deberta-v3-base

🔥 QFIRE — a prompt firewall for LLM applications (proxy + CLI + benchmark harness) in Rust (100+ firewall rules).

Full-stack social sentiment intelligence platform for market topics and tickers powered by fine-tuned DeBERTa-v3 with subject-aware sentiment analysis.

Applying zero-shot learning on classification task.

This repository contains my solution for the Kaggle competition Automated Essay Scoring 2.0. The goal of this project is to develop an automated system capable of scoring essays based on their content and quality using machine learning techniques.

This repository presents a complete deep learning pipeline for detecting extremist content in social media text.

AI-powered tool for detecting and classifying prompt safety, identifying potential injections and jailbreaks in both raw text and code inputs.

Word Sense Disambiguation with DeBERTa-v3-base, centroid classification, PCA/t-SNE visualization, and error analysis

Entity framing describes the narrative role assigned to a person, organization, country, or social group in a news article.

CLEF2025_Task4a is the official repository for Subtask 4a of the CLEF2025 CheckThat Lab. It provides a complete pipeline for building and evaluating multi-label classifiers to detect scientific discourse in Twitter data. This project supports reproducible research in social media analysis and NLP.

This repository contains the code and models for the LegalLens-2024 shared task, which detects legal violations in unstructured text and associates them with affected individuals using transformer models like BERT, RoBERTa, and DeBERTa for Named Entity Recognition (L-NER) and Natural Language Inference (L-NLI).

An AI-powered browser extension to detect fake and AI-generated product reviews in real-time. Pair-built with a modern React/Plasmo dashboard interface and a fine-tuned DeBERTa-v3 NLP transformer backend.

An end-to-end retrieval-augmented extractive Question Answering (QA) system for multi-hop reasoning on the HotpotQA benchmark.