ArHFcloud/mistral-7b-detell

Model Overview

ArHFcloud/mistral-7b-detell is a specialized 7 billion parameter language model, built upon the Mistral-7B-Instruct-v0.2 architecture. It has been fine-tuned using QLoRA to address a specific challenge in multiple-choice questions: the presence of "tells."

Key Capabilities

• De-telling Multiple-Choice Questions: The primary function of this model is to identify and rewrite answer choices in multiple-choice questions. This process aims to eliminate surface-level clues or artifacts that might inadvertently reveal the correct answer, thereby improving the quality and fairness of assessments.
• QLoRA Fine-tuned: The model leverages QLoRA (Quantized Low-Rank Adapters) for efficient fine-tuning, with the adapter weights merged directly into the base model for seamless loading and inference.
• Based on Mistral-7B-Instruct-v0.2: Inherits the strong foundational capabilities of the Mistral-7B-Instruct-v0.2 model, providing a robust base for its specialized task.

Good For

• Educational Content Creation: Ideal for developers and educators looking to generate or refine multiple-choice questions for quizzes, exams, or learning platforms, ensuring that questions test knowledge rather than pattern recognition.
• Assessment Design: Useful in scenarios where unbiased and robust assessment questions are critical, helping to prevent test-takers from guessing answers based on superficial cues.
• Research in NLP and Education: Can serve as a tool or baseline for research into question generation, fairness in AI-generated content, and the impact of linguistic features on assessment validity.