Multilingual BIO span tagger for cause/effect/signal (mDeBERTa-v3), consumed by reasongraph
This repository is dedicated to a Ph.D. research project that systematically investigates the effectiveness of multilingual transformer models on the task of stance detection. The goal is to not only benchmark these models but also to analyze their ability to handle linguistic challenges, transfer knowledge, and perform under dataset constraints.
SemEval-2026 Task 3 (DimABSA) — Habib University. Four-stage multilingual pipeline for dimensional aspect-based sentiment analysis. 2nd place Tatar, 6th Russian.
MCP server for multilingual place/location extraction (cities, countries, regions, landmarks) in 13 languages, backed by mDeBERTa/ONNX models.
NAMAA Community's submission to the AraGenre 2026 shared task on hierarchical Arabic genre classification (broad genre + fine-grained specific genre, from short Arabic text). This collection holds 10 fine-tuned/ensembled encoder models (E5, mDeBERTa, Qwen2.5-7B few-shot, multi-embedding fusion, and several dev-tuned ensembles)
To associate your repository with the mdeberta topic, visit your repo's landing page and select "manage topics."