Qwen3-8B SOC v11 LoRA Adapter
LoRA adapter fine-tuned from Qwen/Qwen3-8B on a distilled SOC analyst SFT dataset.
Training
- Base model: Qwen/Qwen3-8B
- Method: BF16 base LoRA
- LoRA rank: 32
- LoRA alpha: 64
- Max sequence length: 8192
- Epochs: 1
- Learning rate: 1e-4
- Dataset: distilled SOC v11 accepted SFT
- Branches: L1 investigation, advanced reasoning, agentic tool-use, ticket writing, L2 review
Intended use
SOC analyst assistance, alert triage, evidence-grounded investigation, case ticket writing, and L2-style review.
Notes
This is a LoRA adapter, not a merged full model. Use with base model Qwen/Qwen3-8B.
