Investigating Assistant Bias in LLM User Simulators Using a Role Vector

• URL: http://arxiv.org/abs/2609.00608v1
• Date: Tue, 01 Sep 2026 02:49:30 GMT
• Title: Investigating Assistant Bias in LLM User Simulators Using a Role Vector
• Abstract summary: LLM-based user simulators are increasingly used to evaluate autonomous agents at scale.<n>Assistant bias is baked in during model training, which role-playing prompts fail to override.<n>We observe two findings: (i) the user direction is identifiable in activations, elicits user-like behaviors, and captures characteristics distinct from assistant traits.
• Score: 21.70379807235766
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: LLM-based user simulators are increasingly used to evaluate autonomous agents at scale, in place of costly human evaluations. Despite this promise, these simulators exhibit "assistant bias," a tendency to cooperate and pursue task goals. They rarely reproduce the frustration or disengagement that real users exhibit, compromising evaluation validity. Prior work outlines that this bias is baked in during model training, which role-playing prompts fail to override. We analyze this bias from model activations, extracting a user role vector by contrasting how the model represents user versus assistant perspectives on the same dialogue. We observe two findings: (i) the user direction is identifiable in activations, elicits user-like behaviors, and captures characteristics distinct from assistant traits; and (ii) although user-role activation associates with simulation realism and steering strengthens it, it can exaggerate user behaviors and override individual user profiles. Together, our findings provide a representation-level analysis of LLM user simulators, confirming that assistant bias is structurally identifiable and that user behavior can be directionally analyzed.

Related papers

• Verifiable User Simulation for Search and Recommendation Systems [9.794615910946002]
  Large-language-model (LLM) based user simulation is increasingly adopted for evaluating search engines, recommender systems, and retrieval-augmented generation pipelines.<n>This tutorial introduces a proposed design-and-audit framework that treats a user simulator as a verifiable engineering artefact composed of seven auditable components.
  arXiv  Detail & Related papers  (2026-06-12T14:09:56Z)
• Preference-Aware Rubric Learning for Personalized Evaluation [59.539429430690156]
  Existing evaluation methods fail to capture user-specific preferences embedded in long-term interaction histories.<n>We propose Personalized Evaluation as Learning, a paradigm that formulates personalized evaluation as a learning problem rather than a static judgment.<n>Experiments show that PARL consistently induces high-fidelity rubrics that reliably identify user-aligned responses and generalize across users.
  arXiv  Detail & Related papers  (2026-05-29T17:00:55Z)
• The Illusion of Intervention: Your LLM-Simulated Experiment is an Observational Study [56.649987197786096]
  Large language models (LLMs) show potential as simulators of human behavior.<n>LLMs can induce unintended shifts in latent user attributes, causing user drift.<n>We formalize the confounding or selection bias that can arise due to user drift.
  arXiv  Detail & Related papers  (2026-05-20T06:09:41Z)
• Measuring and Mitigating the Distributional Gap Between Real and Simulated User Behaviors [61.610957638373826]
  We introduce a method to measure the distributional gap between real and simulated user behaviors.<n>Given a dataset of real and simulated conversations, our method extracts representations of user behavior from each conversation.<n>We show that most simulators behave similarly, while a few stand apart.
  arXiv  Detail & Related papers  (2026-05-08T15:09:25Z)
• Taming Actor-Observer Asymmetry in Agents via Dialectical Alignment [59.536125286960186]
  Multi-agent frameworks assigning specialized roles are increasingly adopted to enable self-reflection and mutual auditing.<n>We find it simultaneously induces a human-like cognitive bias known as Actor-Observer Asymmetry.<n>ReTAS guides agents to synthesize conflicting viewpoints into an objective consensus.
  arXiv  Detail & Related papers  (2026-04-21T15:05:58Z)
• Mind the Sim2Real Gap in User Simulation for Agentic Tasks [101.69142591891234]
  We formalize the Sim2Real gap in user simulation and present the first study running the full $$-bench protocol with real humans.<n> Behaviorally, LLM simulators are excessively cooperative, stylistically uniform, and lack realistic frustration or ambiguity.<n>These findings highlight the importance of human validation when using LLM-based user simulators in the agent development cycle.
  arXiv  Detail & Related papers  (2026-03-11T19:12:31Z)
• Flipping the Dialogue: Training and Evaluating User Language Models [31.119620506835677]
  We introduce purpose-built User Language Models (User LMs)<n>User LMs are models post-trained to simulate human users in multi-turn conversations.<n>We show how User LMs align better with human behavior and achieve better simulation robustness than existing simulation methods.
  arXiv  Detail & Related papers  (2025-10-08T01:04:36Z)
• Addressing Personalized Bias for Unbiased Learning to Rank [56.663619153713434]
  Unbiased learning to rank (ULTR) aims to learn unbiased ranking models from biased user behavior logs.<n>We propose a novel user-aware inverse-propensity-score estimator for learning-to-rank objectives.
  arXiv  Detail & Related papers  (2025-08-28T14:01:31Z)
• Personas within Parameters: Fine-Tuning Small Language Models with Low-Rank Adapters to Mimic User Behaviors [1.8352113484137629]
  A long-standing challenge in developing accurate recommendation models is simulating user behavior, mainly due to the complex nature of user interactions.<n>We propose an approach to extracting robust user representations using a frozen Large Language Models (LLMs) and simulating cost-effective, resource-efficient user agents powered by fine-tuned Small Language Models (SLMs)<n>Our experiments provide compelling empirical evidence of the efficacy of our methods, demonstrating that user agents developed using our approach have the potential to bridge the gap between offline metrics and real-world performance of recommender systems.
  arXiv  Detail & Related papers  (2025-08-18T22:14:57Z)
• How Reliable is Your Simulator? Analysis on the Limitations of Current LLM-based User Simulators for Conversational Recommendation [14.646529557978512]
  We analyze the limitations of using Large Language Models in constructing user simulators for Conversational Recommender System. Data leakage, which occurs in conversational history and the user simulator's replies, results in inflated evaluation results. We propose SimpleUserSim, employing a straightforward strategy to guide the topic toward the target items.
  arXiv  Detail & Related papers  (2024-03-25T04:21:06Z)
• User Behavior Simulation with Large Language Model based Agents [116.74368915420065]
  We propose an LLM-based agent framework and design a sandbox environment to simulate real user behaviors. Based on extensive experiments, we find that the simulated behaviors of our method are very close to the ones of real humans.
  arXiv  Detail & Related papers  (2023-06-05T02:58:35Z)

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.