Logo Lanfrica
  • Home
  • Atlas
  • Insights
  • Docs
  • Sign in

© 2026 Lanfrica. All rights reserved. All copyrights of the resources shown on the Lanfrica website belong to the original copyright holders, unless explicitly stated otherwise.

Rethinking Cross-lingual Alignment: Balancing Transfer and Cultural Erasure in Multilingual LLMs

Domain:

natural language processing

Record type:

paper
Creator:
HanAgrBri
Host:avatar
Cross-lingual alignment (CLA) aims to align multilingual representations, enabling Large Language Models (LLMs) to seamlessly transfer knowledge across languages. While intuitive, we hypothesize, this pursuit of representational convergence can inadvertently cause "cultural erasure", the functional loss of providing culturally-situated responses that should diverge based on the query language. In this work, we systematically analyze this trade-off by introducing a holistic evaluation framework, the transfer-localization plane, which quantifies both desirable knowledge transfer and undesirable cultural erasure. Using this framework, we re-evaluate recent CLA approaches and find that they consistently improve factual transfer at the direct cost of cultural localization across all six languages studied. Our investigation into the internal representations of these models reveals a key insight: universal factual transfer and culturally-specific knowledge are optimally steerable at different model layers. Based on this finding, we propose Surgical Steering, a novel inference-time method that disentangles these two objectives. By applying targeted activation steering to distinct layers, our approach achieves a better balance between the two competing dimensions, effectively overcoming the limitations of current alignment techniques.

Visit

arxiv.org

Tasks

transfer learning

Tags

Computation and LanguageArtificial Intelligence

Similar

Scaling Intermediate-Task Data for Cross-Lingual Transfer in Multilingual LLMsCross-Lingual Auto Evaluation for Assessing Multilingual LLMsParallel Tokenizers: Rethinking Vocabulary Design for Cross-Lingual TransferLLMs Beyond English: Scaling the Multilingual Capability of LLMs with Cross-Lingual FeedbackMultimodal Alignment Tasks and Zero-Shot Cross-Lingual Transfer in XLM-RWord Alignment Precision and F1 Score Degradation in Cross-lingual NER Transfer

Scaling Intermediate-Task Data for Cross-Lingual Transfer in Multilingual LLMs

Intermediate-task training---fine-tuning a pretrained model on an intermediate task before fine-tuni

Cross-Lingual Auto Evaluation for Assessing Multilingual LLMs

Evaluating machine-generated text remains a significant challenge in NLP, especially for non-English

Parallel Tokenizers: Rethinking Vocabulary Design for Cross-Lingual Transfer

Tokenization defines the foundation of multilingual language models by determining how words are rep

LLMs Beyond English: Scaling the Multilingual Capability of LLMs with Cross-Lingual Feedback

To democratize large language models (LLMs) to most natural languages, it is imperative to make thes

Multimodal Alignment Tasks and Zero-Shot Cross-Lingual Transfer in XLM-R

The introduction of pretrained cross-lingual language models brought decisive improvements to multil

Word Alignment Precision and F1 Score Degradation in Cross-lingual NER Transfer

Cross-lingual Named Entity Recognition (NER) leverages knowledge transfer between languages to ident