When Does Mixing Help? Analyzing Query Embedding Interpolation in Multilingual Dense Retrieval

2026-06-11

AI Summary

This study investigates the impact of mixed-language queries on multilingual dense retrieval, a mechanism previously not well understood. By systematically interpolating monolingual queries from parallel translations in embedding space at controlled ratios, the authors evaluate the effectiveness of mixing strategies. Experiments conducted with models such as BGE-M3 on the mMARCO dataset reveal that in 88 out of 105 scenarios, the optimal mixing ratio outperforms the best monolingual query. Non-English indexes generally benefit from mixed queries, whereas indexes containing English perform best with purely English queries. The work uncovers an asymmetry in mixing gains driven by English dominance and establishes, for the first time, a negative correlation between mixing gain and linguistic typological distance, indicating that sensitivity to query mixing is both structured and predictable.

Abstract

While mixed-language querying is ubiquitous in multilingual communities, the sensitivity of dense retrievers to such queries remains poorly understood. We present a ratio-controlled study on mMARCO that systematically evaluates retrieval performance by varying the mixing proportion of parallel query translations via embedding-level mixing -- constructing mixed queries as an interpolation of monolingual embeddings. Experiments with BGE-M3 demonstrate that an optimal mixing ratio outperforms the best monolingual endpoint in 88/105 cases. We uncover a distinct asymmetry driven by English dominance: mixing is uniformly beneficial when retrieving from non-English document indices, whereas indices containing English are best served by pure English queries. Furthermore, English acts as the strongest mixing partner for every non-English document language. Finally, when controlling for English dominance, mixing gains correlate negatively with typological distance. We conclude that language-mix sensitivity is structured and predictable, and we validate the robustness of these patterns across model families and scales.

Problem

• multilingual dense retrieval
• mixed-language querying
• query embedding interpolation
• language mixing
• retrieval performance

Innovation

• embedding interpolation
• multilingual dense retrieval
• language mixing
• typological distance
• English dominance