Can LLMs Cook Jamaican Couscous? A Study of Cultural Novelty in Recipe Generation

TL;DR

This work addresses whether large language models (LLMs) can generate culturally meaningful recipe adaptations beyond dominant cultures. It extends the GlobalFusion framework to create LLMFusion, enabling direct comparison between eight LLMs and human adaptations across 130 countries, using five divergence-based metrics anchored in Jensen–Shannon divergence. The findings show that LLMs produce novelty that weakly tracks cultural distance, exhibit systematic over-divergence, and fail to ground culturally salient elements like ingredients, with cultural information only weakly encoded in internal layers. These results highlight fundamental limitations of current LLMs for culturally oriented generation and underscore the need for artifact-based evaluation and culture-aware model design for sensitive applications.

Abstract

Large Language Models (LLMs) are increasingly used to generate and shape cultural content, ranging from narrative writing to artistic production. While these models demonstrate impressive fluency and generative capacity, prior work has shown that they also exhibit systematic cultural biases, raising concerns about stereotyping, homogenization, and the erasure of culturally specific forms of expression. Understanding whether LLMs can meaningfully align with diverse cultures beyond the dominant ones remains a critical challenge. In this paper, we study cultural adaptation in LLMs through the lens of cooking recipes, a domain in which culture, tradition, and creativity are tightly intertwined. We build on the \textit{GlobalFusion} dataset, which pairs human recipes from different countries according to established measures of cultural distance. Using the same country pairs, we generate culturally adapted recipes with multiple LLMs, enabling a direct comparison between human and LLM behavior in cross-cultural content creation. Our analysis shows that LLMs fail to produce culturally representative adaptations. Unlike humans, the divergence of their generated recipes does not correlate with cultural distance. We further provide explanations for this gap. We show that cultural information is weakly preserved in internal model representations, that models inflate novelty in their production by misunderstanding notions such as creativity and tradition, and that they fail to identify adaptation with its associated countries and to ground it in culturally salient elements such as ingredients. These findings highlight fundamental limitations of current LLMs for culturally oriented generation and have important implications for their use in culturally sensitive applications.

Figures (12)

• Figure 1: Prompt example to generate multiple variations of a recipe given a specific NATIONALITY and a RECIPE-NAME.
• Figure 2: Example of a Jamaican Cultural Variations paired with Moroccan Couscous. The top left shows the ingredient set of Moroccan couscous reference recipes, while the top right displays ingredients characteristic of Jamaican cuisine, exemplified by Jerk Chicken. The bottom-right part presents human Jamaican couscous adaptations, and the bottom-left part presents an LLM-generated recipe using our protocol. While the human adaptation balances ingredients drawn from both Moroccan and Jamaican culinary traditions, the LLM-generated recipe predominantly relies on Jamaican ingredients, with minimal reuse of elements from the Moroccan reference.
• Figure 3: Correlations between the novelty metrics and the cultural distances for both human and LLMs. The color and its intensity provide information on the value of the coefficient, while the presence of the blue dot in each cell indicate a p value inferior to 0.05.
• Figure 4: Divergence increase rate between LLMs and human recipes. A positive value means that LLM-generated recipes have an increased divergence compared to human recipes.
• Figure 5: Average divergence gap between traditional recipes and novel and creative ones. The color and its intensity provide information on the value of the coefficient, while the presence of the blue dot in each cell indicate a p value inferior to 0.05.