tokhey/question-generator-model-qwen2.5-1.5b

TEXT GENERATIONConcurrent Unit Cost:1Model Size:1.5BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Jul 14, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

The tokhey/question-generator-model-qwen2.5-1.5b is a 1.5 billion parameter Qwen2.5-Instruct model fine-tuned for generating English multiple-choice questions. It specializes in creating exam-style questions that mimic the structure, wording, and difficulty of official Egyptian Ministry of Education examinations. This model excels at producing curriculum-aligned grammar, vocabulary, reading, and writing questions with plausible distractors and explanations, making it ideal for educational platforms and assessment systems.

Loading preview...

Overview

This model, tokhey/question-generator-model-qwen2.5-1.5b, is a specialized fine-tuned version of Qwen2.5-1.5B-Instruct (1.5 billion parameters, 32K context length) designed for generating English multiple-choice questions. Unlike general-purpose question generators, it is specifically trained to replicate the style, structure, and difficulty of official Egyptian Ministry of Education (General Secondary Stage) examinations.

Key Capabilities

  • Generates Ministry-style English MCQs in structured JSON format.
  • Produces Grammar, Vocabulary, Reading, and Writing questions.
  • Creates plausible distractors and provides explanations for each question.
  • Can generate multiple questions in a single request.
  • Optimized to assess understanding rather than rote memorization.

Training Details

The model was fine-tuned using LoRA with Supervised Fine-Tuning (SFT) on two datasets:

  • Official Egyptian Ministry Dataset: Manually collected from past exams, converted to an instruction-following format with curriculum metadata.
  • Synthetic Dataset: Generated using GLM-5.2 to enhance generalization and diversity while maintaining the Ministry examination style.

Recommended Usage

For optimal performance, users should provide curriculum information in their prompts, including:

  • Unit, Topic, Grammar Focus
  • Target Skills, Life Skills, Core Values
  • Desired JSON output schema

Intended Applications

  • Educational Platforms and AI Tutors
  • Question Generation Systems
  • Ministry-style Practice Exams and Curriculum Assessment Systems
  • Intelligent Learning Platforms

Limitations

  • Primarily optimized for Egyptian Ministry English examinations.
  • Performance is best when detailed curriculum information is supplied in the prompt.
  • Output quality is dependent on prompt specificity.

Finetunes
1 models