paper
When Does Mixing Help? Analyzing Query Embedding Interpolation in Multilingual Dense Retrieval
paperactiveprovisional
when-does-mixing-help-analyzing-query-embedding-interpolation-in-multilingual-dense-retrieval-6ac68b48·1 events·first seen 5d agoAliases: When Does Mixing Help? Analyzing Query Embedding Interpolation in Multilingual Dense Retrieval
Co-occurring entities
More like this (12)
hybrid dense-sparse retrievalSelf-Augmenting Retrieval for Diffusion Language ModelsGranite Embedding Multilingual R2Uncertainty-Aware Hybrid Retrieval for Long-Document RAGCLP: Collocation-Length Prediction for Zero-Loss Adaptive Multi-Token InferenceHow abundant are good interpolators?Semantic Neighbor MixingWhen English Isn't the Best Teacher: Source Language Effects in Cross-Lingual In-Context LearningLeveraging Audio-LLMs to Filter Speech-to-Speech Training Datasentence embeddingsAttention Expansion: Enhancing Keyphrase Extraction from Long Documents with Attention-Augmented Contextualized EmbeddingsSparse Mixture-of-Experts
Recent events (1)
Embedding interpolation study reveals structured benefits of mixed-language queries in multilingual dense retrieval
A ratio-controlled study on mMARCO evaluates how mixing proportions of parallel query translations via embedding-level interpolation affect multilingual dense retrieval performance. Using BGE-M3, the authors find that an optimal mixing ratio outperforms the best monolingual endpoint in 88 of 105 cases, with a clear asymmetry driven by English dominance. Mixing is uniformly beneficial for non-English document indices, while English-containing indices are best served by pure English queries, and mixing gains correlate negatively with typological distance when controlling for English dominance.