mjpsm/activity-generation-model-v1.1

TEXT GENERATIONPricing:Input $0.04 / Cached $0.008 / Output $0.08Concurrent Unit Cost:1Model Size:0.5BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Aug 21, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

The mjpsm/activity-generation-model-v1.1 is a 0.5 billion parameter instruction-tuned model, fine-tuned from Qwen/Qwen2.5-0.5B-Instruct, designed by mjpsm for conditional educational activity generation. It specializes in generating small, micro-progression learning activities in JSON format based on a learner's goal, previous activity, and knowledge submission. This model excels at producing concise, focused next steps for learning workflows, ensuring 100% valid JSON and schema compliance.

Loading preview...

Activity Generation Model v1.1 Overview

The mjpsm/activity-generation-model-v1.1 is a specialized 0.5 billion parameter model, fine-tuned from Qwen/Qwen2.5-0.5B-Instruct, for generating educational activities within the MyVillage learning workflow. Its core function is to produce a single, small, and highly focused next learning activity based on a learner's broader goal, their most recently completed activity, and their knowledge submission.

Key Capabilities & Features

  • Micro-Progression Focus: Generates the smallest meaningful next step, avoiding large assignments or multi-step projects.
  • Structured Output: Returns activities in a strict JSON format with title, description, and instructions fields.
  • Contextual Generation: Utilizes Village goal, previous activity title, and knowledge submission to tailor activities.
  • Improved Conciseness: Version 1.1 significantly reduces instruction and description lengths, with average instruction length decreasing by 35.5% and description length by 32.7% compared to v1.
  • High Compliance: Maintains 100% valid JSON and exact schema compliance.
  • Performance: Achieves 95% single-sentence instructions and 90% micro-activity heuristic pass rate on benchmarks.

Ideal Use Cases

This model is best suited for learning systems that require:

  • Incremental Learning: Systems where activities should progress incrementally rather than assigning large projects.
  • Automated Activity Suggestions: Generating short, focused next steps after a learner completes an activity and submits their learning.
  • Structured Data Integration: Applications that can consume and validate JSON-formatted activity suggestions.

Finetunes
2 models