Deep Persona: A Psychologically Grounded Architecture and Evaluation Framework for Role-Playing Agents and Simulations
Paper proposes Deep Persona, a three-layer psychologically grounded architecture for LLM role-playing agents plus a reference-free evaluation framework.
Deep Persona organizes personas into hierarchical levels of observable expression, latent beliefs, and core motivational drives, using scripted determinism and bounded agency to keep the model a reactive engine. The authors introduce a reference-free evaluation framework that benchmarks dialogue naturalness against empirical human distributions using clinical psychological instruments and adversarial stress-tests. Findings show LLMs achieve high pragmatic fluency but systematic weaknesses in emotional expression and joint attention; structured personas better align with human conversational behavior.
- Three-layer persona architecture: expression, latent beliefs, motivational drives
- Scripted determinism and bounded agency constrain the role-playing model
- Reference-free evaluation via clinical instruments and adversarial stress-tests
- LLMs fluent pragmatically but weak in emotional expression and joint attention
Full article157 words · extracted from huggingface.co · click to collapse
Existing approaches to persona simulation with Large Language Models (LLMs) mostly rely on shallow character descriptions that fail to sustain coherent character behavior across extended interactions. We introduce Deep Persona, a psychologically grounded, three-layered architecture that organizes personas into hierarchical levels of observable expression, latent beliefs, and core motivational drives, for constructing highly convincing role-playing agents. Governed by the principles of scripted determinism and bounded agency, the architecture restricts the model to a reactive engine guided by a structured internal script. We further propose a reference-free evaluation framework that benchmarks dialogue naturalness against empirical human distributions using established psychological clinical instruments and adversarial stress-tests. Empirical evaluation reveals that while LLMs achieve high pragmatic fluency, they exhibit systematic limitations in emotional expression and joint attention. In addition, we present a case study of two Deep Personas and evaluate them using the proposed framework, demonstrating that structured personas can produce interactions that more closely align with human conversational behavior.
Text extracted automatically; images, tables and formatting may be missing. Original: https://huggingface.co/papers/2609.22255