While large language models (LLMs) demonstrate remarkable success in multilingual translation, their internal core translation mechanisms, even at the fundamental word level, remain insufficiently understood. To address this critical gap, this work introduces a systematic framework for interpreting the mechanism behind LLM translation from the perspective of computational components. This paper first proposes subspace-intervened path patching for precise, fine-grained causal analysis, enabling the detection of components crucial to translation tasks and subsequently characterizing their behavioral patterns in human-interpretable terms. Comprehensive experiments reveal that translation is predominantly driven by a sparse subset of components: specialized attention heads serve critical roles in extracting source language, translation indicators, and positional features, which are then integrated and processed by specific multi-layer perceptrons (MLPs) into intermediary English-centric latent representations before ultimately yielding the final translation. The significance of these findings is underscored by the empirical demonstration that targeted fine-tuning a minimal parameter subset (< 5%) enhances translation performance while preserving general capabilities. This result further indicates that these crucial components generalize effectively to sentence-level translation and are instrumental in elucidating more intricate translation tasks. Code is available at this URL.

Your cell phone can handle basic language translation, but bespoke tools can offer a much more immersive experience. Hans Christian Andersen once said, "To travel is to live," and while that's a romantic notion, he probably wasn't careening through Gyeongju, South Korea, at midnight in the back of a taxi with a driver who didn't speak a lick of English. Today's world traveler has it awfully easy when it comes to understanding the local lingo, as even a basic modern cell phone app can offer a pretty good translation of common phrases delivered in everything from Abkhaz to Zulu. Type or speak a sentence or two into the app, tap a button, and out it returns in the language of your choice. Tap another button, and your phone can even speak those sentences aloud.

Figure 5 shows an schema explaining the extraction of the entities. Each step is depicted in a triplet format: subject,predicate,object . Blue (italics) information is the information extracted at each step. For each step outlined with a dotted rectangle (), the information extracted is the subject; otherwise, the information extracted is the object. Figure 6 show an example of multilingual alignment for the languages considered in the high-resource use case: English, Arabic, Spanish and Russian.

We then train and release a 10.7B-parameter

Neural Information Processing Systems http://nips.cc/

Awordvectorspace-sometimes referred toasawordembedding-associates similar wordsina vocabularywithsimilar vectors.