Current Aspect-Based Sentiment Analysis (ABSA) often relies on coarse-grained categorical labels, such as Positive and Negative, and this often leads to fail capturing the subtle intensity of emotional expression in real-world text. To address this issue, the SemEval-2026 Shared Task 3: Dimensional ABSA (DimABSA) extends the traditional ABSA by replacing categorical sentiment polarity with continuous valence-arousal (VA) scores. In this paper, we describe our system for Subtask 1 (Dimensional Aspect Sentiment Regression) of Track A (DimABSA). Our system utilizes a DeBERTa-v3-large backbone, enhanced by a prompt-based learning strategy that concatenates aspect information with the context. And we employ multi-sample dropout and a weighted aggregation of the hidden states from the last four layers to capture rich semantic representations. Our experimental results across all provided domains on different languages demonstrate the effectiveness of integrating consistency regularization with dimensional contrastive learning for fine-grained sentiment regression.