KCLarity at SemEval-2026 Task 6: Encoder and Zero-Shot Approaches to Political Evasion Detection

TL;DR

This paper explores encoder and zero-shot models for detecting political evasion in discourse, comparing formulations and training variants, with RoBERTa-large and GPT-5.2 showing strong results.

Contribution

It introduces two modeling approaches for evasion detection, evaluates auxiliary training and zero-shot methods, and compares encoder-based and decoder-only models.

Findings

RoBERTa-large achieves top performance on the public test set.

Zero-shot GPT-5.2 generalizes better on hidden evaluation data.

Both formulations perform comparably in detection accuracy.

Abstract

This paper describes the KCLarity team's participation in CLARITY, a shared task at SemEval 2026 on classifying ambiguity and evasion techniques in political discourse. We investigate two modelling formulations: (i) directly predicting the clarity label, and (ii) predicting the evasion label and deriving clarity through the task taxonomy hierarchy. We further explore several auxiliary training variants and evaluate decoder-only models in a zero-shot setting under the evasion-first formulation. Overall, the two formulations yield comparable performance. Among encoder-based models, RoBERTa-large achieves the strongest results on the public test set, while zero-shot GPT-5.2 generalises better on the hidden evaluation set.