SemEval-2026 Task 5 — Graded word-sense plausibility rating via a 3-component ensemble: zero-shot NLI, gloss-informed DeBERTa regressor, and a novel Likert-distribution head with uncertainty estimation.
natural-language-processing transformers pytorch word-sense-disambiguation ensemble-learning model-calibration semeval uncertainty-estimation semantic-segmentation natural-language-inference lexical-semantics huggingface deberta semeval-2025 narrative-understanding
-
Updated
May 19, 2026 - Jupyter Notebook