ZiyiYe/Con-J-Qwen2-7B

Introduction

Con-J-Qwen2-7B (learning the generative Judge using self-generated Contrastive judgments) is an advanced generative judge built on Qwen2-7B-Instruct architecture and dataset Skywork/Skywork-Reward-Preference-80K-v0.1.

Con-J-Qwen2-7B is trained from preference data. We prompt the pre-trained Qwen2-7B-Instruct model to generate positive and negative judgments, both supported with rationales in natural language form. Then the self-generated contrastive judgment pairs are used to train the generative judge with Direct Preference Optimization (DPO). By doing this, Con-J learns to act as a generative judge and provides accurate and supprting rationales.

Usage

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer

model_name = "ZiyiYe/Con-J-Qwen2-7B"
model = AutoModelForCausalLM.from_pretrained(model_name, device_map='auto', trust_remote_code=True)
tokenizer = AutoTokenizer.from_pretrained(model_name)

question = "What is the range of the numeric output of a sigmoid node in a neural network?"
answer1 = "The output of a sigmoid node is bounded between -1 and 1."
answer2 = "The output of a sigmoid node is bounded between 0 and 1."

# Format and tokenize the conversations
CON_J_PROMPT = """作为一个评价专家，给定一个问题和它的两个可能的回答，请选出哪一个回答在连贯性、准确性、覆盖度和上述定义的整体质量方面最为符合。请用JSON格式输出你的判断, 其中"原因"是你提供的解释，"更好的回答"是整数类型的1或2，例如{{"原因": "你的解释", "更好的回答": 1}}。以下是问题和候选回答的内容：
    \n问题：{instruction}
回答1：{output_1}
回答2：{output_2}"""
user_prompt = CON_J_PROMPT.format(instruction=question, output_1=answer1, output_2=answer2)
system_prompt = ""
messages = [
    {"role": "system", "content": system_prompt,},
    {"role": "user", "content": user_prompt},
]
prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
prompt = tokenizer([prompt], return_tensors="pt")

# Generate judgment for the given prompt
with torch.no_grad():
    generated_ids = model.generate(prompt.input_ids, do_sample=False, max_new_tokens=2048,)
generated_ids = [
    output_ids[len(input_ids):] for input_ids, output_ids in zip(prompt.input_ids, generated_ids)
]
response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]

# response: {"原因": "回答1中的-1是错误的，因为sigmoid函数的实际输出范围是0到1，而不是包括-1。回答2准确地描述了sigmoid函数的输出范围是0到1。",\n "更好的回答": 2}

Performance

Model	Infinity- Preference	Ultra- Feedback	PKU- SafeRLHF	Reward-Bench
				Chat	Chat-H	Safety	Reasoning
Llama3.1-8B	59.0	62.9	66.4	80.7	49.8	64.0	68.1
Llama3.1-70B	64.0	71.4	67.6	97.2	70.2	82.8	86.0
Qwen2-7B	59.0	64.5	67.2	91.3	44.8	73.6	69.0
Qwen2.5-72B	70.0	66.0	58.7	86.6	61.4	74.5	90.7
Auto-J	69.0	63.9	66.9	93.0	40.0	65.5	50.5
Prometheus 2	68.0	63.3	63.0	85.5	49.1	77.1	76.5
GPT-4o	75.0	72.2	69.6	95.3	74.3	87.6	86.9
Con-J (ours)	81.0	73.0	68.4	91.3	79.6	88.0	87.1

Reference

@misc{ye2024scalarrewardmodellearning,
      title={Beyond Scalar Reward Model: Learning Generative Judge from Preference Data}, 
      author={Ziyi Ye and Xiangsheng Li and Qiuchi Li and Qingyao Ai and Yujia Zhou and Wei Shen and Dong Yan and Yiqun Liu},
      year={2024},
      eprint={2410.03742},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2410.03742}, 
}