{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:SPFMQKRGCEW3LWF575MW4WKAHV","short_pith_number":"pith:SPFMQKRG","schema_version":"1.0","canonical_sha256":"93cac82a26112db5d8bdff596e59403d561f5837f8f348447548e3c0ad54fd1d","source":{"kind":"arxiv","id":"2305.15047","version":3},"attestation_state":"computed","paper":{"title":"Ghostbuster: Detecting Text Ghostwritten by Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Dan Klein, Eve Fleisig, Nicholas Tomlin, Vivek Verma","submitted_at":"2023-05-24T11:37:10Z","abstract_excerpt":"We introduce Ghostbuster, a state-of-the-art system for detecting AI-generated text. Our method works by passing documents through a series of weaker language models, running a structured search over possible combinations of their features, and then training a classifier on the selected features to predict whether documents are AI-generated. Crucially, Ghostbuster does not require access to token probabilities from the target model, making it useful for detecting text generated by black-box models or unknown model versions. In conjunction with our model, we release three new datasets of human-"},"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":"2305.15047","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2023-05-24T11:37:10Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"486a3a71c80bd20d8727d395975b4f38a2c0ae98ec9c3cbfc867fd34b9308d46","abstract_canon_sha256":"2b306c6ace4d0f7848ba4f6af1b8df47a31cdc3ed5e79396da8d749b4c68ff80"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:04:58.595222Z","signature_b64":"AesabYsB3Vck7fGzTed8L9HbXqScMi0c5+v00l7ub+1xKmAy8pKqkmMQ1/qc6i3zSU94cVc/XG/2/ob24gK6Dw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"93cac82a26112db5d8bdff596e59403d561f5837f8f348447548e3c0ad54fd1d","last_reissued_at":"2026-07-05T08:04:58.594746Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:04:58.594746Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Ghostbuster: Detecting Text Ghostwritten by Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Dan Klein, Eve Fleisig, Nicholas Tomlin, Vivek Verma","submitted_at":"2023-05-24T11:37:10Z","abstract_excerpt":"We introduce Ghostbuster, a state-of-the-art system for detecting AI-generated text. Our method works by passing documents through a series of weaker language models, running a structured search over possible combinations of their features, and then training a classifier on the selected features to predict whether documents are AI-generated. Crucially, Ghostbuster does not require access to token probabilities from the target model, making it useful for detecting text generated by black-box models or unknown model versions. In conjunction with our model, we release three new datasets of human-"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2305.15047","kind":"arxiv","version":3},"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/2305.15047/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":"2305.15047","created_at":"2026-07-05T08:04:58.594804+00:00"},{"alias_kind":"arxiv_version","alias_value":"2305.15047v3","created_at":"2026-07-05T08:04:58.594804+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2305.15047","created_at":"2026-07-05T08:04:58.594804+00:00"},{"alias_kind":"pith_short_12","alias_value":"SPFMQKRGCEW3","created_at":"2026-07-05T08:04:58.594804+00:00"},{"alias_kind":"pith_short_16","alias_value":"SPFMQKRGCEW3LWF5","created_at":"2026-07-05T08:04:58.594804+00:00"},{"alias_kind":"pith_short_8","alias_value":"SPFMQKRG","created_at":"2026-07-05T08:04:58.594804+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.20761","citing_title":"Findings of the Counter Turing Test: AI-Generated Text Detection","ref_index":16,"is_internal_anchor":false},{"citing_arxiv_id":"2410.23728","citing_title":"GigaCheck: Detecting LLM-generated Content via Object-Centric Span Localization","ref_index":77,"is_internal_anchor":false},{"citing_arxiv_id":"2502.11336","citing_title":"ExaGPT: Example-Based Machine-Generated Text Detection for Human Interpretability","ref_index":38,"is_internal_anchor":false},{"citing_arxiv_id":"2605.20761","citing_title":"Findings of the Counter Turing Test: AI-Generated Text Detection","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2509.20924","citing_title":"RLCracker: Evaluating the Worst-Case Vulnerability of LLM Watermarks with Adaptive RL Attacks","ref_index":34,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/SPFMQKRGCEW3LWF575MW4WKAHV","json":"https://pith.science/pith/SPFMQKRGCEW3LWF575MW4WKAHV.json","graph_json":"https://pith.science/api/pith-number/SPFMQKRGCEW3LWF575MW4WKAHV/graph.json","events_json":"https://pith.science/api/pith-number/SPFMQKRGCEW3LWF575MW4WKAHV/events.json","paper":"https://pith.science/paper/SPFMQKRG"},"agent_actions":{"view_html":"https://pith.science/pith/SPFMQKRGCEW3LWF575MW4WKAHV","download_json":"https://pith.science/pith/SPFMQKRGCEW3LWF575MW4WKAHV.json","view_paper":"https://pith.science/paper/SPFMQKRG","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2305.15047&json=true","fetch_graph":"https://pith.science/api/pith-number/SPFMQKRGCEW3LWF575MW4WKAHV/graph.json","fetch_events":"https://pith.science/api/pith-number/SPFMQKRGCEW3LWF575MW4WKAHV/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/SPFMQKRGCEW3LWF575MW4WKAHV/action/timestamp_anchor","attest_storage":"https://pith.science/pith/SPFMQKRGCEW3LWF575MW4WKAHV/action/storage_attestation","attest_author":"https://pith.science/pith/SPFMQKRGCEW3LWF575MW4WKAHV/action/author_attestation","sign_citation":"https://pith.science/pith/SPFMQKRGCEW3LWF575MW4WKAHV/action/citation_signature","submit_replication":"https://pith.science/pith/SPFMQKRGCEW3LWF575MW4WKAHV/action/replication_record"}},"created_at":"2026-07-05T08:04:58.594804+00:00","updated_at":"2026-07-05T08:04:58.594804+00:00"}