# CulTrace: Tracing Internal Cultural Reasoning in Large Language Models

arXiv: https://arxiv.org/abs/2508.08879 · HTML: https://mlsanity.com/paper/2508.08879/

## Key facts

- Cited by 6 papers in the map (data through 14 Oct 2026).
- Placed in Reasoning benchmarks and evaluation, one of 17 directions in LLM reasoning.
- Builds on 3 of its 31 references that are in the map.
- First posted on arXiv 12 Aug 2025.

## Placed in

- LLM reasoning / [Reasoning benchmarks and evaluation](https://mlsanity.com/lr/reasoning-benchmarks-and-evaluation/)
- Mechanistic interpretability / [Knowledge and representations inside LLMs](https://mlsanity.com/mechinterp/knowledge-and-representations-inside-llms/)

## Builds on (most cited first, 3 of 3)

- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [Gemma 3 Technical Report](https://mlsanity.com/paper/2503.19786/) (2025)
- [Eliciting Latent Predictions from Transformers with the Tuned Lens](https://mlsanity.com/paper/2303.08112/) (2023)

## Built on by (most cited first, 6 of 6)

- [Reflections and New Directions for Human-Centered Large Language Models](https://mlsanity.com/paper/2605.06901/) (2026)
- [Understanding the Interplay between LLMs' Utilisation of Parametric and Contextual Knowledge: A keynote at ECIR 2025](https://mlsanity.com/paper/2603.09654/) (2026)
- [Expressing Social Emotions: Misalignment Between LLMs and Human Cultural Emotion Norms](https://mlsanity.com/paper/2604.16757/) (2026)
- [Will Humanity Be Rendered Obsolete by AI?](https://mlsanity.com/paper/2510.22814/) (2025)
- [Can LLMs Cook Jamaican Couscous? A Study of Cultural Novelty in Recipe Generation](https://mlsanity.com/paper/2602.10964/) (2026)
- [When a Name Is Not a Name: A Benchmark Dataset and Distilled Reasoning for Culturally Entangled Bangla Homographs in Low-Resource LLMs](https://mlsanity.com/paper/2607.17828/) (2026)

## Abstract (arXiv)

The growing deployment of large language models (LLMs) across diverse cultural contexts necessitates a deeper understanding of models' hidden representations of different cultures. Prior work has evaluated cultural awareness in LLMs by analysing their outputs. This approach overlooks how cultures are represented within the model parameters, missing why models generate incorrect responses. To bridge this gap, we propose CulTrace, a mechanistic interpretability-based method that probes the internal representations of LLMs for cultural knowledge. With CulTrace, we inspect how cultural knowledge is processed across layers and how it is integrated during cultural QA. We find a consistent staged trajectory of cultural reasoning. Models first engage with the question's domain, then resolve the relevant culture, and finally narrow in on an answer. We also demonstrate that models' cultural reasoning is imbalanced, showing delayed relevant culture resolution and more confusion with less-represented cultures.