{"paper":{"title":"Luminol-AIDetect: Fast Zero-shot Machine-Generated Text Detection based on Perplexity under Text Shuffling","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Machine-generated text exhibits a distinct dispersion in perplexity after random shuffling, unlike the stable variability of human text.","cross_cats":["cs.AI","cs.CY"],"primary_cat":"cs.CL","authors_text":"Andrea Tagarelli, Lucio La Cava","submitted_at":"2026-04-28T16:58:55Z","abstract_excerpt":"Machine-generated text (MGT) detection requires identifying structurally invariant signals across generation models, rather than relying on model-specific fingerprints. In this respect, we hypothesize that while large language models excel at local semantic consistency, their autoregressive nature results in a specific kind of structural fragility compared to human writing. We propose Luminol-AIDetect, a novel, zero-shot statistical approach that exposes this fragility through coherence disruption. By applying a simple randomized text-shuffling procedure, we demonstrate that the resulting shif"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"By applying a simple randomized text-shuffling procedure, we demonstrate that the resulting shift in perplexity serves as a principled, model-agnostic discriminant, as MGT displays a characteristic dispersion in perplexity-under-shuffling that differs markedly from the more stable structural variability of human-written text.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the autoregressive nature of large language models produces a specific structural fragility that is reliably exposed by randomized text shuffling and remains distinguishable from human text across all content domains, languages, and adversarial modifications.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Luminol-AIDetect detects machine-generated text zero-shot by extracting perplexity-based features from original and shuffled text versions, using density estimation and ensemble prediction to exploit greater structural fragility in AI output.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Machine-generated text exhibits a distinct dispersion in perplexity after random shuffling, unlike the stable variability of human text.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"ba02b989d1598893e5b14bd6911fc49f913721dd0f173260d0062fb495805b35"},"source":{"id":"2604.25860","kind":"arxiv","version":2},"verdict":{"id":"c0ac9851-017e-4a24-b825-b3ed6a66ed14","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-07T16:00:13.870041Z","strongest_claim":"By applying a simple randomized text-shuffling procedure, we demonstrate that the resulting shift in perplexity serves as a principled, model-agnostic discriminant, as MGT displays a characteristic dispersion in perplexity-under-shuffling that differs markedly from the more stable structural variability of human-written text.","one_line_summary":"Luminol-AIDetect detects machine-generated text zero-shot by extracting perplexity-based features from original and shuffled text versions, using density estimation and ensemble prediction to exploit greater structural fragility in AI output.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the autoregressive nature of large language models produces a specific structural fragility that is reliably exposed by randomized text shuffling and remains distinguishable from human text across all content domains, languages, and adversarial modifications.","pith_extraction_headline":"Machine-generated text exhibits a distinct dispersion in perplexity after random shuffling, unlike the stable variability of human text."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.25860/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"ai_meta_artifact","ran_at":"2026-05-21T03:39:30.220230Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T20:43:08.771850Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"13c1b544a80cb3ca6b4b83f0bedbe6c1746f7de58543f1b486b0dad60d6c9001"},"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"}