{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2019:R77CAC2QRSBJBAJT4B6Y3DWHQI","short_pith_number":"pith:R77CAC2Q","schema_version":"1.0","canonical_sha256":"8ffe200b508c82908133e07d8d8ec78223ca34ec8faf1a67c6ee3599358597a9","source":{"kind":"arxiv","id":"1911.00650","version":2},"attestation_state":"computed","paper":{"title":"Automatic Detection of Generated Text is Easiest when Humans are Fooled","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Chris Callison-Burch, Daniel Duckworth, Daphne Ippolito, Douglas Eck","submitted_at":"2019-11-02T04:52:00Z","abstract_excerpt":"Recent advancements in neural language modelling make it possible to rapidly generate vast amounts of human-sounding text. The capabilities of humans and automatic discriminators to detect machine-generated text have been a large source of research interest, but humans and machines rely on different cues to make their decisions. Here, we perform careful benchmarking and analysis of three popular sampling-based decoding strategies---top-$k$, nucleus sampling, and untruncated random sampling---and show that improvements in decoding methods have primarily optimized for fooling humans. This comes "},"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":"1911.00650","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2019-11-02T04:52:00Z","cross_cats_sorted":[],"title_canon_sha256":"126e4a78ba8628ad69bb28ec3c66e2683ad260ad5279761faec4c57eed61e85b","abstract_canon_sha256":"18fa690a319e60676abdb46c288ecc5815505e6a214915d90f44f66e7d6cd295"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T01:01:21.343417Z","signature_b64":"mcwRyEnSLK8NejwshhxZWyDzY7PdylMCJkjKUiemDA9rlqKmRvYGmMflnyE8D2buiTAt2Ps4+zu0Ux34JFy7DA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8ffe200b508c82908133e07d8d8ec78223ca34ec8faf1a67c6ee3599358597a9","last_reissued_at":"2026-07-05T01:01:21.342965Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T01:01:21.342965Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Automatic Detection of Generated Text is Easiest when Humans are Fooled","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Chris Callison-Burch, Daniel Duckworth, Daphne Ippolito, Douglas Eck","submitted_at":"2019-11-02T04:52:00Z","abstract_excerpt":"Recent advancements in neural language modelling make it possible to rapidly generate vast amounts of human-sounding text. The capabilities of humans and automatic discriminators to detect machine-generated text have been a large source of research interest, but humans and machines rely on different cues to make their decisions. Here, we perform careful benchmarking and analysis of three popular sampling-based decoding strategies---top-$k$, nucleus sampling, and untruncated random sampling---and show that improvements in decoding methods have primarily optimized for fooling humans. This comes "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1911.00650","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/1911.00650/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":"1911.00650","created_at":"2026-07-05T01:01:21.343020+00:00"},{"alias_kind":"arxiv_version","alias_value":"1911.00650v2","created_at":"2026-07-05T01:01:21.343020+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1911.00650","created_at":"2026-07-05T01:01:21.343020+00:00"},{"alias_kind":"pith_short_12","alias_value":"R77CAC2QRSBJ","created_at":"2026-07-05T01:01:21.343020+00:00"},{"alias_kind":"pith_short_16","alias_value":"R77CAC2QRSBJBAJT","created_at":"2026-07-05T01:01:21.343020+00:00"},{"alias_kind":"pith_short_8","alias_value":"R77CAC2Q","created_at":"2026-07-05T01:01:21.343020+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":6,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2607.01867","citing_title":"An Exploratory Study on LLM-Generated Code and Comments in Code Repositories","ref_index":48,"is_internal_anchor":false},{"citing_arxiv_id":"2605.20761","citing_title":"Findings of the Counter Turing Test: AI-Generated Text Detection","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2605.20761","citing_title":"Findings of the Counter Turing Test: AI-Generated Text Detection","ref_index":12,"is_internal_anchor":false},{"citing_arxiv_id":"2303.11156","citing_title":"Can AI-Generated Text be Reliably Detected?","ref_index":90,"is_internal_anchor":false},{"citing_arxiv_id":"2511.12468","citing_title":"Detecting LLM-Assisted Academic Dishonesty using Keystroke Dynamics","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2005.14165","citing_title":"Language Models are Few-Shot Learners","ref_index":26,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/R77CAC2QRSBJBAJT4B6Y3DWHQI","json":"https://pith.science/pith/R77CAC2QRSBJBAJT4B6Y3DWHQI.json","graph_json":"https://pith.science/api/pith-number/R77CAC2QRSBJBAJT4B6Y3DWHQI/graph.json","events_json":"https://pith.science/api/pith-number/R77CAC2QRSBJBAJT4B6Y3DWHQI/events.json","paper":"https://pith.science/paper/R77CAC2Q"},"agent_actions":{"view_html":"https://pith.science/pith/R77CAC2QRSBJBAJT4B6Y3DWHQI","download_json":"https://pith.science/pith/R77CAC2QRSBJBAJT4B6Y3DWHQI.json","view_paper":"https://pith.science/paper/R77CAC2Q","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=1911.00650&json=true","fetch_graph":"https://pith.science/api/pith-number/R77CAC2QRSBJBAJT4B6Y3DWHQI/graph.json","fetch_events":"https://pith.science/api/pith-number/R77CAC2QRSBJBAJT4B6Y3DWHQI/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/R77CAC2QRSBJBAJT4B6Y3DWHQI/action/timestamp_anchor","attest_storage":"https://pith.science/pith/R77CAC2QRSBJBAJT4B6Y3DWHQI/action/storage_attestation","attest_author":"https://pith.science/pith/R77CAC2QRSBJBAJT4B6Y3DWHQI/action/author_attestation","sign_citation":"https://pith.science/pith/R77CAC2QRSBJBAJT4B6Y3DWHQI/action/citation_signature","submit_replication":"https://pith.science/pith/R77CAC2QRSBJBAJT4B6Y3DWHQI/action/replication_record"}},"created_at":"2026-07-05T01:01:21.343020+00:00","updated_at":"2026-07-05T01:01:21.343020+00:00"}