BERTScore uses bert-base-chinese. ROUGE uses character-level tokenization for Chinese. ROUGE decreases because the fine-tuned model learned to give concise, on-format answers rather than verbose outputs — BERTScore (semantic similarity) is the primary metric.

Model Comparison

Model	Dataset	Rank	BERTScore (ft)	Δ
MedQwen-1.5B (v1)	original split	r=8	0.6620	+5.61%
MedQwen-1.5B (v2)	80/10/10 split	r=16	0.6726	+8.34%
MedQwen-3B	original split	r=16	0.6751	+5.92%
MedQwen-7B	original split	r=8	0.6670	+7.11%

Training Configuration

Parameter	Value
Base model	Qwen2.5-1.5B-Instruct
LoRA rank	16
LoRA alpha	32
LoRA dropout	0.05
Target modules	q_proj, k_proj, v_proj, o_proj, gate_proj, up_proj, down_proj
Training samples	24,531 (80% of 30,664)
Validation samples	3,066 (10%) — used for early stopping
Test samples	3,067 (10%) — held out for final evaluation
Max sequence length	256
Batch size (effective)	8
Learning rate	2e-4
LR schedule	Cosine with warmup
Epochs	3
Early stopping	patience=5, eval every 500 steps
Hardware	GCP L4 24GB
Precision	FP16

Usage

python

from transformers import AutoTokenizer, AutoModelForCausalLM
from peft import PeftModel
import torch

base_model_id = "Qwen/Qwen2.5-1.5B-Instruct"
adapter_id    = "mellee030/MedQwen-1.5B-LoRA-r16"

tokenizer = AutoTokenizer.from_pretrained(base_model_id, trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(
    base_model_id, dtype=torch.float16, device_map="auto", trust_remote_code=True
)
model = PeftModel.from_pretrained(model, adapter_id)
model.eval()

messages = [
    {"role": "system", "content": "你是一个专业的医疗问答助手，请根据用户的问题给出准确、简洁的医疗建议。"},
    {"role": "user",   "content": "糖尿病的早期症状有哪些？"},
]
text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
inputs = tokenizer(text, return_tensors="pt").to(model.device)

with torch.no_grad():
    outputs = model.generate(**inputs, max_new_tokens=256, temperature=0.7, do_sample=True)
print(tokenizer.decode(outputs[0][inputs.input_ids.shape[1]:], skip_special_tokens=True))

Dataset

~30,664 Chinese medical Q&A pairs derived from a structured Chinese medical knowledge base containing disease records with fields such as name, description, symptoms, causes, and treatments. Split 80/10/10 with seed=42 for reproducibility.

Disclaimer

This model is for research purposes only and should not be used as a substitute for professional medical advice.