Advancements in Multilingual AI and Cross-Lingual Information Retrieval

The field of multilingual AI and cross-lingual information retrieval is rapidly advancing, with a focus on improving the performance of large language models (LLMs) across multiple languages. Recent studies have highlighted the importance of considering the linguistic composition of training data and the need for effective strategies to mitigate the trade-off between cross-lingual and mono-lingual performance. The development of new benchmarks and evaluation frameworks, such as the AI Language Proficiency Monitor and theMarco-Bench-MIF, is facilitating the assessment of LLMs' capabilities across different languages and tasks. Furthermore, research on adapting definition modeling to new languages and on multimodal foundation models' ability to understand schematic diagrams is expanding the scope of multilingual AI. Notably, the introduction of the HanjaBridge technique has shown significant improvements in Korean language understanding, and the MapIQ benchmark is providing insights into the performance of multimodal large language models on map question answering tasks. Overall, the field is moving towards more inclusive and transparent AI systems, with a growing emphasis on evaluating and addressing the performance gaps between high- and low-resource languages. Noteworthy papers include the AI Language Proficiency Monitor, which provides a comprehensive multilingual benchmark for evaluating LLM performance, and the HanjaBridge paper, which presents a novel technique for improving Korean language understanding. The MapIQ benchmark is also noteworthy for its evaluation of multimodal large language models on map question answering tasks.

Sources

Improving Korean-English Cross-Lingual Retrieval: A Data-Centric Study of Language Composition and Model Merging

The AI Language Proficiency Monitor -- Tracking the Progress of LLMs on Multilingual Benchmarks

Banzhida: Advancing Large Language Models for Tibetan with Curated Data and Continual Pre-Training

Adapting Definition Modeling for New Languages: A Case Study on Belarusian

Can Multimodal Foundation Models Understand Schematic Diagrams? An Empirical Study on Information-Seeking QA over Scientific Papers

HanjaBridge: Resolving Semantic Ambiguity in Korean LLMs via Hanja-Augmented Pre-Training

How Many Instructions Can LLMs Follow at Once?

AI Governance InternationaL Evaluation Index (AGILE Index) 2025

MapIQ: Benchmarking Multimodal Large Language Models for Map Question Answering

Cross-lingual Few-shot Learning for Persian Sentiment Analysis with Incremental Adaptation

Marco-Bench-MIF: On Multilingual Instruction-Following Capability of Large Language Models

POLYCHARTQA: Benchmarking Large Vision-Language Models with Multilingual Chart Question Answering

Toxicity-Aware Few-Shot Prompting for Low-Resource Singlish Translation

The first open machine translation system for the Chechen language

Are Knowledge and Reference in Multilingual Language Models Cross-Lingually Consistent?

HATS: Hindi Analogy Test Set for Evaluating Reasoning in Large Language Models

Built with on top of