mlsanityData to 7 Oct 2026

Cross-Lingual Sentiment Misalignment: Auditing Multilingual Language Models for Inversion Risk, Dialectal Representation, and Affective Stability

arXiv 2602.17469 ↗First posted 19 Feb 2026cs.CL cs.HCcited by 5

Not on the map. Why?

Abstract (arXiv)

Recent advances in multilingual representation learning aim to bridge the performance gap between high- and low-resource languages, yet their ability to preserve affective meaning across languages remains underexplored, particularly for underrepresented languages like Bengali. This research addresses cross-lingual sentiment misalignment between Bengali and English by introducing a controlled benchmarking framework evaluating four multilingual transformer models on parallel Bengali-English sentence pairs, stratified by dialect, to assess their representational stability. We demonstrate that a compressed model architecture exhibits a 28.7% "Sentiment Inversion Rate," fundamentally misinterpreting positive semantics as negative (or vice versa). Consequently, we identify a cross-lingual sentiment skew that we call "Asymmetric Empathy," where models systematically dampen or artificially amplify the affective weight of Bengali text relative to its exact English counterpart. Finally, we expose a key vulnerability regarding dialectal representation: a "Modern Bias" in the regional model, which exhibits a 57% increase in alignment error when processing the formal Bengali register compared to modern colloquial text. As foundational encoders continue to serve as safety classifiers and reward models for LLM pipelines, cross-lingual reliability becomes a critical concern. We therefore advocate for the integration of "Affective Stability" metrics into future cross-lingual benchmarks to detect and penalize polarity inversions, particularly in low-resource settings.

Key facts. Cited by 5 papers in the map (data through 7 Oct 2026). Builds on 18 of its 31 references that are in the map. First posted on arXiv 19 Feb 2026.

Builds on

18 of 31 references.

References · 18

  1. 2019Unsupervised Cross-lingual Representation Learning at Scalecited by 3,907
  2. 2020XTREME: A Massively Multilingual Multi-task Benchmark for Evaluating Cross-lingual Generalizationcited by 738
  3. 2021XTREME-R: Towards More Challenging and Nuanced Multilingual Evaluationcited by 167
  4. 2021BanglaBERT: Language Model Pretraining and Benchmarks for Low-Resource Language Understanding Evaluation in Banglacited by 123
  5. 2024Breaking the Curse of Multilinguality with Cross-lingual Expert Language Modelscited by 36
  6. 2023BenLLMEval: A Comprehensive Evaluation into the Potentials and Pitfalls of Large Language Models on Bengali NLPcited by 23
  7. 2024The "Colonial Impulse" of Natural Language Processing: An Audit of Bengali Sentiment Analysis Tools and Their Identity-based Biasescited by 16
  8. 2025CULEMO: Cultural Lenses on Emotion -- Benchmarking LLMs for Cross-Cultural Emotion Understandingcited by 14
  9. 2024Social Bias in Large Language Models For Bangla: An Empirical Study on Gender and Religious Biascited by 13
  10. 2025MuBench: Assessment of Multilingual Capabilities of Large Language Models Across 61 Languagescited by 13
  11. 2025Revisiting Multilingual Data Mixtures in Language Model Pretrainingcited by 10
  12. 2025Script Fragmentation and Format: What Drives the English-Bengali Performance Gap in Open LLMs?cited by 8
6 more
  1. 2025LionGuard 2: Building Lightweight, Data-Efficient & Localised Multilingual Content Moderatorscited by 7
  2. 2025BnMMLU: Measuring Massive Multitask Language Understanding in Bengalicited by 6
  3. 2020The Impact of Indirect Machine Translation on Sentiment Classificationcited by 4
  4. 2024Exploring Bengali Religious Dialect Biases in Large Language Models with Evaluation Perspectivescited by 4
  5. 2025MMAFFBen: A Multilingual and Multimodal Affective Analysis Benchmark for Evaluating LLMs and VLMscited by 4
  6. 2025Reasoning Beyond Labels: Measuring LLM Sentiment in Low-Resource, Culturally Nuanced Contextscited by 2

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 5.

5 papers.

  1. 2026Many Dialects, Many Languages, One Cultural Lens: Evaluating Multilingual VLMs for Bengali Culture Understanding Across Historically Linked Languages and Regional Dialectscited by 3
  2. 2026Register Shifts Break LLM Safety: A Bengali Benchmark with Culturally Grounded Harmscited by 1
  3. 2026OracleZoom: On-Policy Self-Distillation Inspired Reference-Constrained Recursive Image Super Resolutioncited by 1
  4. 2026BEFORE THE FLIP: Measuring Hidden Score Shifts In Quantized Vision Language Models Before The Answer Changes for Visual Question Answeringcited by 1
  5. 2026New Evidence, Same Choice: Testing Physical Experiment Selection in Vision Language Modelscited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025Parallel Tokenizers: Rethinking Encoder Models' Vocabulary Design in Cross-Lingual Transfer of Low-Resource Languages
  2. 2026Cross-Lingual Bias in Large Language Models: A Comparative Analysis of English and Swahili
  3. 2024SentiXRL: An advanced large language Model Framework for Multilingual Fine-Grained Emotion Classification in Complex Text Environment
  4. 2024Exploring Alignment in Shared Cross-lingual Spaces
  5. 2024Evaluating Zero-Shot Multilingual Aspect-Based Sentiment Analysis with Large Language Models
  6. 2026BenSyc: Benchmarking Conversational Sycophancy and Human Alignment in LLMs for Bengali Contexts
  7. 2026Benchmarking Bengali Dialectal Bias: A Multi-Stage Framework Integrating RAG-Based Translation and Human-Augmented RLAIF
  8. 2024TEII: Think, Explain, Interact and Iterate with Large Language Models to Solve Cross-lingual Emotion Detection