Researchers have developed KSE-Web, a system for searching in the low-resource language of Khmer. They analyzed hybrid retrieval methods that combine traditional search algorithms with large language models (LLMs). The study found that traditional BM25 algorithm performed best, but combining it with LLMs improved results. However, using LLMs alone did not outperform non-expanded retrieval. The researchers also noted that direct expansion by LLM can introduce errors and noise. This work highlights the challenges of searching in low-resource languages and provides a foundation for future research. --- Why it matters: This matters to AI researchers because it sheds light on the limitations of using large language models in low-resource languages, where annotated data is scarce. The findings have implications for developing more effective search systems for such languages. Source: https://arxiv.org/abs/2608.21365