# NCL-BU at SemEval-2026 Task 3: Fine-tuning XLM-RoBERTa for Multilingual Dimensional Sentiment Regression

Source: [NCL-BU at SemEval-2026 Task 3: Fine-tuning XLM-RoBERTa for Multilingual Dimensional Sentiment Regression](<https://www.sota2.com/research/paper/69dc78597d0172140ec788ba>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-05-12T05:00:04.826Z

Authors: Tong Wu, Nicolay Rusnachenko, Huizhi Liang

Published: 2026-04-10T03:38:53.000Z

## Abstract

Dimensional Aspect-Based Sentiment Analysis (DimABSA) extends traditional ABSA from categorical polarity labels to continuous valence-arousal (VA) regression. This paper describes a system developed for Track A, Subtask 1 (Dimensional Aspect Sentiment Regression), aiming to predict real-valued VA scores in the \[1, 9\] range for each given aspect in a text. A fine-tuning approach based on XLM-RoBERTa-base is adopted, with dual regression heads with sigmoid-scaled outputs for valence and arousal prediction. Separate models are trained for each language-domain pair (English and Chinese across restaurant, laptop, and finance domains), and training and development sets are merged for final test predictions. In development experiments, the fine-tuning approach is compared against several large language models under a few-shot prompting setting, demonstrating that task-specific fine-tuning outperforms these LLM-based methods across all evaluation datasets.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [VA Regression on English Laptop eng\_lap (test)](<https://www.sota2.com/research/sota/va-regression-on-english-laptop-eng_lap-test.md>): VA Regression; English Laptop eng\_lap (test); RMSE (VA): 1.4562.

- [VA Regression on English Restaurant (test)](<https://www.sota2.com/research/sota/va-regression-on-english-restaurant-test.md>): VA Regression; English Restaurant (test); RMSE (VA): 1.4861.

- [VA Regression on Chinese Laptop zho\_lap (test)](<https://www.sota2.com/research/sota/va-regression-on-chinese-laptop-zho_lap-test.md>): VA Regression; Chinese Laptop zho\_lap (test); RMSE (VA): 0.751.

- [VA Regression on Chinese Restaurant zho\_res (test)](<https://www.sota2.com/research/sota/va-regression-on-chinese-restaurant-zho_res-test.md>): VA Regression; Chinese Restaurant zho\_res (test); RMSE (VA): 0.9553.

- [VA Regression on Chinese Finance zho\_fin (test)](<https://www.sota2.com/research/sota/va-regression-on-chinese-finance-zho_fin-test.md>): VA Regression; Chinese Finance zho\_fin (test); RMSE\_VA: 0.5391.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2604.08923>)

[PDF](<https://arxiv.org/pdf/2604.08923.pdf>)