mjpsm/activity-generation-v1.3.5-qwen0.5b-427

TEXT GENERATIONPricing:Input $0.04 / Cached $0.008 / Output $0.08Concurrent Unit Cost:1Model Size:0.5BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Sep 22, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

The mjpsm/activity-generation-v1.3.5-qwen0.5b-427 model is a 0.5 billion parameter Qwen2.5-based language model fine-tuned by mjpsm for generating next-activity suggestions within the MyVillage platform. It specializes in producing structured JSON outputs containing activity titles, descriptions, and instructions, leveraging a 32768-token context length. This model is specifically optimized for educational activity generation based on learner submissions and village goals, distinguishing it from general-purpose LLMs.

Loading preview...

Overview

This model, mjpsm/activity-generation-v1.3.5-qwen0.5b-427, is a specialized 0.5 billion parameter language model fine-tuned from Qwen/Qwen2.5-0.5B-Instruct. Its primary purpose is to generate next-activity suggestions for the MyVillage platform, focusing on educational and learning contexts.

Key Capabilities

  • Activity Generation: Produces structured JSON output including title, description, and instructions for educational activities.
  • Contextual Understanding: Processes inputs such as village goals, previous activity titles, knowledge submissions, and specific wisdom objects (book name, type, chapter, content).
  • Learner-Centric Design: Treats knowledge submissions as the strongest evidence of a learner's current state, guiding activity generation to avoid repetition and align with long-term village goals.
  • Behavioral Nuances: Trained to handle vague submissions conservatively and avoid common linguistic shortcuts like the word "but" in its outputs.

Training Details

The model was trained using supervised fine-tuning with QLoRA on a dataset of 427 examples. It underwent 4 epochs with a learning rate of 0.0001 and a LoRA rank of 16, targeting a context length of 1024 tokens during training.

Important Limitation

This is an experimental checkpoint based on a small synthetic dataset. Its behavioral quality requires thorough evaluation on held-out MyVillage scenarios before any production deployment.