{"paper":{"title":"DecepGPT: Schema-Driven Deception Detection with Multicultural Datasets and Robust Multimodal Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"DecepGPT augments existing deception benchmarks with cue descriptions and reasoning chains, adds a large multicultural dataset, and introduces two modules to reach state-of-the-art detection that transfers across domains and cultures.","cross_cats":["cs.AI"],"primary_cat":"cs.CV","authors_text":"Chunmei Zhu, Dongliang Zhu, Hui Ma, Jiajian Huang, Jiayu Zhang, Xiaochun Cao, Zitong Yu","submitted_at":"2026-03-25T04:06:36Z","abstract_excerpt":"Multimodal deception detection aims to identify deceptive behavior by analyzing audiovisual cues for forensics and security. In these high-stakes settings, investigators need verifiable evidence connecting audiovisual cues to final decisions, along with reliable generalization across domains and cultural contexts. However, existing benchmarks provide only binary labels without intermediate reasoning cues. Datasets are also small with limited scenario coverage, leading to shortcut learning. We address these issues through three contributions. First, we construct reasoning datasets by augmenting"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Experiments on three established benchmarks and our novel dataset demonstrate that our method achieves state-of-the-art performance in both in-domain and cross-domain scenarios, while exhibiting superior transferability across diverse cultural contexts.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the manually added cue-level descriptions and reasoning chains in the augmented datasets accurately reflect genuine deception signals rather than annotator bias or post-hoc rationalization.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"A new 1695-sample multicultural dataset plus two modules for stable multimodal fusion and modality consistency yield state-of-the-art deception detection with cross-cultural transfer.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"DecepGPT augments existing deception benchmarks with cue descriptions and reasoning chains, adds a large multicultural dataset, and introduces two modules to reach state-of-the-art detection that transfers across domains and cultures.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"4ccd40db0597bf55f22b7e718bf892fa238c18819d89dd28456b25a4e3ddd58e"},"source":{"id":"2603.23916","kind":"arxiv","version":3},"verdict":{"id":"ad8bbeef-18ec-4ea9-ac94-5afa242a328f","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T01:05:36.984856Z","strongest_claim":"Experiments on three established benchmarks and our novel dataset demonstrate that our method achieves state-of-the-art performance in both in-domain and cross-domain scenarios, while exhibiting superior transferability across diverse cultural contexts.","one_line_summary":"A new 1695-sample multicultural dataset plus two modules for stable multimodal fusion and modality consistency yield state-of-the-art deception detection with cross-cultural transfer.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the manually added cue-level descriptions and reasoning chains in the augmented datasets accurately reflect genuine deception signals rather than annotator bias or post-hoc rationalization.","pith_extraction_headline":"DecepGPT augments existing deception benchmarks with cue descriptions and reasoning chains, adds a large multicultural dataset, and introduces two modules to reach state-of-the-art detection that transfers across domains and cultures."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2603.23916/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"}