MemeCULT-1K: Benchmarking South Asian Cultural Context and Humor Understanding of Multimodal Models

Tawsif Dipto, Author 2, Author 3, Author 4, Author 5

multimodalcultural contextbenchmarkingmeme understanding

2609.01772

Builder Relevance

80%

2h ago

Abstract

The paper introduces MemeCULT-1K, a benchmark for evaluating the cultural context and humor understanding of multimodal models using South Asian memes.

Reality Card

Core Claim

Providing minimal cultural context significantly improves the performance of Vision Language Models on meme understanding tasks.

Method / Result

Mean SBERT similarity improves from 44.6 to 56.4 (+11.8) across all models and languages.

Limitations

Closed-source models struggle with entity and reference misidentification, while open-source models face broader cultural knowledge gaps.

Paper to code

Verified implementation resources so builders can test the paper’s claims instead of stopping at the abstract.

No verified implementation link has been attached yet. AIBuzzHub will keep this panel separate from unverified search results.

← Back to all papers