PersonaLLM: Workshop on LLM Persona Modeling

Workshop @ NeurIPS 2025

Hilton Mexico City Reforma, Mexico City | Monday, Dec 1, 2025

Large language models (LLMs) exhibit strong instruction-following capabilities, which have led to their widespread use in building AI agents and personalized chatbots that engage in natural, context-aware interactions. Crucially, the adaptability of LLMs across diverse contexts opens new opportunities for personification, allowing the creation of distinct human-like personas with consistent traits and behaviors. Such advances provide an important foundation for exploring human cognition, emotion, and social interaction through LLM persona modeling.

LLM persona modeling leverages LLMs to simulate human characters, or personas, that can represent target audiences or specific roles in experimental and applied settings. By integrating demographic, social, and behavioral cues into prompts, LLMs can generate detailed and contextually grounded profiles. These personas have broad applications, ranging from marketing research and social science to product development and healthcare. However, key challenges remain, particularly regarding accuracy, bias, and ethical implications.

The PersonaLLM workshop aims to provide a timely, interdisciplinary forum on LLM persona modeling, bringing together perspectives from AI, psychology, cognitive science, and human-computer interaction. The workshop will foster dialogue across disciplines, highlight emerging methods and applications, and chart future directions for developing robust, ethical, and scientifically grounded approaches to persona modeling with LLMs. Our mission aligns with NeurIPS in advancing responsible, interpretable, and socially grounded machine learning research as LLMs become embedded in real-world systems.

Workshop Goals

The workshop aims to bridge NLP, psychology, and social sciences in understanding and developing LLM personas. We focus on:

Conceptualizing LLM personas through computational and psychological lenses
Building robust evaluation methods for persona consistency and effectiveness
Exploring broader social and ethical implications of anthropomorphism in LLMs

Call for Papers

We invite contributions from NLP, psychology, cognitive science, HCI, sociology, and related fields. The workshop encompasses both technical innovations and interdisciplinary insights across the following key research areas:

Discovering and Steering LLM Personas: Systematically identifying, extracting, and responsibly steering latent personas embedded in LLMs through probing techniques, controllable generation, and prompt/representation engineering.
Evaluating Persona Authenticity and Consistency: Developing metrics, benchmarks, and methodologies to assess persona modeling consistency, psychological authenticity, and behavioral effectiveness against real human data.
Psychological and Cognitive Foundations: Incorporating established theories from psychology and cognitive science (personality theories, cognitive biases, social identity, cultural psychology) into computational persona modeling.
Responsible Development and Ethical Considerations: Balancing user engagement with privacy, consent, and cultural sensitivities; mitigating stereotyping and bias; ensuring informed consent and ethical deployment guidelines.
Technical Innovations and Methodologies: Novel architectures, training paradigms, multi-modal representations, few-shot learning, persona transfer, memory mechanisms, and efficient fine-tuning approaches.
Real-World Applications and Case Studies: Therapeutic and mental health settings, legal contexts, education, market research, creative industries, and accessibility technologies.

Submissions may report new results, negative findings, benchmarks, or visionary perspectives.

Submission Guidelines

We welcome submissions in the following formats:

Long papers: up to 9 pages (excluding references) for original research, comprehensive studies, or position papers.
Short papers: up to 4 pages (excluding references) for preliminary results, focused contributions, or negative findings.
Demo papers: 2–4 pages (excluding references) presenting systems, tools, or datasets. Must include a link to reproducible resources (code, data, or demo).
Extended abstracts: up to 2 pages (excluding references) for early-stage ideas or work-in-progress. These are non-archival and suitable for informal publication.
Literature review (special track): 2–4 pages summarizing and discussing relevant research published or accepted within the past 12 months. This track allows authors to synthesize and reflect on recent developments in persona modeling. Must include full citations and publication venues. Authors can present their work as posters or talks.

Formatting requirements:

Use the official NeurIPS 2025 style files
Submit a single PDF
For long papers, short papers, demo papers, extended abstracts, and literature review: submissions should be anonymized following NeurIPS anonymization guidelines
Appendix: include any supplementary material in the same PDF — no page limit for the appendix

Important Dates

Submission opens: Oct 17, 2025
Submission deadline: Nov 4, 2025 (AoE)
Notification of decisions: Nov 7, 2025
Paper awards notification: Nov 17, 2025
Workshop: Monday, Dec 1, 2025 (at NeurIPS 2025 in Mexico City)
Submission link: OpenReview

The workshop is non‑archival; authors are free to publish revised versions elsewhere. Every submission will receive at least two reviews from our program committee, and accepted papers will be presented as oral talks or posters.

Questions? Contact us at personanlp.workshop@gmail.com