qwen3-vl-reranker-2b deploy issue

[newbecoder]

i wanna deploy qwen3-vl-reranker-2b model, and i tried 3 methods:

1. transformers
2. vllm local method: https://github.com/QwenLM/Qwen3-VL-Embedding/blob/main/examples/reranker_vllm.ipynb
3. vllm openai method: https://github.com/vllm-project/vllm/blob/main/examples/pooling/score/vision_rerank_api_online.py

take method 1 as standard answer, i found that method 2 have little difference, but the result of method 3 is exactly wired：all around 0.5X score. anyone can give me some advice?

my query is image, and documents are images

here is my code:

import requests
import json

def rerank(query_image_url, document_image_urls):
    
    documents = {
        "content": []
    }
    
    for url in document_image_urls:
        documents["content"].append({
            "type": "image_url",
            "image_url": {"url": url}
        })
    
    query = {
        "content": [
            {
                "type": "image_url",
                "image_url": {"url": query_image_url}
            }
        ]
    }
    
    payload = {
        "model": "Qwen3-VL-Reranker-2B",
        # "instruction": "Retrieval relevant image or text with user's query",
        "query": query,
        "documents": documents,
    }
    # print(payload)
    
    response = requests.post(
        "http://localhost:8000/v2/rerank",  
        json=payload,
        headers={"Content-Type": "application/json"}
    )
    
    return response.json()

def score(query, documents):
    query_vllm = {
        "content": [
            query
        ]
    }
    documents_vllm = {
        "content": documents
    }

    payload = {
        "model": "Qwen3-VL-Reranker-2B",
        "instruction": "Retrieval relevant image or text with user's query",
        "queries": query_vllm,
        "documents": documents_vllm,
    }
    
    response = requests.post(
        "http://localhost:8000/score",  
        json=payload,
        headers={"Content-Type": "application/json"}
    )
    
    return response.json()

if __name__ == "__main__":
    query_image =  ""
    doc_images = []
    
    result = rerank(query_image, doc_images)
    print("Result :", result)