Cherish at SemEval-2026 Task 2: Enhancing RoBERTa-Based Models for Emotional Valence and Arousal Prediction in Ecological Essays with Personalized PLoRA and Temporal Embeddings

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

Summary

Team Cherish developed a system for SemEval-2026 Task 2, aimed at predicting emotional valence and arousal variations over time in ecological essays. Their approach integrates personalization and temporal data into a transformer-based architecture, utilizing a RoBERTa-large backbone. This model is enhanced with Personalized PLoRA and a temporal embedding module, designed to retain broad semantic understanding while adapting to individual users and emotional changes across different timeframes. The system secured 13th place among 29 teams in Subtask 1, achieving a Pearson's r composite score of 0.596 for valence prediction and 0.505 for arousal prediction. Although the team also participated in Subtask 2a, technical inference issues resulted in zero variance predictions and an undefined correlation score.

Key takeaway

For NLP Engineers developing emotion prediction systems, consider integrating personalized and temporal components into your transformer architectures. Your models can achieve better performance in dynamic, user-generated text by adapting to individual emotional shifts over time. Specifically, explore augmenting RoBERTa-based backbones with techniques like PLoRA and temporal embeddings to enhance valence and arousal prediction accuracy.

Key insights

A RoBERTa-based model enhanced with personalized PLoRA and temporal embeddings effectively predicts emotional valence and arousal in ecological essays.

Principles

• Emotional dynamics require personalization and temporal context.
• Transformer models can adapt to individual users and temporal shifts.
• Combining general semantic knowledge with fine-grained adaptation is key.

Method

The system uses a RoBERTa-large encoder, augmented with a Personalized PLoRA module for user adaptation and a temporal embedding module to capture emotional shifts over time.

In practice

• Integrate PLoRA for user-specific model adaptation.
• Employ temporal embeddings for time-sensitive text analysis.

Topics

• Emotional Valence Prediction
• Arousal Prediction
• RoBERTa-large
• PLoRA
• Temporal Embeddings
• SemEval-2026

Best for: Research Scientist, AI Scientist, NLP Engineer

Related on AIssential

• 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.
• An End-to-End Hybrid Framework for Rumour Detection in Low-Resources Algerian Dialect — A hybrid framework combining transformer embeddings with classical classifiers enables effective rumour detection in low-resource Algerian dialect.
• From Affect Prediction to Affect Forecasting: Evidence for Distinct Information Sources in Longitudinal Text — Current affect prediction relies on text semantics, while future affective change forecasting benefits from numeric trajectory dynamics.
• Exploration of symbolic methods for emotion detection for Portuguese — Symbolic emotion detection in Portuguese shows stable relative performance despite domain-specific absolute variations.
• Learning Perspectivist Social Meaning via Demographic-Conditioned Fusion Embeddings — Social meaning is perspectival; models can predict how interpretations vary across demographic groups using fusion embeddings.
• Aspect-Based Sentiment Analysis in the Accommodation Domain Using the BERTimbau Model — BERTimbau model fine-tuning strategies for Portuguese ABSA show trade-offs between recall and precision.

Open in AIssential →

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