ALPS-Lab at SemEval-2026 Task 3: A Multilingual Generative LLM Approach for Dimensional Aspect Sentiment Analysis

2026-06-25 · Source: Paper Index on ACL Anthology · Field: Technology & Digital — Artificial Intelligence & Machine Learning · Depth: Expert, short

Summary

ALPS-Lab introduced a Supervised Fine-Tuning (SFT) approach for the SemEval-2026 Task 3, focusing on Dimensional Aspect Sentiment Analysis (DimABSA). This method utilizes the Gemma-3 27B large language model, enhanced with QLoRA for efficient fine-tuning across multilingual datasets. A key finding is that merging data from various languages significantly boosts performance, particularly in low-resource language domains. The proposed system also incorporates a post-processing step to eliminate duplicate outputs, ensuring accurate evaluation of aspect-level sentiment intensities. This research was presented at the 20th International Workshop on Semantic Evaluation in July 2026.

Key takeaway

For NLP Engineers developing multilingual sentiment analysis systems, especially in low-resource settings, you should consider adopting a Supervised Fine-Tuning approach with QLoRA. Leveraging multilingual data merging can significantly improve model performance. Evaluate the Gemma-3 27B model or similar generative LLMs with this strategy, ensuring robust post-processing to refine output accuracy for aspect-level sentiment intensity predictions.

Key insights

Multilingual data merging and QLoRA fine-tuning enhance aspect-level sentiment analysis with LLMs.

Principles

• Multilingual data merging boosts low-resource domain performance.
• QLoRA enables efficient LLM fine-tuning.

Method

Apply SFT with Gemma-3 27B and QLoRA, fine-tuning on merged multilingual datasets, followed by duplicate output post-processing for accurate evaluation.

In practice

• Use QLoRA for efficient LLM fine-tuning.
• Combine multilingual data for better results.
• Implement post-processing to refine LLM outputs.

Topics

• Dimensional Aspect Sentiment Analysis
• Large Language Models
• QLoRA
• Supervised Fine-Tuning
• Multilingual NLP
• SemEval

Best for: Research Scientist, AI Scientist, NLP Engineer

Related on AIssential

• mdok-style at SemEval-2026 Task 9: Finetuning LLMs for Multilingual Polarization Detection — Finetuning mid-sized LLMs with QLoRA and augmented data improves multilingual online polarization detection.
• Beyond Sentiment Classification: A Generative Framework for Emotion Intensity Evaluation in Text — Generative LLMs fine-tuned on continuous emotion intensity scores outperform classification and generalize to unseen affective dimensions.
• Zero-Shot to Full-Resource: Cross-lingual Transfer Strategies for Aspect-Based Sentiment Analysis — Cross-lingual transfer strategies for ABSA vary in effectiveness based on model architecture and resource availability.
• System Report for CCL25-Eval Task 5: New Dataset and LoRA-Fine-Tuned Qwen2.5 — Domain-specific LoRA fine-tuning on Qwen2.5-14B with CCPoetry-49K significantly enhances classical Chinese poetry understanding and translation.
• Building an LLM fine-tuning Dataset — Building multi-speaker Reddit datasets for LLM fine-tuning is complex but achievable, with efficient QLoRA methods requiring minimal samples.
• Cross-lingual Relation Extraction with Large Language Models: Zero-Shot, Few-Shot, and Fine-Tuned Evaluation on Romanian — QLoRA fine-tuning significantly boosts cross-lingual relation extraction performance for low-resource languages, outperforming prompt-only LLMs.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Paper Index on ACL Anthology.