This paper describes the system developed bythe team for SemEval-2026 Task 3: Di-mensional Aspect-Based Sentiment Analysis(DimABSA). Unlike traditional categorical sen-timent analysis, predicting continuous Valenceand Arousal (VA) scores across multiple lan-guages and domains poses significant theoret-ical and engineering challenges. To systemat-ically address data scarcity and cross-domaindistribution shifts, we propose a highly robustframework. First, we implement a translation-based data augmentation strategy with preciseHTML-tag alignment to mitigate low-resourceconstraints. Second, we introduce an unsuper-vised opinion extraction module based on syn-tactic dependency parsing to explicitly capturesentiment-bearing words. Third, we designa Tripartite Feature Fusion architecture builtupon both encoder-only (DeBERTa-v3) andcausal LLM (Qwen2.5) models to dynamicallyaggregate global and localized aspect-opinionembeddings. Finally, we apply an unsupervisedTest-Time Adaptation (TTA) mechanism to cal-ibrate normalization layers on the fly. Our sys-tem demonstrates highly competitive perfor-mance while offering critical insights into thelimitations of LLMs in cross-lingual sentimenttransfer.

This paper describes the system developed for the Medical Visual Answer Localization (MVAL) task at MedGenVidQA 2026. Accurately locating surgical or instructional steps in medical videos is inherently challenging due to audio-visual asynchrony and the visual homogeneity of surgical scenes. We propose a Cascade Multi-modal Alignment Framework that integrates Large Language Models (LLMs) to bridge the semantic-temporal gap. Our pipeline utilizes WhisperX for word-level speech transcription to ensure precise textual anchoring. We then employ Gemini3 as a high-level semantic ranker to generate multi-scale textual priors. Crucially, we transform these discrete semantic scores into a continuous 1D Gaussian Soft Prior, which is injected as an attention bias into our cross-modal fusion network. This mechanism preserves global temporal context while guiding the model to focus on query-relevant frames. Our system achieves highly competitive performance on the validation leaderboard, particularly under strict evaluation metrics, reaching an IoU@0.7 of 67.5%.

"Recent advances in large-scale pre-training have substantially enhanced the robustness and generalization capabilities of foundation models (e.g., Qwen3 and Llama-4). However, when fine-tuning them on downstream tasks, these models often latch onto dataset-specific biases, learning spurious correlations tied to easy-to-learn but non-robust features. This undermines their performance under distribution shifts, despite strong in-distribution (ID) accuracy. Existing fine-tuning methods, including full-parameter and parameter-efficient techniques, primarily optimize for ID performance and largely overlook out-of-distribution (OOD) robustness. Meanwhile, debiasing has been explored in full fine-tuning, while debiasing strategies on Parameter-Efficient Fine-Tuning (PEFT) remain underexplored. To this end, in this paper, we propose Enhanced Debiased Gradient Extraction (EDGE), a lightweight gradient projection-based method that explicitly suppresses bias-amplifying updates during fine-tuning process. EDGE is a model-agnostic, and plug-and-play debiasing method that operates without relying on predefined bias types or labels.It seamlessly integrates with both full and parameter-efficient fine-tuning, and generalizes acrossNLP and vision tasks. Experiments on synthetic and real-world benchmarks demonstrate thatEDGE effectively reduces bias and consistently improves OOD generalization, offering a unified and practical framework for robust adaptation under dataset bias."