on a variety of NLU and NLG tasks, it costs 5"},"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":"2108.13487","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2021-08-30T19:18:24Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"47cdeb28952efc8f094f3c301d6a2a877892d1b60c1e5fa0347e2a2966725b48","abstract_canon_sha256":"473dbefe78a3359a326f11d846c322e81c701acd2a0e61cbe53fe03cd3e8c588"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T03:10:14.687572Z","signature_b64":"Kak8+6vOG0fp9OJcaI1EUvX01DViqF1Ea/YLRFl6rhNsvNp6xHH1Gfe2XWJTMexPPlJ5bYToZootYM5+TsjEDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e6fe284b5fa29e5c85fe12fa46f5a0bcd78d5f6c2daa5dbff00dca8af435b8d6","last_reissued_at":"2026-07-05T03:10:14.687070Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T03:10:14.687070Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Want To Reduce Labeling Cost? GPT-3 Can Help","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Chenguang Zhu, Michael Zeng, Shuohang Wang, Yang Liu, Yichong Xu","submitted_at":"2021-08-30T19:18:24Z","abstract_excerpt":"Data annotation is a time-consuming and labor-intensive process for many NLP tasks. Although there exist various methods to produce pseudo data labels, they are often task-specific and require a decent amount of labeled data to start with. Recently, the immense language model GPT-3 with 175 billion parameters has achieved tremendous improvement across many few-shot learning tasks. In this paper, we explore ways to leverage GPT-3 as a low-cost data labeler to train other models. We find that, to make the downstream model achieve the same performance on a variety of NLU and NLG tasks, it costs 5"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2108.13487","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/2108.13487/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":"2108.13487","created_at":"2026-07-05T03:10:14.687134+00:00"},{"alias_kind":"arxiv_version","alias_value":"2108.13487v1","created_at":"2026-07-05T03:10:14.687134+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2108.13487","created_at":"2026-07-05T03:10:14.687134+00:00"},{"alias_kind":"pith_short_12","alias_value":"437CQS27UKPF","created_at":"2026-07-05T03:10:14.687134+00:00"},{"alias_kind":"pith_short_16","alias_value":"437CQS27UKPFZBP6","created_at":"2026-07-05T03:10:14.687134+00:00"},{"alias_kind":"pith_short_8","alias_value":"437CQS27","created_at":"2026-07-05T03:10:14.687134+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.04273","citing_title":"Characterizing initial human-AI proof formalization workflows","ref_index":255,"is_internal_anchor":false},{"citing_arxiv_id":"2305.02301","citing_title":"Distilling Step-by-Step! Outperforming Larger Language Models with Less Training Data and Smaller Model Sizes","ref_index":105,"is_internal_anchor":false},{"citing_arxiv_id":"2605.20809","citing_title":"Refining and Reusing Annotation Guidelines for LLM Annotation","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2604.08578","citing_title":"Structured Exploration and Exploitation of Label Functions for Automated Data Annotation","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2604.17037","citing_title":"Dynamic Emotion and Personality Profiling for Multimodal Deception Detection","ref_index":30,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/437CQS27UKPFZBP6CL5EN5NAXT","json":"https://pith.science/pith/437CQS27UKPFZBP6CL5EN5NAXT.json","graph_json":"https://pith.science/api/pith-number/437CQS27UKPFZBP6CL5EN5NAXT/graph.json","events_json":"https://pith.science/api/pith-number/437CQS27UKPFZBP6CL5EN5NAXT/events.json","paper":"https://pith.science/paper/437CQS27"},"agent_actions":{"view_html":"https://pith.science/pith/437CQS27UKPFZBP6CL5EN5NAXT","download_json":"https://pith.science/pith/437CQS27UKPFZBP6CL5EN5NAXT.json","view_paper":"https://pith.science/paper/437CQS27","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2108.13487&json=true","fetch_graph":"https://pith.science/api/pith-number/437CQS27UKPFZBP6CL5EN5NAXT/graph.json","fetch_events":"https://pith.science/api/pith-number/437CQS27UKPFZBP6CL5EN5NAXT/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/437CQS27UKPFZBP6CL5EN5NAXT/action/timestamp_anchor","attest_storage":"https://pith.science/pith/437CQS27UKPFZBP6CL5EN5NAXT/action/storage_attestation","attest_author":"https://pith.science/pith/437CQS27UKPFZBP6CL5EN5NAXT/action/author_attestation","sign_citation":"https://pith.science/pith/437CQS27UKPFZBP6CL5EN5NAXT/action/citation_signature","submit_replication":"https://pith.science/pith/437CQS27UKPFZBP6CL5EN5NAXT/action/replication_record"}},"created_at":"2026-07-05T03:10:14.687134+00:00","updated_at":"2026-07-05T03:10:14.687134+00:00"}