machalek29/qwen3-0.6b-state-lifetime-tutor-n250
The machalek29/qwen3-0.6b-state-lifetime-tutor-n250 is a 0.8 billion parameter Qwen3-based model, fine-tuned to act as a Python state-lifetime tutor. It is specifically designed to identify and ask a single, non-compound question about mutable-state lifetime bugs in short Python programs. This model focuses on pinpointing relevant declarations, assignments, or mutations without providing corrected code or direct solutions, operating with a 32768 token context length.
Loading preview...
Overview
The machalek29/qwen3-0.6b-state-lifetime-tutor-n250 is a specialized 0.8 billion parameter language model built on the Qwen3 architecture. Its core function is to serve as a Python state-lifetime tutor, identifying single mutable-state lifetime bugs within short Python programs. The model's unique characteristic is its constrained output: it will quote or identify the relevant code segment and then ask exactly one non-compound question about object creation, ownership, or shared references, without ever emitting corrected code.
Key Capabilities
- Precise Bug Identification: Accurately pinpoints declarations, assignments, or mutations related to state-lifetime bugs.
- Targeted Questioning: Formulates a single, focused question about the object's lifecycle, ownership, or references.
- Strict Adherence to Tutoring Role: Never provides direct code corrections or solutions, encouraging user learning.
- High Reliability: Achieved 100% spec adherence and robustness on 36 held-out evaluation scenarios, including adversarial cases.
Training and Usage
The model was trained using LoRA on the machalek29/state-lifetime-tutor-v1 dataset (first 250 examples) for 96 steps. It requires a specific system prompt and greedy decoding (do_sample=False) with thinking off for optimal performance, as this matches its training conditions. Its 32768 token context length allows for analysis of reasonably sized code snippets.