Exploring Cross-lingual Latent Transplantation: Mutual Opportunities and Open Challenges

arXiv cs.CL / 4/13/2026

💬 OpinionSignals & Early TrendsIdeas & Deep AnalysisModels & Research

共有:

Key Points

The paper proposes XTransplant, a cross-lingual latent transplantation framework that transfers latent activations across languages during inference to better use an LLM’s internal multilingual knowledge.
Experiments indicate mutually beneficial improvements in both multilingual capability and cultural adaptability, with the strongest gains observed for low-resource languages and cultures.
The study finds architectural roles: attention modules help multilingual understanding, while feed-forward modules capture more culture-specific information.
Detailed analysis covers XTransplant’s stability, effectiveness, and generalizability, including an upper-bound probe that suggests current LLMs underutilize their available multilingual potential.
The authors position XTransplant as a new lens for designing and evaluating cross-lingual interactions, while highlighting remaining open challenges in exploiting these capabilities broadly.

Abstract

Current large language models (LLMs) often exhibit imbalances in multilingual capabilities and cultural adaptability, largely attributed to their English-centric pre-training data. In this paper, we introduce and investigate cross-lingual latent transplantation (XTransplant), a probing framework which aims to further exploit the model's internalized multilingual knowledge during inference and examine its effects on the multilingual capability and cultural adaptability of LLMs. XTransplant framework enables models to harness the complementary strengths of both English and non-English resources by transplanting latent activations across languages. Through extensive analysis, we empirically demonstrate that XTransplant, a form of cross-lingual interaction, has mutually beneficial effects on the multilingual capability and cultural adaptability of LLMs, particularly for low-resource languages and cultures. We further reveal that attention modules play a pivotal role in supporting multilingual understanding, while feed-forward modules are more adept at capturing culture-specific knowledge. In addition, we conduct in-depth analysis of XTransplant's stability, effectiveness, and generalizability. By probing the upper bound performance of XTransplant, we expose the considerable underutilization of current LLMs' multilingual potential-a challenge that remains open. We hope our analysis offers a new lens for advancing cross-lingual interactions and better leveraging models' internalized multilingual knowledge.