# LLM Biases

arXiv: https://arxiv.org/abs/2604.26960 · HTML: https://mlsanity.com/paper/2604.26960/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Agentic recommender systems, one of 16 directions in LLM agents and tool use.
- Builds on 18 of its 66 references that are in the map.
- First posted on arXiv 7 Apr 2026.

## Placed in

- LLM agents and tool use / [Agentic recommender systems](https://mlsanity.com/agents/agentic-recommender-systems/)

## Builds on (most cited first, 18 of 18)

- [Efficient Streaming Language Models with Attention Sinks](https://mlsanity.com/paper/2309.17453/) (2023)
- [How Bad is Training on Synthetic Data? A Statistical Analysis of Language Model Collapse](https://mlsanity.com/paper/2404.05090/) (2024)
- [OptiMUS-0.3: Using Large Language Models to Model and Solve Optimization Problems at Scale](https://mlsanity.com/paper/2407.19633/) (2024)
- [PLUM: Adapting Pre-trained Language Models for Industrial-scale Generative Recommendations](https://mlsanity.com/paper/2510.07784/) (2025)
- [Large Language Models for Market Research: A Data-augmentation Approach](https://mlsanity.com/paper/2412.19363/) (2024)
- [Large Language Models for Supply Chain Decisions](https://mlsanity.com/paper/2507.21502/) (2025)
- [AI Agents for Inventory Control: Human-LLM-OR Complementarity](https://mlsanity.com/paper/2602.12631/) (2026)
- [Transformer Choice Net: A Transformer Neural Network for Choice Prediction](https://mlsanity.com/paper/2310.08716/) (2023)
- [Understanding the Training and Generalization of Pretrained Transformer for Sequential Decision Making](https://mlsanity.com/paper/2405.14219/) (2024)
- [Artificial Intelligence Clones](https://mlsanity.com/paper/2501.16996/) (2025)
- [Large Language Newsvendor: Decision Biases and Cognitive Mechanisms](https://mlsanity.com/paper/2512.12552/) (2025)
- [Assured autonomy: How operations research powers and orchestrates generative AI systems](https://mlsanity.com/paper/2512.23978/) (2025)
- [Ask, Clarify, Optimize: Human-LLM Agent Collaboration for Smarter Inventory Control](https://mlsanity.com/paper/2601.00121/) (2025)
- [Rethinking Supply Chain Planning: A Generative Paradigm](https://mlsanity.com/paper/2509.03811/) (2025)
- [RideAgent: An LLM-Enhanced Optimization Framework for Automated Taxi Fleet Operations](https://mlsanity.com/paper/2505.06608/) (2025)
- [Supracompetitive Pricing Under AI Monoculture](https://mlsanity.com/paper/2601.01279/) (2026)
- [LLM-SAA: LLM-persona Generated Distributions for Decision-making](https://mlsanity.com/paper/2602.06357/) (2026)
- [LLM-Inspired Pretrain-Then-Finetune for Small-Data, Large-Scale Optimization](https://mlsanity.com/paper/2602.03690/) (2026)

## Abstract (arXiv)

Transformer-based agentic AI is rapidly being deployed on major platforms to help users shop, watch, and navigate content with less effort. While these systems can deliver impressive performance, a key concern is whether they may be less reliable than they appear. We ask a simple but fundamental question: whether the mechanisms that make transformer-based agents effective can also induce systematic biases or distortions? We study this question through a theoretical analysis of transformer-based generative recommenders, in which the next user interaction is generated sequentially from the user history. Focusing on how the model allocates attention across historical evidence, we identify four bias channels: (i) Positional bias: stronger positional encoding shifts influence toward recent history, improving responsiveness but potentially reducing stability and long-term diversity; (ii) Popularity amplification: small frequency differences in data can be magnified into disproportionate exposure, contributing to Matthew effects and echo chambers; (iii) Latent driver bias: when important drivers of user choices are not directly observed, the model can place overly concentrated weight on a small subset of past events, creating overconfident attributions. (iv) Synthetic data bias: when users increasingly follow AI suggestions and platforms retrain on model-shaped synthetic logs, outputs can concentrate over time, and long-tail alternatives can disappear first. Our analysis highlights mechanism-level reliability risks that may not be visible in offline performance metrics. The four bias channels indicate that large-scale deployment may systematically distort exposure and choice. For managers, the immediate implication is to treat these as operational risk factors and to monitor concentration and drift over time, rather than assuming that performance gains alone guarantee reliability.