---
base_model: Qwen/Qwen2.5-0.5B-Instruct
library_name: transformers
pipeline_tag: text-generation
tags:
- fine-tuned
- activity-generation
- education
- structured-output
- qwen
---

# Activity Generation Model v0.1

This model generates one next educational activity from a student's
knowledge submission.

## Input

A student knowledge submission.

## Output

The model is trained to return JSON with:

- `title`
- `description`
- `instructions`
- `activityType`
- `estimatedMinutes`

Allowed activity types:

- `REFLECTION`
- `RESEARCH`
- `COLLABORATE`
- `CREATE`
- `PRACTICE`
- `EXPERIENCE`
- `TEACH`

## Training data

The proof-of-concept training dataset contained approximately 200
synthetic input-output examples derived from real knowledge-submission
patterns.

## Limitations

This is an early proof-of-concept model trained on a small and imbalanced
dataset. It may favor common activity types such as CREATE or PRACTICE,
produce inaccurate time estimates, or return malformed JSON.