{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:72OXHZG3EFBAUKCJR4IITTSUNS","short_pith_number":"pith:72OXHZG3","schema_version":"1.0","canonical_sha256":"fe9d73e4db21420a28498f1089ce546c85cfc0b79ddf4535ef83241f2cce83ad","source":{"kind":"arxiv","id":"2303.12767","version":2},"attestation_state":"computed","paper":{"title":"Can we trust the evaluation on ChatGPT?","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Haewoon Kwak, Jisun An, Rachith Aiyappa, Yong-Yeol Ahn","submitted_at":"2023-03-22T17:32:56Z","abstract_excerpt":"ChatGPT, the first large language model (LLM) with mass adoption, has demonstrated remarkable performance in numerous natural language tasks. Despite its evident usefulness, evaluating ChatGPT's performance in diverse problem domains remains challenging due to the closed nature of the model and its continuous updates via Reinforcement Learning from Human Feedback (RLHF). We highlight the issue of data contamination in ChatGPT evaluations, with a case study of the task of stance detection. We discuss the challenge of preventing data contamination and ensuring fair model evaluation in the age of"},"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":"2303.12767","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2023-03-22T17:32:56Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"055904088b3d4825ed2036a8281e86194ef6ea92f0d62e043017a053e6f6aabc","abstract_canon_sha256":"5581ba4d732af0e9c589455c9eab07572380937f45e99e594ad16e213061a076"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:57:51.519670Z","signature_b64":"Bb4alL8nQum9qFZKGOyo0LdTKLq0+5KK7R4Md14t2RNjddbIYJj2wcr7vmDyYMfluBNPssKcocIgtkDen0amDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"fe9d73e4db21420a28498f1089ce546c85cfc0b79ddf4535ef83241f2cce83ad","last_reissued_at":"2026-07-05T08:57:51.519201Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:57:51.519201Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Can we trust the evaluation on ChatGPT?","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Haewoon Kwak, Jisun An, Rachith Aiyappa, Yong-Yeol Ahn","submitted_at":"2023-03-22T17:32:56Z","abstract_excerpt":"ChatGPT, the first large language model (LLM) with mass adoption, has demonstrated remarkable performance in numerous natural language tasks. Despite its evident usefulness, evaluating ChatGPT's performance in diverse problem domains remains challenging due to the closed nature of the model and its continuous updates via Reinforcement Learning from Human Feedback (RLHF). We highlight the issue of data contamination in ChatGPT evaluations, with a case study of the task of stance detection. We discuss the challenge of preventing data contamination and ensuring fair model evaluation in the age of"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2303.12767","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/2303.12767/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":"2303.12767","created_at":"2026-07-05T08:57:51.519262+00:00"},{"alias_kind":"arxiv_version","alias_value":"2303.12767v2","created_at":"2026-07-05T08:57:51.519262+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2303.12767","created_at":"2026-07-05T08:57:51.519262+00:00"},{"alias_kind":"pith_short_12","alias_value":"72OXHZG3EFBA","created_at":"2026-07-05T08:57:51.519262+00:00"},{"alias_kind":"pith_short_16","alias_value":"72OXHZG3EFBAUKCJ","created_at":"2026-07-05T08:57:51.519262+00:00"},{"alias_kind":"pith_short_8","alias_value":"72OXHZG3","created_at":"2026-07-05T08:57:51.519262+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":6,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.26749","citing_title":"Structure Before Collapse: Transient semantic geometry in next-token prediction","ref_index":89,"is_internal_anchor":false},{"citing_arxiv_id":"2305.09620","citing_title":"AI-Augmented Surveys: Leveraging Large Language Models and Surveys for Opinion Prediction","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2401.00761","citing_title":"Identifying the Achilles' Heel: An Iterative Method for Dynamically Uncovering Factual Errors in Large Language Models","ref_index":1,"is_internal_anchor":false},{"citing_arxiv_id":"2309.11495","citing_title":"Chain-of-Verification Reduces Hallucination in Large Language Models","ref_index":92,"is_internal_anchor":false},{"citing_arxiv_id":"2402.13228","citing_title":"Smaug: Fixing Failure Modes of Preference Optimisation with DPO-Positive","ref_index":137,"is_internal_anchor":false},{"citing_arxiv_id":"2604.24819","citing_title":"Programming with Data: Test-Driven Data Engineering for Self-Improving LLMs from Raw Corpora","ref_index":1,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/72OXHZG3EFBAUKCJR4IITTSUNS","json":"https://pith.science/pith/72OXHZG3EFBAUKCJR4IITTSUNS.json","graph_json":"https://pith.science/api/pith-number/72OXHZG3EFBAUKCJR4IITTSUNS/graph.json","events_json":"https://pith.science/api/pith-number/72OXHZG3EFBAUKCJR4IITTSUNS/events.json","paper":"https://pith.science/paper/72OXHZG3"},"agent_actions":{"view_html":"https://pith.science/pith/72OXHZG3EFBAUKCJR4IITTSUNS","download_json":"https://pith.science/pith/72OXHZG3EFBAUKCJR4IITTSUNS.json","view_paper":"https://pith.science/paper/72OXHZG3","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2303.12767&json=true","fetch_graph":"https://pith.science/api/pith-number/72OXHZG3EFBAUKCJR4IITTSUNS/graph.json","fetch_events":"https://pith.science/api/pith-number/72OXHZG3EFBAUKCJR4IITTSUNS/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/72OXHZG3EFBAUKCJR4IITTSUNS/action/timestamp_anchor","attest_storage":"https://pith.science/pith/72OXHZG3EFBAUKCJR4IITTSUNS/action/storage_attestation","attest_author":"https://pith.science/pith/72OXHZG3EFBAUKCJR4IITTSUNS/action/author_attestation","sign_citation":"https://pith.science/pith/72OXHZG3EFBAUKCJR4IITTSUNS/action/citation_signature","submit_replication":"https://pith.science/pith/72OXHZG3EFBAUKCJR4IITTSUNS/action/replication_record"}},"created_at":"2026-07-05T08:57:51.519262+00:00","updated_at":"2026-07-05T08:57:51.519262+00:00"}