machalek29/qwen3-0.6b-state-lifetime-tutor-n500
The machalek29/qwen3-0.6b-state-lifetime-tutor-n500 is a 0.8 billion parameter Qwen3-based language model specifically fine-tuned to act as a Python state-lifetime tutor. It identifies mutable-state lifetime bugs in short Python programs and asks a single, non-compound question about object creation, ownership, or shared references. This model excels at providing targeted, pedagogical feedback without offering direct code corrections, making it suitable for educational applications.
Loading preview...
Model Overview
The machalek29/qwen3-0.6b-state-lifetime-tutor-n500 is a specialized language model built upon the Qwen/Qwen3-0.6B base, featuring 0.8 billion parameters and a 32768-token context length. Its unique purpose is to function as a Python state-lifetime tutor, designed to help users understand specific types of bugs in their code.
Key Capabilities
- Bug Identification: Given a short Python program with a single mutable-state lifetime bug, the model can accurately quote or identify the relevant declaration, assignment, or mutation causing the issue.
- Targeted Questioning: It asks precisely one non-compound question related to the object's creation, ownership, or shared references, guiding the user towards understanding the underlying problem.
- Pedagogical Approach: Crucially, the model never emits corrected code or states the correction directly, even when prompted. This design encourages active learning and problem-solving by the user.
Training and Evaluation
The model was fine-tuned using LoRA (r=16, alpha=16) on the machalek29/state-lifetime-tutor-v1 dataset, specifically the first 500 examples. Training involved 189 steps, achieving a final loss of 0.5029. Evaluation was conducted on 36 held-out scenarios (24 clean, 12 adversarial) using a frozen LLM judge, demonstrating:
- 100% Spec Adherence on clean scenarios.
- 100% Robustness on adversarial scenarios.
- 97% Mechanical Check Pass.
Recommended Usage
To ensure optimal performance as observed during training and evaluation, users should apply the provided chat template with the system prompt "You are a Python state-lifetime tutor." It is critical to disable thinking (enable_thinking=False) and use greedy decoding (do_sample=False).