You can deploy this model and start earning money today!

phai17/
lab21-2A202601575-qwen35-triage-vi

Name: lab21-2A202601575-qwen35-triage-vi
Availability: InStock

Views

No views yet

Lab 21: Qwen3.5-4B LoRA Fine-Tuning for Customer Service Triage (Vietnamese)

Học viên: Hoàng Văn Phái
MSSV: 2A202601575
Module: AICB-P2T3 · Ngày 21 · Chương 5 — Fine-tuning & An Toàn (Track 3)
Base Model: unsloth/Qwen3.5-4B
Adapter: LoRA Without Regret (target_modules=text-linear, r=16, alpha=32, lr=1e-4)

Run	Target Task (Triage Acc)	Regression (General QA)	Format (Valid JSON)	Latency (ms)
(a) Base + Naive Prompt	0.000	0.758	0.000	3308.5
(b) Base + Optimized Prompt	0.765	0.758	1.000	1037.9
(c) LoRA Fine-tune (`correct`)	0.970 (+0.205)	0.522 (-0.236)	1.000	1428.5

submission/REPORT.md: Toàn bộ báo cáo phân tích chi tiết.
submission/REFLECTION.md: Trả lời phản tư cá nhân.
results/: Đầy đủ các file JSON đo đạc (verdict.json, autopsy.json, baselines_frozen.json, runs.csv, qualitative.json, mask_proof.json).
adapter_model.safetensors & adapter_config.json: Trọng số LoRA adapter.