# User-Assistant Bias in LLMs

arXiv: https://arxiv.org/abs/2508.15815 · HTML: https://mlsanity.com/paper/2508.15815/

## Key facts

- Cited by 3 papers in the map (data through 14 Oct 2026).
- Builds on 30 of its 47 references that are in the map.
- First posted on arXiv 16 Aug 2025.

## Builds on (most cited first, 30 of 30)

- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [ReAct: Synergizing Reasoning and Acting in Language Models](https://mlsanity.com/paper/2210.03629/) (2022)
- [Direct Preference Optimization: Your Language Model is Secretly a Reward Model](https://mlsanity.com/paper/2305.18290/) (2023)
- [Qwen2.5 Technical Report](https://mlsanity.com/paper/2412.15115/) (2024)
- [Finetuned Language Models Are Zero-Shot Learners](https://mlsanity.com/paper/2109.01652/) (2021)
- [Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback](https://mlsanity.com/paper/2204.05862/) (2022)
- [Lost in the Middle: How Language Models Use Long Contexts](https://mlsanity.com/paper/2307.03172/) (2023)
- [s1: Simple test-time scaling](https://mlsanity.com/paper/2501.19393/) (2025)
- [Towards Understanding Sycophancy in Language Models](https://mlsanity.com/paper/2310.13548/) (2023)
- [Large Language Models Cannot Self-Correct Reasoning Yet](https://mlsanity.com/paper/2310.01798/) (2023)
- [Discovering Language Model Behaviors with Model-Written Evaluations](https://mlsanity.com/paper/2212.09251/) (2022)
- [LIMO: Less is More for Reasoning](https://mlsanity.com/paper/2502.03387/) (2025)
- [The Instruction Hierarchy: Training LLMs to Prioritize Privileged Instructions](https://mlsanity.com/paper/2404.13208/) (2024)
- [LLMs Get Lost In Multi-Turn Conversation](https://mlsanity.com/paper/2505.06120/) (2025)
- [UltraFeedback: Boosting Language Models with Scaled AI Feedback](https://mlsanity.com/paper/2310.01377/) (2023)
- [Generating Sequences by Learning to Self-Correct](https://mlsanity.com/paper/2211.00053/) (2022)
- [REFINER: Reasoning Feedback on Intermediate Representations](https://mlsanity.com/paper/2304.01904/) (2023)
- [Simple synthetic data reduces sycophancy in large language models](https://mlsanity.com/paper/2308.03958/) (2023)
- [Platypus: Quick, Cheap, and Powerful Refinement of LLMs](https://mlsanity.com/paper/2308.07317/) (2023)
- [Does Prompt Formatting Have Any Impact on LLM Performance?](https://mlsanity.com/paper/2411.10541/) (2024)
- [SycEval: Evaluating LLM Sycophancy](https://mlsanity.com/paper/2502.08177/) (2025)
- [Shepherd: A Critic for Language Model Generation](https://mlsanity.com/paper/2308.04592/) (2023)
- [On the Emergence of Position Bias in Transformers](https://mlsanity.com/paper/2502.01951/) (2025)
- [IHEval: Evaluating Language Models on Following the Instruction Hierarchy](https://mlsanity.com/paper/2502.08745/) (2025)
- [ChatBug: A Common Vulnerability of Aligned LLMs Induced by Chat Templates](https://mlsanity.com/paper/2406.12935/) (2024)
- [Feedback Friction: LLMs Struggle to Fully Incorporate External Feedback](https://mlsanity.com/paper/2506.11930/) (2025)
- [Reasoning Isn't Enough: Examining Truth-Bias and Sycophancy in LLMs](https://mlsanity.com/paper/2506.21561/) (2025)
- [Emergence of Episodic Memory in Transformers: Characterizing Changes in Temporal Structure of Attention Scores During Training](https://mlsanity.com/paper/2502.06902/) (2025)
- [Repairs in a Block World: A New Benchmark for Handling User Corrections with Multi-Modal Language Models](https://mlsanity.com/paper/2409.14247/) (2024)
- [Investigating the Scaling Effect of Instruction Templates for Training Multimodal Language Model](https://mlsanity.com/paper/2412.08307/) (2024)

## Built on by (most cited first, 3 of 3)

- [Learning When to Trust via Selective Context Preference Optimization](https://mlsanity.com/paper/2608.06377/) (2026)
- [Who Does What? Archetypes of Roles Assigned to LLMs During Human-AI Decision-Making](https://mlsanity.com/paper/2602.11924/) (2026)
- [Caught in the Story: Narrative Captivity in Multi-turn LLMs Conversation](https://mlsanity.com/paper/2609.03407/) (2026)

## Abstract (arXiv)

Modern large language models (LLMs) are typically trained and deployed using structured role tags (e.g. system, user, assistant, tool) that explicitly mark the source of each piece of context. While these tags are essential for instruction following and controllability, asymmetries in the training data associated with different role tags can potentially introduce inductive biases. In this paper, we study this phenomenon by formalizing user-assistant bias, defined as the tendency of an LLM to preferentially rely on information from either the user or assistant role when they provide incompatible information about the same entity in the context history. We introduce a task-agnostic benchmark UserAssist and evaluate such bias in 52 frontier models. We observe that most of the instruction-tuned models exhibit strong user bias, whereas base and reasoning models are close to neutral. Using controlled fine-tuning experiments, we isolate which post-training recipes drive the observed user-assistant bias. We find that human-preference alignment amplifies user bias, while reasoning fine-tuning reduces it. Finally, we show that user-assistant bias can be bidirectionally controlled via direct preference optimization (DPO) on UserAssist-train, and that the resulting bias reliably generalizes to two realistic multi-turn debate datasets spanning philosophical opinions and natural argumentative exchanges on factual/policy topics. These results reveal an underexplored consequence of role-tagged training and provide a principled framework to diagnose and control tag-induced biases in modern LLMs.