{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:4WDAVJQFULQZT6WZVMBO3ORO3G","short_pith_number":"pith:4WDAVJQF","schema_version":"1.0","canonical_sha256":"e5860aa605a2e199fad9ab02edba2ed993cd10a4c96bfb3db60440356bdd6c62","source":{"kind":"arxiv","id":"2103.12407","version":4},"attestation_state":"computed","paper":{"title":"Detecting Hate Speech with GPT-3","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Annie Collins, Ke-Li Chiu, Rohan Alexander","submitted_at":"2021-03-23T09:17:22Z","abstract_excerpt":"Sophisticated language models such as OpenAI's GPT-3 can generate hateful text that targets marginalized groups. Given this capacity, we are interested in whether large language models can be used to identify hate speech and classify text as sexist or racist. We use GPT-3 to identify sexist and racist text passages with zero-, one-, and few-shot learning. We find that with zero- and one-shot learning, GPT-3 can identify sexist or racist text with an average accuracy between 55 per cent and 67 per cent, depending on the category of text and type of learning. With few-shot learning, the model's "},"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":"2103.12407","kind":"arxiv","version":4},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2021-03-23T09:17:22Z","cross_cats_sorted":[],"title_canon_sha256":"d7f12eebb7f321cfbfe4553e12bdf6d51fceb0198c876f2524c8e9f9e8d9af57","abstract_canon_sha256":"874083c98cb3c891a6233cdbe5476c2d17406f140c6dbb98eea66a8dbe517ad9"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T04:08:14.772204Z","signature_b64":"RF1PrHDxOrJsVFLmKVdqrD3SizxJNsmjFq/2+QQ7WzE+KUYIl5NYsi9UeccdOv0YPMvBAUsyldJIBsApu28hDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e5860aa605a2e199fad9ab02edba2ed993cd10a4c96bfb3db60440356bdd6c62","last_reissued_at":"2026-07-05T04:08:14.771731Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T04:08:14.771731Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Detecting Hate Speech with GPT-3","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Annie Collins, Ke-Li Chiu, Rohan Alexander","submitted_at":"2021-03-23T09:17:22Z","abstract_excerpt":"Sophisticated language models such as OpenAI's GPT-3 can generate hateful text that targets marginalized groups. Given this capacity, we are interested in whether large language models can be used to identify hate speech and classify text as sexist or racist. We use GPT-3 to identify sexist and racist text passages with zero-, one-, and few-shot learning. We find that with zero- and one-shot learning, GPT-3 can identify sexist or racist text with an average accuracy between 55 per cent and 67 per cent, depending on the category of text and type of learning. With few-shot learning, the model's "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2103.12407","kind":"arxiv","version":4},"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/2103.12407/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":"2103.12407","created_at":"2026-07-05T04:08:14.771789+00:00"},{"alias_kind":"arxiv_version","alias_value":"2103.12407v4","created_at":"2026-07-05T04:08:14.771789+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2103.12407","created_at":"2026-07-05T04:08:14.771789+00:00"},{"alias_kind":"pith_short_12","alias_value":"4WDAVJQFULQZ","created_at":"2026-07-05T04:08:14.771789+00:00"},{"alias_kind":"pith_short_16","alias_value":"4WDAVJQFULQZT6WZ","created_at":"2026-07-05T04:08:14.771789+00:00"},{"alias_kind":"pith_short_8","alias_value":"4WDAVJQF","created_at":"2026-07-05T04:08:14.771789+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2408.12935","citing_title":"AI Safety Landscape for Large Language Models: Taxonomy, State-of-the-art, and Future Directions","ref_index":136,"is_internal_anchor":false},{"citing_arxiv_id":"2605.20584","citing_title":"QwenSafe: Multimodal Content Rating Description Identification via Preference-Aligned VLMs","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2205.01068","citing_title":"OPT: Open Pre-trained Transformer Language Models","ref_index":113,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/4WDAVJQFULQZT6WZVMBO3ORO3G","json":"https://pith.science/pith/4WDAVJQFULQZT6WZVMBO3ORO3G.json","graph_json":"https://pith.science/api/pith-number/4WDAVJQFULQZT6WZVMBO3ORO3G/graph.json","events_json":"https://pith.science/api/pith-number/4WDAVJQFULQZT6WZVMBO3ORO3G/events.json","paper":"https://pith.science/paper/4WDAVJQF"},"agent_actions":{"view_html":"https://pith.science/pith/4WDAVJQFULQZT6WZVMBO3ORO3G","download_json":"https://pith.science/pith/4WDAVJQFULQZT6WZVMBO3ORO3G.json","view_paper":"https://pith.science/paper/4WDAVJQF","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2103.12407&json=true","fetch_graph":"https://pith.science/api/pith-number/4WDAVJQFULQZT6WZVMBO3ORO3G/graph.json","fetch_events":"https://pith.science/api/pith-number/4WDAVJQFULQZT6WZVMBO3ORO3G/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/4WDAVJQFULQZT6WZVMBO3ORO3G/action/timestamp_anchor","attest_storage":"https://pith.science/pith/4WDAVJQFULQZT6WZVMBO3ORO3G/action/storage_attestation","attest_author":"https://pith.science/pith/4WDAVJQFULQZT6WZVMBO3ORO3G/action/author_attestation","sign_citation":"https://pith.science/pith/4WDAVJQFULQZT6WZVMBO3ORO3G/action/citation_signature","submit_replication":"https://pith.science/pith/4WDAVJQFULQZT6WZVMBO3ORO3G/action/replication_record"}},"created_at":"2026-07-05T04:08:14.771789+00:00","updated_at":"2026-07-05T04:08:14.771789+00:00"}