Investigating Assistant Bias in LLM User Simulators Using a Role Vector

arXiv:2609.00608 · cs.CL, cs.HC · Submitted 2026-09-01 · Read on arXiv

cs.CL, cs.HC

Submitted: 2026-09-01

Updated: 2026-10-03

Code: https://github.com/SAP/mirrorbench

Terminology

• Steering — 133× in this paper · explained in 2 episodes
• Benchmark — 6× in this paper · explained in 2 episodes
• Cosine Similarity — 6× in this paper · explained in 3 episodes
• LLM judge — 5× in this paper · explained in 2 episodes
• Activation Steering — 4× in this paper · explained in 4 episodes
• Alignment — 4× in this paper · explained in 1 episode
• Consistency — 4× in this paper · explained in 3 episodes
• Activation Addition — 3× in this paper · explained in 1 episode
• Fidelity — 3× in this paper · explained in 2 episodes
• Frustration — 3× in this paper · explained in 2 episodes
• Transformer — 3× in this paper · explained in 2 episodes
• Accuracy — 2× in this paper · explained in 1 episode
• Commentary — 2× in this paper · explained in 2 episodes
• Hallucination — 2× in this paper · explained in 4 episodes
• Abstention — 1× in this paper · explained in 1 episode
• Accountability — 1× in this paper · explained in 2 episodes
• Adam — 1× in this paper · explained in 2 episodes
• Agentic Evaluation — 1× in this paper · explained in 1 episode
• Agentic Task — 1× in this paper · explained in 1 episode
• Agentic Tasks — 1× in this paper · explained in 1 episode
• Aggregation — 1× in this paper · explained in 1 episode
• Benchmarking — 1× in this paper · explained in 4 episodes
• Clustering — 1× in this paper · explained in 2 episodes
• Contrastive Learning — 1× in this paper · explained in 5 episodes
• Core — 1× in this paper · explained in 2 episodes
• Generalization — 1× in this paper · explained in 13 episodes
• Generative AI — 1× in this paper · explained in 2 episodes
• Robustness — 1× in this paper · explained in 5 episodes
• Sycophancy — 1× in this paper · explained in 4 episodes
• Synthetic Data Generation — 1× in this paper · explained in 5 episodes
• Trace — 1× in this paper · explained in 2 episodes
• alpha — 1× in this paper · explained in 1 episode

Sources

• Where is the Mind? Persona Vectors and LLM Individuation
• Persona Vectors: Monitoring and Controlling Character Traits in Language Models
• Beyond Cooperative Simulators: Generating Realistic User Personas for Robust Evaluation of LLM Agents
• User Simulation in the Era of Generative AI: User Modeling, Synthetic Data Generation, and System Evaluation
• To Steer or Not to Steer? Mechanistic Error Reduction with Abstention for Language Models
• DiscoverLLM: From Executing Intents to Discovering Them
• Lost in Simulation: LLM-Simulated Users are Unreliable Proxies for Human Users in Agentic Evaluations
• The Assistant Axis: Situating and Stabilizing the Default Persona of Language Models
• Scalable and Transferable Black-Box Jailbreaks for Language Models via Persona Modulation
• Emotion Concepts and their Function in a Large Language Model
• Valence-Arousal Subspace in LLMs: Circular Emotion Geometry and Multi-Behavioral Control
• Mind the Sim2Real Gap in User Simulation for Agentic Tasks
• Programming by Chat: A Large-Scale Behavioral Analysis of 11,579 Real-World AI-Assisted IDE Sessions
• RealUserSim: Bridging the Reality Gap in Agent Benchmarking via Grounded User Simulation
• DIAL: Direct Iterative Adversarial Learning for Realistic Multi-Turn Dialogue Simulation
• Steering Language Models With Activation Engineering
• HumanLM: Simulating Users with State Alignment Beats Response Imitation

Related papers

• Exploring Solution Divergence and Its Effect on Large Language Model Problem Solving
• Ishigaki-IDS-Bench: A Benchmark for Generating Information Delivery Specification from BIM Information Requirements
• Subliminal Steering: Stronger Encoding of Hidden Signals
• MedStruct-S: A Benchmark for Key Discovery, Key-Conditioned QA and Semi-Structured Extraction from OCR Clinical Reports
• The End of Transformers? On Challenging Attention and the Rise of Sub-Quadratic Architectures
• Untangling the Mechanisms of Misleading Context in Medical Question Answering