{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:45NSUTS75WRT6WJYIBUFZWMS5E","short_pith_number":"pith:45NSUTS7","schema_version":"1.0","canonical_sha256":"e75b2a4e5feda33f593840685cd992e912b635900f829633e716cc79fba80d72","source":{"kind":"arxiv","id":"2303.15056","version":2},"attestation_state":"computed","paper":{"title":"ChatGPT Outperforms Crowd-Workers for Text-Annotation Tasks","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CY"],"primary_cat":"cs.CL","authors_text":"Fabrizio Gilardi, Ma\\\"el Kubli, Meysam Alizadeh","submitted_at":"2023-03-27T09:59:48Z","abstract_excerpt":"Many NLP applications require manual data annotations for a variety of tasks, notably to train classifiers or evaluate the performance of unsupervised models. Depending on the size and degree of complexity, the tasks may be conducted by crowd-workers on platforms such as MTurk as well as trained annotators, such as research assistants. Using a sample of 2,382 tweets, we demonstrate that ChatGPT outperforms crowd-workers for several annotation tasks, including relevance, stance, topics, and frames detection. Specifically, the zero-shot accuracy of ChatGPT exceeds that of crowd-workers for four "},"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":"2303.15056","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2023-03-27T09:59:48Z","cross_cats_sorted":["cs.CY"],"title_canon_sha256":"9b6f8b476a7b541dfb67d3b0e305c154896ba1d2c4557200e2d9bc4419fac2b7","abstract_canon_sha256":"1e367e6fcd5a7ea1758be9819527aac9d60ad558697e8a70b303b9f23240bec9"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T06:32:37.995484Z","signature_b64":"/04x/KrRQ/cHEg6VEZv9POlnCubX2lEGgBnteHO9L5R3w2HZKo3IF/LnFgURRmPcFy4aFFgo+mH/69i45HENCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e75b2a4e5feda33f593840685cd992e912b635900f829633e716cc79fba80d72","last_reissued_at":"2026-07-05T06:32:37.995001Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T06:32:37.995001Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"ChatGPT Outperforms Crowd-Workers for Text-Annotation Tasks","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CY"],"primary_cat":"cs.CL","authors_text":"Fabrizio Gilardi, Ma\\\"el Kubli, Meysam Alizadeh","submitted_at":"2023-03-27T09:59:48Z","abstract_excerpt":"Many NLP applications require manual data annotations for a variety of tasks, notably to train classifiers or evaluate the performance of unsupervised models. Depending on the size and degree of complexity, the tasks may be conducted by crowd-workers on platforms such as MTurk as well as trained annotators, such as research assistants. Using a sample of 2,382 tweets, we demonstrate that ChatGPT outperforms crowd-workers for several annotation tasks, including relevance, stance, topics, and frames detection. Specifically, the zero-shot accuracy of ChatGPT exceeds that of crowd-workers for four "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2303.15056","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/2303.15056/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":"2303.15056","created_at":"2026-07-05T06:32:37.995064+00:00"},{"alias_kind":"arxiv_version","alias_value":"2303.15056v2","created_at":"2026-07-05T06:32:37.995064+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2303.15056","created_at":"2026-07-05T06:32:37.995064+00:00"},{"alias_kind":"pith_short_12","alias_value":"45NSUTS75WRT","created_at":"2026-07-05T06:32:37.995064+00:00"},{"alias_kind":"pith_short_16","alias_value":"45NSUTS75WRT6WJY","created_at":"2026-07-05T06:32:37.995064+00:00"},{"alias_kind":"pith_short_8","alias_value":"45NSUTS7","created_at":"2026-07-05T06:32:37.995064+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":15,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.26749","citing_title":"Structure Before Collapse: Transient semantic geometry in next-token prediction","ref_index":76,"is_internal_anchor":false},{"citing_arxiv_id":"2606.30175","citing_title":"CORTEX: High-Quality Cross-Domain Organization of Web-Scale Corpora through Ontological Corpus Graph","ref_index":77,"is_internal_anchor":false},{"citing_arxiv_id":"2605.28802","citing_title":"Human Label Variation as Stable Signal: Learning Annotator-Specific Explanation Behavior via Cross-Annotator Preference Optimization","ref_index":2,"is_internal_anchor":false},{"citing_arxiv_id":"2508.15503","citing_title":"Guidelines for Empirical Studies in Software Engineering involving Large Language Models","ref_index":44,"is_internal_anchor":false},{"citing_arxiv_id":"2605.17152","citing_title":"Multilingual and Multimodal LLMs in the Wild: Building for Low-Resource Languages","ref_index":150,"is_internal_anchor":false},{"citing_arxiv_id":"2508.15503","citing_title":"Guidelines for Empirical Studies in Software Engineering involving Large Language Models","ref_index":44,"is_internal_anchor":false},{"citing_arxiv_id":"2408.00724","citing_title":"Inference Scaling Laws: An Empirical Analysis of Compute-Optimal Inference for Problem-Solving with Language Models","ref_index":287,"is_internal_anchor":false},{"citing_arxiv_id":"2309.11495","citing_title":"Chain-of-Verification Reduces Hallucination in Large Language Models","ref_index":140,"is_internal_anchor":false},{"citing_arxiv_id":"2409.12917","citing_title":"Training Language Models to Self-Correct via Reinforcement Learning","ref_index":96,"is_internal_anchor":false},{"citing_arxiv_id":"2309.00267","citing_title":"RLAIF vs. RLHF: Scaling Reinforcement Learning from Human Feedback with AI Feedback","ref_index":90,"is_internal_anchor":false},{"citing_arxiv_id":"2305.14233","citing_title":"Enhancing Chat Language Models by Scaling High-quality Instructional Conversations","ref_index":240,"is_internal_anchor":false},{"citing_arxiv_id":"2306.14565","citing_title":"Mitigating Hallucination in Large Multi-Modal Models via Robust Instruction Tuning","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2304.08485","citing_title":"Visual Instruction Tuning","ref_index":17,"is_internal_anchor":false},{"citing_arxiv_id":"2306.05685","citing_title":"Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena","ref_index":17,"is_internal_anchor":false},{"citing_arxiv_id":"2604.13899","citing_title":"Do We Still Need Humans in the Loop? Comparing Human and LLM Annotation in Active Learning for Hostility Detection","ref_index":3,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/45NSUTS75WRT6WJYIBUFZWMS5E","json":"https://pith.science/pith/45NSUTS75WRT6WJYIBUFZWMS5E.json","graph_json":"https://pith.science/api/pith-number/45NSUTS75WRT6WJYIBUFZWMS5E/graph.json","events_json":"https://pith.science/api/pith-number/45NSUTS75WRT6WJYIBUFZWMS5E/events.json","paper":"https://pith.science/paper/45NSUTS7"},"agent_actions":{"view_html":"https://pith.science/pith/45NSUTS75WRT6WJYIBUFZWMS5E","download_json":"https://pith.science/pith/45NSUTS75WRT6WJYIBUFZWMS5E.json","view_paper":"https://pith.science/paper/45NSUTS7","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2303.15056&json=true","fetch_graph":"https://pith.science/api/pith-number/45NSUTS75WRT6WJYIBUFZWMS5E/graph.json","fetch_events":"https://pith.science/api/pith-number/45NSUTS75WRT6WJYIBUFZWMS5E/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/45NSUTS75WRT6WJYIBUFZWMS5E/action/timestamp_anchor","attest_storage":"https://pith.science/pith/45NSUTS75WRT6WJYIBUFZWMS5E/action/storage_attestation","attest_author":"https://pith.science/pith/45NSUTS75WRT6WJYIBUFZWMS5E/action/author_attestation","sign_citation":"https://pith.science/pith/45NSUTS75WRT6WJYIBUFZWMS5E/action/citation_signature","submit_replication":"https://pith.science/pith/45NSUTS75WRT6WJYIBUFZWMS5E/action/replication_record"}},"created_at":"2026-07-05T06:32:37.995064+00:00","updated_at":"2026-07-05T06:32:37.995064+00:00"}