KUET at StanceNakba Shared Task: StanceMoE: Mixture-of-Experts Architecture for Stance Detection

arXiv cs.CL / 4/3/2026

💬 OpinionSignals & Early TrendsIdeas & Deep AnalysisModels & Research

共有:

Key Points

The paper introduces StanceMoE, a context-enhanced Mixture-of-Experts architecture for actor-level stance detection that builds on a fine-tuned BERT encoder.
StanceMoE uses six specialized expert modules to capture different heterogeneous linguistic signals, including semantic orientation, lexical cues, clause/phrase-level patterns, framing indicators, and contrast-driven discourse shifts.
A context-aware gating mechanism dynamically routes and weights expert outputs based on the input text characteristics, aiming to better handle diverse stance expressions.
Experiments on the StanceNakba 2026 Subtask A dataset (1,401 English texts with implicit target actors) show StanceMoE achieves a macro-F1 of 94.26%, outperforming baseline and other BERT-based variants.
The work targets the limitation of transformer stance models that rely on a single unified representation, arguing for adaptive architectures that explicitly model varying discourse and framing patterns.

Abstract

Actor-level stance detection aims to determine an author expressed position toward specific geopolitical actors mentioned or implicated in a text. Although transformer-based models have achieved relatively good performance in stance classification, they typically rely on unified representations that may not sufficiently capture heterogeneous linguistic signals, such as contrastive discourse structures, framing cues, and salient lexical indicators. This motivates the need for adaptive architectures that explicitly model diverse stance-expressive patterns. In this paper, we propose StanceMoE, a context-enhanced Mixture-of-Experts (MoE) architecture built upon a fine-tuned BERT encoder for actor-level stance detection. Our model integrates six expert modules designed to capture complementary linguistic signals, including global semantic orientation, salient lexical cues, clause-level focus, phrase-level patterns, framing indicators, and contrast-driven discourse shifts. A context-aware gating mechanism dynamically weights expert contributions, enabling adaptive routing based on input characteristics. Experiments are conducted on the StanceNakba 2026 Subtask A dataset, comprising 1,401 annotated English texts where the target actor is implicit in the text. StanceMoE achieves a macro-F1 score of 94.26%, outperforming traditional baselines, and alternative BERT-based variants.