Seeing Symbols, Missing Cultures: Probing Vision-Language Models' Reasoning on Fire Imagery and Cultural Meaning
On the map: LLM reasoning → Multimodal reasoning
Abstract (arXiv)
Vision-Language Models (VLMs) often appear culturally competent but rely on superficial pattern matching rather than genuine cultural understanding. We introduce a diagnostic framework to probe VLM reasoning on fire-themed cultural imagery through both classification and explanation analysis. Testing multiple models on Western festivals, non-Western traditions, and emergency scenes reveals systematic biases: models correctly identify prominent Western festivals but struggle with underrepresented cultural events, frequently offering vague labels or dangerously misclassifying emergencies as celebrations. These failures expose the risks of symbolic shortcuts and highlight the need for cultural evaluation beyond accuracy metrics to ensure interpretable and fair multimodal systems.
Key facts. Cited by 3 papers in the map (data through 14 Oct 2026). Placed in Multimodal reasoning, one of 17 directions in LLM reasoning. Builds on 5 of its 12 references that are in the map. First posted on arXiv 27 Sep 2025.
Builds on
5 of 12 references.
References · 5
- 2024Benchmarking Vision Language Models for Cultural Understandingcited by 101
- 2025CultureVLM: Characterizing and Improving Cultural Understanding of Vision-Language Models for over 100 Countriescited by 42
- 2025An Evaluation of Cultural Value Alignment in LLMcited by 28
- 2024See It from My Perspective: How Language Affects Cultural Bias in Image Understandingcited by 17
- 2025Multimodal Cultural Safety: Evaluation Framework and Alignment Strategiescited by 8
Built on by
Citing papers per year: 2026: 3.
3 papers.
- 2026VULCA-Bench: A Multicultural Vision-Language Benchmark for Evaluating Cultural Understandingcited by 3
- 2026Cross-Cultural Expert-Level Art Critique Evaluation with Vision-Language Modelscited by 1
- 2026Computational Humor with Multimodal LLMs: Methods, Datasets, Evaluation, and Challengescited by 0
- 2026Computational Humor with Multimodal LLMs: Methods, Datasets, Evaluation, and Challenges21 Jul 2026
- 2026VULCA-Bench: A Multicultural Vision-Language Benchmark for Evaluating Cultural Understanding12 Jan 2026
- 2026Cross-Cultural Expert-Level Art Critique Evaluation with Vision-Language Models12 Jan 2026
Similar, not cited
Close in wording, no citation either way
- 2025Evaluation of Cultural Competence of Vision-Language Models
- 2025Seeing Culture: A Benchmark for Visual Reasoning and Grounding
- 2025Vision Language Models are Confused Tourists
- 2026Cultural Moment Benchmark: Evaluating Video Cultural Reasoning and Grounding in Southeast Asia
- 2024From Local Concepts to Universals: Evaluating the Multicultural Understanding of Vision-Language Models
- 2026On the Cultural Anachronism and Temporal Reasoning in Vision Language Models
- 2025GIMMICK -- Globally Inclusive Multimodal Multitask Cultural Knowledge Benchmarking
- 2025Toward Socially Aware Vision-Language Models: Evaluating Cultural Competence Through Multimodal Story Generation