{"paper":{"title":"Beyond the Final Actor: Modeling the Dual Roles of Creator and Editor for Fine-Grained LLM-Generated Text Detection","license":"http://creativecommons.org/licenses/by/4.0/","headline":"RACE separates creator intent from editor style to enable four-class detection of LLM-generated text.","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Danding Wang, Juan Cao, Qiang Sheng, Yang Li, Yehan Yang, Zhengjia Wang","submitted_at":"2026-04-06T17:59:55Z","abstract_excerpt":"The misuse of large language models (LLMs) requires precise detection of synthetic text. Existing works mainly follow binary or ternary classification settings, which can only distinguish pure human/LLM text or collaborative text at best. This remains insufficient for the nuanced regulation, as the LLM-polished human text and humanized LLM text often trigger different policy consequences. In this paper, we explore fine-grained LLM-generated text detection under a rigorous four-class setting. To handle such complexities, we propose RACE (Rhetorical Analysis for Creator-Editor Modeling), a fine-"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"RACE outperforms 12 baselines in identifying fine-grained types with low false alarms, offering a policy-aligned solution for LLM regulation.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That Rhetorical Structure Theory graphs and Elementary Discourse Unit features reliably separate creator intent from editor style even when both roles are played by LLMs or when editing is subtle.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"RACE uses Rhetorical Structure Theory for creator logic graphs and Elementary Discourse Unit features for editor style to outperform baselines on four-class LLM text detection.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"RACE separates creator intent from editor style to enable four-class detection of LLM-generated text.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"1217cfe0c92c8742fbe3e493f677288d456d301f5f9fec63c205101d99f2ecd0"},"source":{"id":"2604.04932","kind":"arxiv","version":3},"verdict":{"id":"99fdee9b-4a81-42b3-8849-241e38a5108f","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-10T19:40:57.031848Z","strongest_claim":"RACE outperforms 12 baselines in identifying fine-grained types with low false alarms, offering a policy-aligned solution for LLM regulation.","one_line_summary":"RACE uses Rhetorical Structure Theory for creator logic graphs and Elementary Discourse Unit features for editor style to outperform baselines on four-class LLM text detection.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That Rhetorical Structure Theory graphs and Elementary Discourse Unit features reliably separate creator intent from editor style even when both roles are played by LLMs or when editing is subtle.","pith_extraction_headline":"RACE separates creator intent from editor style to enable four-class detection of LLM-generated text."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.04932/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":2,"snapshot_sha256":"ef9040cb5e8db973a3121c7baf342644e1a94cbeca71c09d4249607926b94e29"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}