Language Drift in Multilingual Retrieval-Augmented Generation: Characterization and Decoding-Time Mitigation
By: Bo Li, Zhenghua Xu, Rui Xie
Potential Business Impact:
Stops AI from switching languages during translation.
Multilingual Retrieval-Augmented Generation (RAG) enables large language models (LLMs) to perform knowledge-intensive tasks in multilingual settings by leveraging retrieved documents as external evidence. However, when the retrieved evidence differs in language from the user query and in-context exemplars, the model often exhibits language drift by generating responses in an unintended language. This phenomenon is especially pronounced during reasoning-intensive decoding, such as Chain-of-Thought (CoT) generation, where intermediate steps introduce further language instability. In this paper, we systematically study output language drift in multilingual RAG across multiple datasets, languages, and LLM backbones. Our controlled experiments reveal that the drift results not from comprehension failure but from decoder-level collapse, where dominant token distributions and high-frequency English patterns dominate the intended generation language. We further observe that English serves as a semantic attractor under cross-lingual conditions, emerging as both the strongest interference source and the most frequent fallback language. To mitigate this, we propose Soft Constrained Decoding (SCD), a lightweight, training-free decoding strategy that gently steers generation toward the target language by penalizing non-target-language tokens. SCD is model-agnostic and can be applied to any generation algorithm without modifying the architecture or requiring additional data. Experiments across three multilingual datasets and multiple typologically diverse languages show that SCD consistently improves language alignment and task performance, providing an effective and generalizable solution in multilingual RAG.
Similar Papers
Evaluating the Effect of Retrieval Augmentation on Social Biases
Computation and Language
Makes AI share unfair ideas more often.
Guided Decoding and Its Critical Role in Retrieval-Augmented Generation
Computation and Language
Makes AI give answers that are correct and useful.
Multilingual Retrieval-Augmented Generation for Knowledge-Intensive Task
Computation and Language
Helps computers answer questions in any language.