# CrossCult-KIBench: A Benchmark for Cross-Cultural Knowledge Insertion in MLLMs

Source: [CrossCult-KIBench: A Benchmark for Cross-Cultural Knowledge Insertion in MLLMs](<https://www.sota2.com/research/paper/69fd6e165497b3b670c89424>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-05-12T05:00:04.826Z

Authors: Zhen Zeng, Leijiang Gu, Feng Li, Jing Yu, Zenglin Shi

Published: 2026-05-07T12:26:11.000Z

## Abstract

Multimodal Large Language Models (MLLMs), trained primarily on English-centric data, frequently generate culturally inappropriate or misaligned responses in cross-cultural settings. To mitigate this, we introduce the task of cross-cultural knowledge insertion, which focuses on adapting models to specific cultural contexts while preserving their original behavior in other cultures. To facilitate research in this area, we introduce CrossCult-KIBench, a comprehensive evaluation benchmark for assessing both the effectiveness of knowledge insertion and its unintended side effects on non-target cultures. The benchmark includes 9,800 image-grounded cases covering 49 culturally relevant visual scenarios across English, Chinese, and Arabic language-culture groups. It supports evaluation in both single-insert and sequential-insert settings. We also propose Memory-Conditioned Knowledge Insertion (MCKI) as a baseline method. MCKI retrieves relevant cultural knowledge from an external memory using frozen MLLM representations, prepending matched entries as conditional prompts when applicable. Extensive experiments on CrossCult-KIBench reveal that current approaches struggle to balance effective cultural adaptation with behavioral preservation, highlighting a key challenge in developing culturally-aware MLLMs. Our work thus underscores an important research direction for developing more culturally adaptive and responsible MLLMs.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Sequential-insert knowledge insertion on CrossCult-KIBench 1.0 (test)](<https://www.sota2.com/research/sota/sequential-insert-knowledge-insertion-on-crosscult-kibench-1-0-test.md>): Sequential-insert knowledge insertion; CrossCult-KIBench 1.0 (test); Final Relational ROUGE-L: 85.57.

- [Knowledge Insertion on CrossCult-KIBench single-insert](<https://www.sota2.com/research/sota/knowledge-insertion-on-crosscult-kibench-single-insert.md>): Knowledge Insertion; CrossCult-KIBench single-insert; Reliability (ROUGE-L): 79.83.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2605.06115>)

[PDF](<https://arxiv.org/pdf/2605.06115.pdf>)
