{"paper":{"title":"PICon: A Multi-Turn Interrogation Framework for Evaluating Persona Agent Consistency","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Persona agents exhibit more contradictions than humans when tested with chained multi-turn questions.","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Chaeeun Shim, Edward Choi, Hwajung Hong, Junhee Lee, Junseong Choi, Minseo Kim, Sujeong Im","submitted_at":"2026-03-26T16:34:34Z","abstract_excerpt":"Large language model (LLM)-based persona agents are rapidly being adopted as scalable proxies for human participants across diverse domains. Yet there is no systematic method for verifying whether a persona agent's responses remain free of contradictions and factual inaccuracies throughout an interaction. A principle from interrogation methodology offers a lens: no matter how elaborate a fabricated identity, systematic interrogation will expose its contradictions. We apply this principle to propose PICon, an evaluation framework that probes persona agents through logically chained multi-turn q"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"even systems previously reported as highly consistent fail to meet the human baseline across all three dimensions, revealing contradictions and evasive responses under chained questioning.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the specific multi-turn interrogation protocol fairly measures consistency without LLM-specific biases such as prompt sensitivity or tendencies to evade that differ from human response patterns.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"PICon demonstrates that even highly regarded LLM persona agents exhibit more contradictions and evasive responses than real humans across internal, external, and retest consistency dimensions under chained interrogation.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Persona agents exhibit more contradictions than humans when tested with chained multi-turn questions.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"f904b2317e128226d1f87b3df88c5c19ac1ec1461ce8c33c40666f9240d91702"},"source":{"id":"2603.25620","kind":"arxiv","version":4},"verdict":{"id":"ef3b4730-8d71-46b0-96bb-638144b3260d","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T00:26:26.350873Z","strongest_claim":"even systems previously reported as highly consistent fail to meet the human baseline across all three dimensions, revealing contradictions and evasive responses under chained questioning.","one_line_summary":"PICon demonstrates that even highly regarded LLM persona agents exhibit more contradictions and evasive responses than real humans across internal, external, and retest consistency dimensions under chained interrogation.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the specific multi-turn interrogation protocol fairly measures consistency without LLM-specific biases such as prompt sensitivity or tendencies to evade that differ from human response patterns.","pith_extraction_headline":"Persona agents exhibit more contradictions than humans when tested with chained multi-turn questions."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2603.25620/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"}