machalek29/qwen3-0.6b-state-lifetime-tutor-n62-v2

TEXT GENERATIONPricing:Input $0.04 / Cached $0.008 / Output $0.08Concurrent Unit Cost:1Model Size:0.8BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Aug 19, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

The machalek29/qwen3-0.6b-state-lifetime-tutor-n62-v2 is a 0.8 billion parameter Qwen3-based model fine-tuned by machalek29. It is specifically designed to act as a Python state-lifetime tutor, identifying single mutable-state lifetime bugs in short Python programs. The model's primary function is to ask a single, non-compound question about object creation, ownership, or shared references, without providing corrected code. It has a context length of 32768 tokens.

Loading preview...

Python State-Lifetime Tutor (n=62)

This model, based on Qwen/Qwen3-0.6B with 0.8 billion parameters, functions as a specialized tutor for Python state-lifetime bugs. It is fine-tuned to identify a single mutable-state lifetime bug within a short Python program and then pose exactly one non-compound question related to the object's creation, ownership, or shared references. Crucially, it is designed not to provide corrected code or state the correction directly.

Key Capabilities

  • Bug Identification: Pinpoints a single mutable-state lifetime bug in Python code.
  • Targeted Questioning: Generates a precise, non-compound question about the object's lifecycle.
  • Strict Tutoring Role: Adheres to a pedagogical approach, prompting the user's understanding rather than offering solutions.
  • Optimized for Specific Prompt: Requires a specific system prompt and greedy decoding (do_sample=False) for intended behavior, as it was trained under these conditions.

Training and Evaluation

The model was trained using LoRA on the machalek29/state-lifetime-tutor-v2 dataset, specifically the first 62 examples by rank. It achieved a final training loss of 1.9411 over 24 steps. Evaluation on 36 held-out scenarios (24 clean, 12 adversarial) showed 50% spec adherence on clean scenarios and 67% robustness on adversarial ones, with an 86% mechanical check pass rate.

Good for

  • Educational tools focused on Python state and lifetime concepts.
  • Interactive learning environments where students are guided to discover solutions themselves.
  • Automated code review systems that provide hints rather than direct fixes for specific bug types.

Finetunes
2 models