{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:TLVLBZXLXPUP3MMLZVXCE4F6WJ","short_pith_number":"pith:TLVLBZXL","schema_version":"1.0","canonical_sha256":"9aeab0e6ebbbe8fdb18bcd6e2270beb25f386b7378992db39b4145729627175e","source":{"kind":"arxiv","id":"2405.01470","version":1},"attestation_state":"computed","paper":{"title":"WildChat: 1M ChatGPT Interaction Logs in the Wild","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"A corpus of one million real ChatGPT conversations was assembled from users who opted in for free access.","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Claire Cardie, Jack Hessel, Wenting Zhao, Xiang Ren, Yejin Choi, Yuntian Deng","submitted_at":"2024-05-02T17:00:02Z","abstract_excerpt":"Chatbots such as GPT-4 and ChatGPT are now serving millions of users. Despite their widespread use, there remains a lack of public datasets showcasing how these tools are used by a population of users in practice. To bridge this gap, we offered free access to ChatGPT for online users in exchange for their affirmative, consensual opt-in to anonymously collect their chat transcripts and request headers. From this, we compiled WildChat, a corpus of 1 million user-ChatGPT conversations, which consists of over 2.5 million interaction turns. We compare WildChat with other popular user-chatbot intera"},"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":true,"formal_links_present":false},"canonical_record":{"source":{"id":"2405.01470","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-05-02T17:00:02Z","cross_cats_sorted":[],"title_canon_sha256":"050f37ac94c115de31819627bb2d1efd7df3e488fcfb6923270c7f118167b764","abstract_canon_sha256":"bbea487f1520b1acdffe8c1ab90cb28c38e1c7a4e1eda33deb94693d5ca43609"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-17T23:38:48.151870Z","signature_b64":"9oDpxXCl2at1tZzzCVL/Ixx30kRE4637wJPaD1z/8yCe+gJklD4nNGt+1g2Y/mObwMF84I35UeEe0QHSi4BqBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"9aeab0e6ebbbe8fdb18bcd6e2270beb25f386b7378992db39b4145729627175e","last_reissued_at":"2026-05-17T23:38:48.151382Z","signature_status":"signed_v1","first_computed_at":"2026-05-17T23:38:48.151382Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"WildChat: 1M ChatGPT Interaction Logs in the Wild","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"A corpus of one million real ChatGPT conversations was assembled from users who opted in for free access.","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Claire Cardie, Jack Hessel, Wenting Zhao, Xiang Ren, Yejin Choi, Yuntian Deng","submitted_at":"2024-05-02T17:00:02Z","abstract_excerpt":"Chatbots such as GPT-4 and ChatGPT are now serving millions of users. Despite their widespread use, there remains a lack of public datasets showcasing how these tools are used by a population of users in practice. To bridge this gap, we offered free access to ChatGPT for online users in exchange for their affirmative, consensual opt-in to anonymously collect their chat transcripts and request headers. From this, we compiled WildChat, a corpus of 1 million user-ChatGPT conversations, which consists of over 2.5 million interaction turns. We compare WildChat with other popular user-chatbot intera"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"we compiled WildChat, a corpus of 1 million user-ChatGPT conversations... offers the most diverse user prompts, contains the largest number of languages, and presents the richest variety of potentially toxic use-cases for researchers to study.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The opt-in consent process produces a representative sample of ChatGPT users without major selection bias from users who choose to participate or from the free-access incentive.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"WildChat releases a dataset of 1 million ChatGPT conversations with timestamps, demographics, and headers, claimed to be the most diverse and multilingual such resource available.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"A corpus of one million real ChatGPT conversations was assembled from users who opted in for free access.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"f73a3f640728c594acc0f0d03d22c617a5e4893a169a212d3dc909226787b862"},"source":{"id":"2405.01470","kind":"arxiv","version":1},"verdict":{"id":"71d93274-81a8-481b-a528-8c289477e40d","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-16T10:40:02.109654Z","strongest_claim":"we compiled WildChat, a corpus of 1 million user-ChatGPT conversations... offers the most diverse user prompts, contains the largest number of languages, and presents the richest variety of potentially toxic use-cases for researchers to study.","one_line_summary":"WildChat releases a dataset of 1 million ChatGPT conversations with timestamps, demographics, and headers, claimed to be the most diverse and multilingual such resource available.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The opt-in consent process produces a representative sample of ChatGPT users without major selection bias from users who choose to participate or from the free-access incentive.","pith_extraction_headline":"A corpus of one million real ChatGPT conversations was assembled from users who opted in for free access."},"references":{"count":54,"sample":[{"doi":"","year":2023,"title":"Model card and evaluations for claude models, Jul 2023","work_id":"c66a93f2-bf62-4115-b48b-7f13aff667f3","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2022,"title":"Bowman, Zac Hatfield-Dodds, Ben Mann, Dario Amodei, Nicholas Joseph, Sam McCandlish, Tom Brown, and Jared Kaplan","work_id":"19bc1946-1565-4328-92ea-a230a57a8bbf","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"Free dolly: Introducing the world’s first truly open instruction-tuned llm, 2023","work_id":"ab41d2b1-6296-4ba7-9db5-c040d5310c3b","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"Bard: A conversational ai tool by google, 2023","work_id":"d5be1bf4-947a-4e2c-9cd2-60f9d14ecfd2","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2020,"title":"Laura Hanu and Unitary team . Detoxify. Github. https://github.com/unitaryai/detoxify, 2020","work_id":"a383ee20-a51e-4a32-a272-058028fdf086","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":54,"snapshot_sha256":"eac8c3d4ebb3dd7dce7d88f4d040e1a4287d8c941d94479142d0986bba471875","internal_anchors":1},"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":"2405.01470","created_at":"2026-05-17T23:38:48.151464+00:00"},{"alias_kind":"arxiv_version","alias_value":"2405.01470v1","created_at":"2026-05-17T23:38:48.151464+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2405.01470","created_at":"2026-05-17T23:38:48.151464+00:00"},{"alias_kind":"pith_short_12","alias_value":"TLVLBZXLXPUP","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_16","alias_value":"TLVLBZXLXPUP3MML","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_8","alias_value":"TLVLBZXL","created_at":"2026-05-18T12:33:37.589309+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":29,"internal_anchor_count":29,"sample":[{"citing_arxiv_id":"2408.15549","citing_title":"WildFeedback: Aligning LLMs With In-situ User Interactions And Feedback","ref_index":41,"is_internal_anchor":true},{"citing_arxiv_id":"2412.16720","citing_title":"OpenAI o1 System Card","ref_index":13,"is_internal_anchor":true},{"citing_arxiv_id":"2503.02574","citing_title":"LLM-Safety Evaluations Lack Robustness","ref_index":61,"is_internal_anchor":true},{"citing_arxiv_id":"2601.14348","citing_title":"Legal Retrieval for Public Defenders","ref_index":43,"is_internal_anchor":true},{"citing_arxiv_id":"2603.23231","citing_title":"PERMA: Benchmarking Personalized Memory Agents via Event-Driven Preference and Realistic Task Environments","ref_index":86,"is_internal_anchor":true},{"citing_arxiv_id":"2605.20654","citing_title":"REFLECTOR: Internalizing Step-wise Reflection against Indirect Jailbreak","ref_index":18,"is_internal_anchor":true},{"citing_arxiv_id":"2605.13339","citing_title":"Probing Persona-Dependent Preferences in Language Models","ref_index":35,"is_internal_anchor":true},{"citing_arxiv_id":"2506.01937","citing_title":"RewardBench 2: Advancing Reward Model Evaluation","ref_index":15,"is_internal_anchor":true},{"citing_arxiv_id":"2506.18841","citing_title":"LongWriter-Zero: Mastering Ultra-Long Text Generation via Reinforcement Learning","ref_index":49,"is_internal_anchor":true},{"citing_arxiv_id":"2510.24235","citing_title":"PaTaRM: Bridging Pairwise and Pointwise Signals via Preference-Aware Task-Adaptive Reward Modeling","ref_index":24,"is_internal_anchor":true},{"citing_arxiv_id":"2512.20856","citing_title":"NVIDIA Nemotron 3: Efficient and Open Intelligence","ref_index":20,"is_internal_anchor":true},{"citing_arxiv_id":"2601.21351","citing_title":"Analytical Provisioning for Attention-FFN Disaggregated LLM Serving under Stochastic Workloads","ref_index":12,"is_internal_anchor":true},{"citing_arxiv_id":"2603.03295","citing_title":"Language Model Goal Selection Differs from Humans' in a Self-Directed Learning Task","ref_index":22,"is_internal_anchor":true},{"citing_arxiv_id":"2605.02897","citing_title":"Same Voice, Different Lab: On the Homogenization of Frontier LLM Personalities","ref_index":6,"is_internal_anchor":true},{"citing_arxiv_id":"2605.14175","citing_title":"Grounded Continuation: A Linear-Time Runtime Verifier for LLM Conversations","ref_index":41,"is_internal_anchor":true},{"citing_arxiv_id":"2604.02686","citing_title":"Beyond Semantic Manipulation: Token-Space Attacks on Reward Models","ref_index":20,"is_internal_anchor":true},{"citing_arxiv_id":"2605.11093","citing_title":"Enabling Performant and Flexible Model-Internal Observability for LLM Inference","ref_index":50,"is_internal_anchor":true},{"citing_arxiv_id":"2605.04029","citing_title":"Stayin' Aligned Over Time: Towards Longitudinal Human-LLM Alignment via Contextual Reflection and Privacy-Preserving Behavioral Data","ref_index":42,"is_internal_anchor":true},{"citing_arxiv_id":"2604.27039","citing_title":"Length Value Model: Scalable Value Pretraining for Token-Level Length Modeling","ref_index":33,"is_internal_anchor":true},{"citing_arxiv_id":"2605.09808","citing_title":"Quantifying the Utility of User Simulators for Building Collaborative LLM Assistants","ref_index":35,"is_internal_anchor":true},{"citing_arxiv_id":"2605.09995","citing_title":"Annotations Mitigate Post-Training Mode Collapse","ref_index":59,"is_internal_anchor":true},{"citing_arxiv_id":"2604.25080","citing_title":"CacheFlow: Efficient LLM Serving with 3D-Parallel KV Cache Restoration","ref_index":8,"is_internal_anchor":true},{"citing_arxiv_id":"2604.25905","citing_title":"A paradox of AI fluency","ref_index":40,"is_internal_anchor":true},{"citing_arxiv_id":"2605.05777","citing_title":"Estimating the Black-box LLM Uncertainty with Distribution-Aligned Adversarial Distillation","ref_index":34,"is_internal_anchor":true},{"citing_arxiv_id":"2605.05678","citing_title":"Chain of Risk: Safety Failures in Large Reasoning Models and Mitigation via Adaptive Multi-Principle Steering","ref_index":35,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/TLVLBZXLXPUP3MMLZVXCE4F6WJ","json":"https://pith.science/pith/TLVLBZXLXPUP3MMLZVXCE4F6WJ.json","graph_json":"https://pith.science/api/pith-number/TLVLBZXLXPUP3MMLZVXCE4F6WJ/graph.json","events_json":"https://pith.science/api/pith-number/TLVLBZXLXPUP3MMLZVXCE4F6WJ/events.json","paper":"https://pith.science/paper/TLVLBZXL"},"agent_actions":{"view_html":"https://pith.science/pith/TLVLBZXLXPUP3MMLZVXCE4F6WJ","download_json":"https://pith.science/pith/TLVLBZXLXPUP3MMLZVXCE4F6WJ.json","view_paper":"https://pith.science/paper/TLVLBZXL","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2405.01470&json=true","fetch_graph":"https://pith.science/api/pith-number/TLVLBZXLXPUP3MMLZVXCE4F6WJ/graph.json","fetch_events":"https://pith.science/api/pith-number/TLVLBZXLXPUP3MMLZVXCE4F6WJ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/TLVLBZXLXPUP3MMLZVXCE4F6WJ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/TLVLBZXLXPUP3MMLZVXCE4F6WJ/action/storage_attestation","attest_author":"https://pith.science/pith/TLVLBZXLXPUP3MMLZVXCE4F6WJ/action/author_attestation","sign_citation":"https://pith.science/pith/TLVLBZXLXPUP3MMLZVXCE4F6WJ/action/citation_signature","submit_replication":"https://pith.science/pith/TLVLBZXLXPUP3MMLZVXCE4F6WJ/action/replication_record"}},"created_at":"2026-05-17T23:38:48.151464+00:00","updated_at":"2026-05-17T23:38:48.151464+00:00"}