Spotting Out-of-Character Behavior: Atomic-Level Evaluation of Persona Fidelity in Open-Ended Generation

• URL: http://arxiv.org/abs/2506.19352v1
• Date: Tue, 24 Jun 2025 06:33:10 GMT
• Title: Spotting Out-of-Character Behavior: Atomic-Level Evaluation of Persona Fidelity in Open-Ended Generation
• Authors: Jisu Shin, Juhyun Oh, Eunsu Kim, Hoyun Song, Alice Oh,
• Abstract summary: We propose an atomic-level evaluation framework that quantifies persona fidelity at a finer granularity.<n>Our three key metrics measure the degree of persona alignment and consistency within and across generations.<n>By analyzing persona fidelity across diverse tasks and personality types, we reveal how task structure and persona desirability influence model adaptability.
• Score: 16.76995815742803
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Ensuring persona fidelity in large language models (LLMs) is essential for maintaining coherent and engaging human-AI interactions. However, LLMs often exhibit Out-of-Character (OOC) behavior, where generated responses deviate from an assigned persona, leading to inconsistencies that affect model reliability. Existing evaluation methods typically assign single scores to entire responses, struggling to capture subtle persona misalignment, particularly in long-form text generation. To address this limitation, we propose an atomic-level evaluation framework that quantifies persona fidelity at a finer granularity. Our three key metrics measure the degree of persona alignment and consistency within and across generations. Our approach enables a more precise and realistic assessment of persona fidelity by identifying subtle deviations that real users would encounter. Through our experiments, we demonstrate that our framework effectively detects persona inconsistencies that prior methods overlook. By analyzing persona fidelity across diverse tasks and personality types, we reveal how task structure and persona desirability influence model adaptability, highlighting challenges in maintaining consistent persona expression.

Related papers

• Persistent Instability in LLM's Personality Measurements: Effects of Scale, Reasoning, and Conversation History [7.58175460763641]
  Even 400B+ models exhibit substantial response variability.<n> Interventions expected to stabilize behavior, such as chain-of-thought reasoning, detailed personas instruction, inclusion of conversation history, can paradoxically increase variability.<n>For safety-critical applications requiring predictable behavior, these findings indicate that personality-based alignment strategies may be fundamentally inadequate.
  arXiv  Detail & Related papers  (2025-08-06T19:11:33Z)
• Traits Run Deep: Enhancing Personality Assessment via Psychology-Guided LLM Representations and Multimodal Apparent Behaviors [46.55948528317124]
  We propose a novel personality assessment framework called textittextbfTraits Run Deep.<n>It employs textittextbfpsychology-informed prompts to elicit high-level personality-relevant semantic representations.<n>It devises a textittextbfText-Centric Trait Fusion Network that anchors rich text semantics to align and integrate asynchronous signals from other modalities.
  arXiv  Detail & Related papers  (2025-07-30T04:12:14Z)
• PersonaFeedback: A Large-scale Human-annotated Benchmark For Personalization [25.45861816665351]
  We introduce PersonaFeedback, a new benchmark that directly evaluates LLMs' ability to provide personalized responses.<n>Unlike existing benchmarks that require models to infer implicit user personas from historical interactions, PersonaFeedback decouples persona inference from personalization.<n> PersonaFeedback consists of 8298 human-annotated test cases, which are categorized into easy, medium, and hard tiers.
  arXiv  Detail & Related papers  (2025-06-15T17:19:19Z)
• IMPersona: Evaluating Individual Level LM Impersonation [28.040025302581366]
  We introduce IMPersona, a framework for evaluating LMs at impersonating specific individuals' writing style and personal knowledge.<n>We demonstrate that even modestly sized open-source models, such as Llama-3.1-8B-Instruct, can achieve impersonation abilities at concerning levels.
  arXiv  Detail & Related papers  (2025-04-06T02:57:58Z)
• On the Fairness, Diversity and Reliability of Text-to-Image Generative Models [68.62012304574012]
  multimodal generative models have sparked critical discussions on their reliability, fairness and potential for misuse.<n>We propose an evaluation framework to assess model reliability by analyzing responses to global and local perturbations in the embedding space.<n>Our method lays the groundwork for detecting unreliable, bias-injected models and tracing the provenance of embedded biases.
  arXiv  Detail & Related papers  (2024-11-21T09:46:55Z)
• ConSiDERS-The-Human Evaluation Framework: Rethinking Human Evaluation for Generative Large Language Models [53.00812898384698]
  We argue that human evaluation of generative large language models (LLMs) should be a multidisciplinary undertaking. We highlight how cognitive biases can conflate fluent information and truthfulness, and how cognitive uncertainty affects the reliability of rating scores such as Likert. We propose the ConSiDERS-The-Human evaluation framework consisting of 6 pillars -- Consistency, Scoring Criteria, Differentiating, User Experience, Responsible, and Scalability.
  arXiv  Detail & Related papers  (2024-05-28T22:45:28Z)
• PsyCoT: Psychological Questionnaire as Powerful Chain-of-Thought for Personality Detection [50.66968526809069]
  We propose a novel personality detection method, called PsyCoT, which mimics the way individuals complete psychological questionnaires in a multi-turn dialogue manner. Our experiments demonstrate that PsyCoT significantly improves the performance and robustness of GPT-3.5 in personality detection.
  arXiv  Detail & Related papers  (2023-10-31T08:23:33Z)
• Automated Evaluation of Personalized Text Generation using Large Language Models [38.2211640679274]
  We present AuPEL, a novel evaluation method that distills three major semantic aspects of the generated text: personalization, quality and relevance, and automatically measures these aspects. We find that, compared to existing evaluation metrics, AuPEL not only distinguishes and ranks models based on their personalization abilities more accurately, but also presents commendable consistency and efficiency for this task.
  arXiv  Detail & Related papers  (2023-10-17T21:35:06Z)
• Position: AI Evaluation Should Learn from How We Test Humans [65.36614996495983]
  We argue that psychometrics, a theory originating in the 20th century for human assessment, could be a powerful solution to the challenges in today's AI evaluations.
  arXiv  Detail & Related papers  (2023-06-18T09:54:33Z)
• Improving Personality Consistency in Conversation by Persona Extending [22.124187337032946]
  We propose a novel retrieval-to-prediction paradigm consisting of two subcomponents, namely, Persona Retrieval Model (PRM) and Posterior-scored Transformer (PS-Transformer) Our proposed model yields considerable improvements in both automatic metrics and human evaluations.
  arXiv  Detail & Related papers  (2022-08-23T09:00:58Z)
• Improving Factual Consistency Between a Response and Persona Facts [64.30785349238619]
  Neural models for response generation produce responses that are semantically plausible but not necessarily factually consistent with facts describing the speaker's persona. We propose to fine-tune these models by reinforcement learning and an efficient reward function that explicitly captures the consistency between a response and persona facts as well as semantic plausibility.
  arXiv  Detail & Related papers  (2020-04-30T18:08:22Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.