A new research paper investigates linguistic biases in large language models (LLMs) when generating recommendations. The study used datasets from Yelp and Walmart, prompting LLMs with variations of American English, Indian English, and Code-Switched Hindi-English. Results indicated that certain models, like mistral-small-3.1 and the llama-3.1 family, showed increased sensitivity to Indian English and Code-Switched prompts for restaurant recommendations. For product recommendations, the llama-3.1-70B model was particularly affected by Code-Switched prompts, influencing categories like beauty and home. AI
Summary written by gemini-2.5-flash-lite from 2 sources. How we write summaries →
IMPACT Highlights potential biases in LLM recommendation systems, suggesting a need for careful prompt engineering and model evaluation across diverse linguistic inputs.
RANK_REASON Academic paper investigating linguistic biases in LLM recommendations.