Qwen3-8B Cybersecurity (merged, v0.1.0-beta)
Beta — not evaluated, not for production.
QLoRA fine-tune of Qwen/Qwen3-8B on cybersecurity Q&A, adapter merged into the base. r=16, alpha=32, 2 epochs, lr 1e-4 cosine.
Generate with enable_thinking=False and call .eval() — the model was trained
with that prompt form.
Security answers may be wrong or outdated. Do not use for real assessment or incident response without expert review.