GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
breakdown881
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Run this model inference with full control and performance in your environment.
Model Details
Model Provider
breakdown881
Model Tree
Qwen/Qwen2.5-7B-Instruct
Input Modalities
Output Modalities
Supported Functionality
Evaluated against ground truth statutory provisions across Civil, Commercial, Corporate, Labor, and Bidding laws:
| Metric / Model | Qwen2.5-7B Base | GPT-4o-mini (Baseline) | VILaw-LLM (SFT) | VILaw-LLM (SFT + DPO) 🏆 |
|---|---|---|---|---|
| Legal Citation Accuracy (0-4) | 1.5 / 4.0 | 2.5 / 4.0 | 3.2 / 4.0 | 3.8 / 4.0 |
| Legal Reasoning Quality (0-3) | 1.2 / 3.0 | 2.1 / 3.0 | 2.5 / 3.0 | 2.8 / 3.0 |
| Practical Actionable Advice (0-3) | 1.1 / 3.0 | 1.6 / 3.0 | 2.2 / 3.0 | 2.5 / 3.0 |
| Overall Score (out of 10) | 3.8 / 10 | 6.2 / 10 | 7.9 / 10 | 9.1 / 10 |
| Hallucination Rate | 58.0% | 32.0% | 16.0% | 0.0% |
python
import torchfrom transformers import AutoModelForCausalLM, AutoTokenizerfrom peft import PeftModelbase_model_id = "Qwen/Qwen2.5-7B-Instruct"adapter_model_id = "your-username/vilaw-dpo-lora" # Or path to local folder# 1. Load Tokenizer & Base Modeltokenizer = AutoTokenizer.from_pretrained(base_model_id)base_model = AutoModelForCausalLM.from_pretrained(base_model_id,torch_dtype=torch.float16,device_map="auto")# 2. Attach DPO LoRA Adaptermodel = PeftModel.from_pretrained(base_model, adapter_model_id)model.eval()# 3. Format with ChatML TemplateSYSTEM_PROMPT = ("Bạn là một chuyên gia tư vấn pháp luật Việt Nam am hiểu sâu sắc các quy định pháp luật. ""Hãy trả lời câu hỏi dựa trên các văn bản quy phạm pháp luật hiện hành, ""viện dẫn chính xác số Điều, Khoản, tên luật và đưa ra lập luận logic, rõ ràng.")messages = [{"role": "system", "content": SYSTEM_PROMPT},{"role": "user", "content": "Thời hiệu khởi kiện yêu cầu giải quyết tranh chấp hợp đồng dân sự là bao lâu?"}]prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)inputs = tokenizer([prompt], return_tensors="pt").to(model.device)# 4. Generate Responsewith torch.no_grad():outputs = model.generate(**inputs,max_new_tokens=512,temperature=0.3,repetition_penalty=1.15)response = tokenizer.decode(outputs[0][len(inputs.input_ids[0]):], skip_special_tokens=True)print(response)
python
from unsloth import FastLanguageModelmodel, tokenizer = FastLanguageModel.from_pretrained(model_name="your-username/vilaw-dpo-lora",max_seq_length=2048,load_in_4bit=True)FastLanguageModel.for_inference(model)# Standard inference workflow...
Qwen2.5-7B-Instructq_proj, k_proj, v_proj, o_proj, gate_proj, up_proj, down_projDPOTrainer with Unsloth PatchDPOTrainerref_model=None parameter-efficient strategy)vbpl.vn).If you utilize VILaw-LLM in your research or application, please cite:
bibtex
@misc{vilaw_llm_2025,title={VILaw-LLM: Vietnamese Legal Reasoning Large Language Model via QLoRA SFT and DPO Alignment},author={breakdown881},year={2025},publisher={Hugging Face},howpublished={\url{https://huggingface.co/breakdown881/vilaw-llm-dpo}}}
This adapter model is distributed under the Apache 2.0 License, aligning with the base model weights of the Qwen series.