Hugging Face's logo

AQ-MedAI
/
Diver-Retriever-4B

Text Ranking
Transformers
Safetensors
Chinese
English
qwen3
text-generation
medical
code
math
reasoning
general
text-embeddings-inference
Model card Files
xet
Community
Diver-Retriever-4B / README.md
prayerdan's picture
prayerdan
Update README.md
ff7f8bc verified
preview code
|
raw
history blame contribute delete
24.7 kB
 ---
license: apache-2.0
tags:
- medical
- code
- math
- reasoning
- general
datasets:
- Raderspace/MATH_qCoT_LLMquery_questionasquery_lexicalquery
- reasonir/reasonir-data
- truehealth/medqa
- AQ-MedAI/PRGB-ZH
metrics:
- accuracy
- recall
base_model:
- Qwen/Qwen3-Embedding-4B
pipeline_tag: text-ranking
language:
- zh
- en
library_name: transformers
---
# Diver-Retriever-4B
## HighLights
The Diver Retriever 4B model is a reasoning-intensive model designed to tackle the challenge of reasonIR and rader.
We combined data from the fields of mathematics, coding, and healthcare.
At the same time, we made precise matching in terms of the difficulty level of the samples, and uniquely
constructed negative samples corresponding to each field. Therefore, this model performs very well on the Bright LeaderBoard
as well as the Mteb-Medical Benchmark.
### Model Description
<!-- Provide a longer summary of what this model is. -->
- **Model type:** Text Embedding
- **Language(s) (NLP):** Bilingual (Chinese & English)
- **Context Length:** 40k
- **Number of Paramaters:** 4B
For more details, including benchmark evaluation, hardware requirements, and inference performance, please refer to our GitHub (https://github.com/AQ-MedAI/Diver).
| **Model** | **#Total Params** | **Context Length** | **Download** | **BRIGHT** |
| :------------------: | :---------------: | :----------------: | :--------------------------------------------------------------------------------------------------------------------------------------------------------: | :----------------: |
| DIVER-Retriever-4B | 4B | 40K | [🤗 HuggingFace]https://huggingface.co/AQ-MedAI/Diver-Retriever-4B <br>[🤖 ModelScope]https://www.modelscope.cn/models/AQ-MedAI/Diver-Retriever-4B | **28.9** |
| DIVER-Retriever-1.7B | 1.7B | 40K | [🤗 HuggingFace]https://huggingface.co/AQ-MedAI/Diver-Retriever-1.7B <br>[🤖 ModelScope]https://www.modelscope.cn/models/AQ-MedAI/Diver-Retriever-1.7B | **27.3** |
| DIVER-Retriever-0.6B | 0.6B | 32K | [🤗 HuggingFace]https://huggingface.co/AQ-MedAI/Diver-Retriever-0.6B <br>[🤖 ModelScope]https://www.modelscope.cn/models/AQ-MedAI/Diver-Retriever-0.6B | **25.2** |
## Evaluation
<table>
<thead>
<tr>
<th>Method</th>
<th style="text-align:right">Avg.</th>
<th style="text-align:right">Bio.</th>
<th style="text-align:right">Earth.</th>
<th style="text-align:right">Econ.</th>
<th style="text-align:right">Psy.</th>
<th style="text-align:right">Rob.</th>
<th style="text-align:right">Stack.</th>
<th style="text-align:right">Sus.</th>
<th style="text-align:right">Leet.</th>
<th style="text-align:right">Pony</th>
<th style="text-align:right">AoPS</th>
<th style="text-align:right">TheoQ.</th>
<th style="text-align:right">TheoT.</th>
</tr>
</thead>
<tbody>
<tr>
<td colspan=12 style="text-align:center"><strong>Evaluate Retriever with Original Query</strong></td>
</tr>
<tr>
<td>BM25</td>
<td style="text-align:right">14.5</td>
<td style="text-align:right">18.9</td>
<td style="text-align:right">27.2</td>
<td style="text-align:right">14.9</td>
<td style="text-align:right">12.5</td>
<td style="text-align:right">13.6</td>
<td style="text-align:right">18.4</td>
<td style="text-align:right">15.0</td>
<td style="text-align:right">24.4</td>
<td style="text-align:right">7.9</td>
<td style="text-align:right">6.2</td>
<td style="text-align:right">10.4</td>
<td style="text-align:right">4.9</td>
</tr>
<tr>
<td>SBERT</td>
<td style="text-align:right">14.9</td>
<td style="text-align:right">15.1</td>
<td style="text-align:right">20.4</td>
<td style="text-align:right">16.6</td>
<td style="text-align:right">22.7</td>
<td style="text-align:right">8.2</td>
<td style="text-align:right">11.0</td>
<td style="text-align:right">15.3</td>
<td style="text-align:right">26.4</td>
<td style="text-align:right">7.0</td>
<td style="text-align:right">5.3</td>
<td style="text-align:right">20.0</td>
<td style="text-align:right">10.8</td>
</tr>
<tr>
<td>gte-Qwen1.5-7B</td>
<td style="text-align:right">22.5</td>
<td style="text-align:right">30.6</td>
<td style="text-align:right">36.4</td>
<td style="text-align:right">17.8</td>
<td style="text-align:right">24.6</td>
<td style="text-align:right">13.2</td>
<td style="text-align:right">22.2</td>
<td style="text-align:right">14.8</td>
<td style="text-align:right">25.5</td>
<td style="text-align:right">9.9</td>
<td style="text-align:right">14.4</td>
<td style="text-align:right">27.8</td>
<td style="text-align:right">32.9</td>
</tr>
<tr>
<td>Qwen3-4B</td>
<td style="text-align:right">5.6</td>
<td style="text-align:right">3.5</td>
<td style="text-align:right">8.0</td>
<td style="text-align:right">2.3</td>
<td style="text-align:right">2.0</td>
<td style="text-align:right">1.6</td>
<td style="text-align:right">1.0</td>
<td style="text-align:right">4.4</td>
<td style="text-align:right">2.1</td>
<td style="text-align:right">0.1</td>
<td style="text-align:right">4.9</td>
<td style="text-align:right">18.0</td>
<td style="text-align:right">19.2</td>
</tr>
<tr>
<td>OpenAI</td>
<td style="text-align:right">17.9</td>
<td style="text-align:right">23.3</td>
<td style="text-align:right">26.7</td>
<td style="text-align:right">19.5</td>
<td style="text-align:right">27.6</td>
<td style="text-align:right">12.8</td>
<td style="text-align:right">14.3</td>
<td style="text-align:right">20.5</td>
<td style="text-align:right">23.6</td>
<td style="text-align:right">2.4</td>
<td style="text-align:right">8.5</td>
<td style="text-align:right">23.5</td>
<td style="text-align:right">11.7</td>
</tr>
<tr>
<td>Google</td>
<td style="text-align:right">20.0</td>
<td style="text-align:right">22.7</td>
<td style="text-align:right">34.8</td>
<td style="text-align:right">19.6</td>
<td style="text-align:right">27.8</td>
<td style="text-align:right">15.7</td>
<td style="text-align:right">20.1</td>
<td style="text-align:right">17.1</td>
<td style="text-align:right">29.6</td>
<td style="text-align:right">3.6</td>
<td style="text-align:right">9.3</td>
<td style="text-align:right">23.8</td>
<td style="text-align:right">15.9</td>
</tr>
<tr>
<td>ReasonIR-8B</td>
<td style="text-align:right">24.4</td>
<td style="text-align:right">26.2</td>
<td style="text-align:right">31.4</td>
<td style="text-align:right">23.3</td>
<td style="text-align:right">30.0</td>
<td style="text-align:right">18.0</td>
<td style="text-align:right"><strong>23.9</strong></td>
<td style="text-align:right">20.5</td>
<td style="text-align:right">35.0</td>
<td style="text-align:right">10.5</td>
<td style="text-align:right"><strong>14.7</strong></td>
<td style="text-align:right">31.9</td>
<td style="text-align:right">27.2</td>
</tr>
<tr>
<td>RaDeR-7B</td>
<td style="text-align:right">25.5</td>
<td style="text-align:right">34.6</td>
<td style="text-align:right">38.9</td>
<td style="text-align:right">22.1</td>
<td style="text-align:right">33.0</td>
<td style="text-align:right">14.8</td>
<td style="text-align:right">22.5</td>
<td style="text-align:right">23.7</td>
<td style="text-align:right">37.3</td>
<td style="text-align:right">5.0</td>
<td style="text-align:right">10.2</td>
<td style="text-align:right">28.4</td>
<td style="text-align:right">35.1</td>
</tr>
<tr>
<td>Seed1.5-Embedding</td>
<td style="text-align:right">27.2</td>
<td style="text-align:right">34.8</td>
<td style="text-align:right"><strong>46.9</strong></td>
<td style="text-align:right"><strong>23.4</strong></td>
<td style="text-align:right">31.6</td>
<td style="text-align:right">19.1</td>
<td style="text-align:right">25.4</td>
<td style="text-align:right">21.0</td>
<td style="text-align:right"><strong>43.2</strong></td>
<td style="text-align:right">4.9</td>
<td style="text-align:right">12.2</td>
<td style="text-align:right">33.3</td>
<td style="text-align:right">30.5</td>
</tr>
<tr>
<td>DIVER-Retriever</td>
<td style="text-align:right"><strong>28.9</strong></td>
<td style="text-align:right"><strong>41.8</strong></td>
<td style="text-align:right">43.7</td>
<td style="text-align:right">21.7</td>
<td style="text-align:right"><strong>35.3</strong></td>
<td style="text-align:right"><strong>21.0</strong></td>
<td style="text-align:right">21.2</td>
<td style="text-align:right"><strong>25.1</strong></td>
<td style="text-align:right">37.6</td>
<td style="text-align:right"><strong>13.2</strong></td>
<td style="text-align:right">10.7</td>
<td style="text-align:right"><strong>38.4</strong></td>
<td style="text-align:right"><strong>37.3</strong></td>
</tr>
<tr>
<td colspan=12 style="text-align:center"><strong>Evaluate Retriever with GPT-4 REASON-query</strong></td>
</tr>
<tr>
<td>BM25</td>
<td style="text-align:right">27.0</td>
<td style="text-align:right"><strong>53.6</strong></td>
<td style="text-align:right"><strong>54.1</strong></td>
<td style="text-align:right">24.3</td>
<td style="text-align:right">38.7</td>
<td style="text-align:right">18.9</td>
<td style="text-align:right">27.7</td>
<td style="text-align:right">26.3</td>
<td style="text-align:right">19.3</td>
<td style="text-align:right">17.6</td>
<td style="text-align:right">3.9</td>
<td style="text-align:right">19.2</td>
<td style="text-align:right">20.8</td>
</tr>
<tr>
<td>SBERT</td>
<td style="text-align:right">17.8</td>
<td style="text-align:right">18.5</td>
<td style="text-align:right">26.3</td>
<td style="text-align:right">17.5</td>
<td style="text-align:right">27.2</td>
<td style="text-align:right">8.8</td>
<td style="text-align:right">11.8</td>
<td style="text-align:right">17.5</td>
<td style="text-align:right">24.3</td>
<td style="text-align:right">10.3</td>
<td style="text-align:right">5.0</td>
<td style="text-align:right">22.3</td>
<td style="text-align:right">23.5</td>
</tr>
<tr>
<td>gte-Qwen1.5-7B</td>
<td style="text-align:right">24.8</td>
<td style="text-align:right">35.5</td>
<td style="text-align:right">43.1</td>
<td style="text-align:right">24.3</td>
<td style="text-align:right">34.3</td>
<td style="text-align:right">15.4</td>
<td style="text-align:right">22.9</td>
<td style="text-align:right">23.9</td>
<td style="text-align:right">25.4</td>
<td style="text-align:right">5.2</td>
<td style="text-align:right">4.6</td>
<td style="text-align:right">28.7</td>
<td style="text-align:right">34.6</td>
</tr>
<tr>
<td>Qwen3-4B</td>
<td style="text-align:right">5.5</td>
<td style="text-align:right">1.3</td>
<td style="text-align:right">17.3</td>
<td style="text-align:right">2.5</td>
<td style="text-align:right">6.2</td>
<td style="text-align:right">1.0</td>
<td style="text-align:right">4.8</td>
<td style="text-align:right">4.5</td>
<td style="text-align:right">3.0</td>
<td style="text-align:right">5.9</td>
<td style="text-align:right">0.0</td>
<td style="text-align:right">7.2</td>
<td style="text-align:right">12.5</td>
</tr>
<tr>
<td>OpenAI</td>
<td style="text-align:right">23.3</td>
<td style="text-align:right">35.2</td>
<td style="text-align:right">40.1</td>
<td style="text-align:right">25.1</td>
<td style="text-align:right">38.0</td>
<td style="text-align:right">13.6</td>
<td style="text-align:right">18.2</td>
<td style="text-align:right">24.2</td>
<td style="text-align:right">24.5</td>
<td style="text-align:right">6.5</td>
<td style="text-align:right">7.7</td>
<td style="text-align:right">22.9</td>
<td style="text-align:right">23.8</td>
</tr>
<tr>
<td>Google</td>
<td style="text-align:right">26.2</td>
<td style="text-align:right">36.4</td>
<td style="text-align:right">45.6</td>
<td style="text-align:right">25.6</td>
<td style="text-align:right">38.2</td>
<td style="text-align:right">18.7</td>
<td style="text-align:right"><strong>29.5</strong></td>
<td style="text-align:right">17.9</td>
<td style="text-align:right">31.1</td>
<td style="text-align:right">3.7</td>
<td style="text-align:right">10.0</td>
<td style="text-align:right">27.8</td>
<td style="text-align:right">30.4</td>
</tr>
<tr>
<td>ReasonIR-8B</td>
<td style="text-align:right">29.9</td>
<td style="text-align:right">43.6</td>
<td style="text-align:right">42.9</td>
<td style="text-align:right"><strong>32.7</strong></td>
<td style="text-align:right">38.8</td>
<td style="text-align:right">20.9</td>
<td style="text-align:right">25.8</td>
<td style="text-align:right"><strong>27.5</strong></td>
<td style="text-align:right">31.5</td>
<td style="text-align:right"><strong>19.6</strong></td>
<td style="text-align:right">7.4</td>
<td style="text-align:right">33.1</td>
<td style="text-align:right">35.7</td>
</tr>
<tr>
<td>RaDeR-7B</td>
<td style="text-align:right">29.2</td>
<td style="text-align:right">36.1</td>
<td style="text-align:right">42.9</td>
<td style="text-align:right">25.2</td>
<td style="text-align:right">37.9</td>
<td style="text-align:right">16.6</td>
<td style="text-align:right">27.4</td>
<td style="text-align:right">25.0</td>
<td style="text-align:right"><strong>34.8</strong></td>
<td style="text-align:right">11.9</td>
<td style="text-align:right"><strong>12.0</strong></td>
<td style="text-align:right">37.7</td>
<td style="text-align:right"><strong>43.4</strong></td>
</tr>
<tr>
<td>DIVER-Retriever</td>
<td style="text-align:right"><strong>32.1</strong></td>
<td style="text-align:right">51.9</td>
<td style="text-align:right">53.5</td>
<td style="text-align:right">29.5</td>
<td style="text-align:right"><strong>41.2</strong></td>
<td style="text-align:right"><strong>21.4</strong></td>
<td style="text-align:right">27.5</td>
<td style="text-align:right">26.1</td>
<td style="text-align:right">33.5</td>
<td style="text-align:right">11.7</td>
<td style="text-align:right">9.5</td>
<td style="text-align:right"><strong>39.3</strong></td>
<td style="text-align:right">39.7</td>
</tr>
<tr>
<td colspan=12 style="text-align:center"><strong>Evaluate retriever with DIVER-QExpand query</strong></td>
</tr>
<tr>
<td>ReasonIR-8B</td>
<td style="text-align:right">32.6</td>
<td style="text-align:right">49.4</td>
<td style="text-align:right">44.7</td>
<td style="text-align:right">32.4</td>
<td style="text-align:right">44.0</td>
<td style="text-align:right">26.6</td>
<td style="text-align:right">31.8</td>
<td style="text-align:right">29.0</td>
<td style="text-align:right">32.3</td>
<td style="text-align:right">12.8</td>
<td style="text-align:right">9.1</td>
<td style="text-align:right"><strong>40.7</strong></td>
<td style="text-align:right">38.4</td>
</tr>
<tr>
<td>+BM25 (Hybrid)</td>
<td style="text-align:right">35.7</td>
<td style="text-align:right">56.8</td>
<td style="text-align:right">53.5</td>
<td style="text-align:right"><strong>33.0</strong></td>
<td style="text-align:right"><strong>48.5</strong></td>
<td style="text-align:right"><strong>29.4</strong></td>
<td style="text-align:right"><strong>34.2</strong></td>
<td style="text-align:right"><strong>32.0</strong></td>
<td style="text-align:right"><strong>35.2</strong></td>
<td style="text-align:right">16.8</td>
<td style="text-align:right">12.9</td>
<td style="text-align:right">39.3</td>
<td style="text-align:right">36.8</td>
</tr>
<tr>
<td>DIVER-Retriever</td>
<td style="text-align:right"><strong>33.9</strong></td>
<td style="text-align:right">54.5</td>
<td style="text-align:right">52.7</td>
<td style="text-align:right">28.8</td>
<td style="text-align:right">44.9</td>
<td style="text-align:right">25.1</td>
<td style="text-align:right">27.4</td>
<td style="text-align:right">29.5</td>
<td style="text-align:right">34.5</td>
<td style="text-align:right">10.0</td>
<td style="text-align:right">14.5</td>
<td style="text-align:right"><strong>40.7</strong></td>
<td style="text-align:right">44.7</td>
</tr>
<tr>
<td>+BM25 (Hybrid)</td>
<td style="text-align:right"><strong>37.2</strong></td>
<td style="text-align:right"><strong>60.0</strong></td>
<td style="text-align:right"><strong>55.9</strong></td>
<td style="text-align:right">31.8</td>
<td style="text-align:right">47.9</td>
<td style="text-align:right">27.1</td>
<td style="text-align:right">33.9</td>
<td style="text-align:right">31.9</td>
<td style="text-align:right">35.1</td>
<td style="text-align:right"><strong>23.1</strong></td>
<td style="text-align:right"><strong>16.8</strong></td>
<td style="text-align:right">36.9</td>
<td style="text-align:right"><strong>46.6</strong></td>
</tr>
</tbody>
</table>
## Usage
<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
### Inference
#### Sentence Transformers Usage
```bash
# Requires transformers>=4.51.0
# Requires sentence-transformers>=2.7.0
from sentence_transformers import SentenceTransformer
# Load the model
model = SentenceTransformer("AQ-MedAI/Diver-Retriever-4B")
# The queries and documents to embed
queries = [
"What is the capital of China?",
"Explain gravity",
]
documents = [
"The capital of China is Beijing.",
"Gravity is a force that attracts two bodies towards each other. It gives weight to physical objects and is responsible for the movement of planets around the sun.",
]
# Encode the queries and documents. Note that queries benefit from using a prompt
# Here we use the prompt called "query" stored under `model.prompts`, but you can
# also pass your own prompt via the `prompt` argument
query_embeddings = model.encode(queries, prompt_name="query")
document_embeddings = model.encode(documents)
# Compute the (cosine) similarity between the query and document embeddings
similarity = model.similarity(query_embeddings, document_embeddings)
print(similarity)
```
#### Transformers Usage
```bash
# Requires transformers>=4.51.0
import torch
import torch.nn.functional as F
from torch import Tensor
from transformers import AutoTokenizer, AutoModel
def last_token_pool(last_hidden_states: Tensor,
attention_mask: Tensor) -> Tensor:
left_padding = (attention_mask[:, -1].sum() == attention_mask.shape[0])
if left_padding:
return last_hidden_states[:, -1]
else:
sequence_lengths = attention_mask.sum(dim=1) - 1
batch_size = last_hidden_states.shape[0]
return last_hidden_states[torch.arange(batch_size, device=last_hidden_states.device), sequence_lengths]
def get_detailed_instruct(task_description: str, query: str) -> str:
return f'Instruct: {task_description}\nQuery:{query}'
# Each query must come with a one-sentence instruction that describes the task
task = 'Given a web search query, retrieve relevant passages that answer the query'
queries = [
get_detailed_instruct(task, 'What is the capital of China?'),
get_detailed_instruct(task, 'Explain gravity')
]
# No need to add instructions for retrieval documents
documents = [
"The capital of China is Beijing.",
"Gravity is a force that attracts two bodies towards each other. It gives weight to physical objects and is responsible for the movement of planets around the sun."
]
input_texts = queries + documents
tokenizer = AutoTokenizer.from_pretrained('AQ-MedAI/Diver-Retriever-4B', padding_side='left')
model = AutoModel.from_pretrained('AQ-MedAI/Diver-Retriever-4B')
max_length = 8192
# Tokenize the input texts
batch_dict = tokenizer(
input_texts,
padding=True,
truncation=True,
max_length=max_length,
return_tensors="pt",
)
batch_dict.to(model.device)
outputs = model(**batch_dict)
embeddings = last_token_pool(outputs.last_hidden_state, batch_dict['attention_mask'])
# normalize embeddings
embeddings = F.normalize(embeddings, p=2, dim=1)
scores = (embeddings[:2] @ embeddings[2:].T)
print(scores.tolist())
# [[0.9319270849227905, 0.5878604054450989], [0.639923095703125, 0.7950234413146973]]
```
### Finetuning
We recommend you to use [swift](https://github.com/modelscope/ms-swift) to finetune our DIVER-Retriever-4B with infonce.
Before starting training, please ensure your environment is properly configured.
```bash
pip install ms-swift -U
# Install from source
pip install git+https://github.com/modelscope/ms-swift.git
pip install transformers -U
# Optional packages
pip install deepspeed # multi-GPU training
pip install liger-kernel # save GPU memory resources
pip install flash-attn --no-build-isolation
```
#### Training Command
Using the infonce loss as an example, the complete training command is as follows:
```bash
nproc_per_node=8
NPROC_PER_NODE=$nproc_per_node \
swift sft \
--model DIVER/DIVER-Retriever-4B \
--task_type embedding \
--model_type qwen3_emb \
--train_type full \
--dataset your_dataset \
--split_dataset_ratio 0.05 \
--eval_strategy steps \
--output_dir output \
--eval_steps 20 \
--num_train_epochs 5 \
--save_steps 20 \
--per_device_train_batch_size 4 \
--per_device_eval_batch_size 4 \
--gradient_accumulation_steps 4 \
--learning_rate 6e-6 \
--loss_type infonce \
--label_names labels \
--dataloader_drop_last true \
--deepspeed zero3
```
## Citation
<!-- If a paper or blog post is introducing the model, the APA and BibTeX information for that should go in this section. -->
If you find our work helpful, feel free to cite it.
```
@misc{long2025divermultistageapproachreasoningintensive,
title={DIVER: A Multi-Stage Approach for Reasoning-intensive Information Retrieval},
author={Meixiu Long and Duolin Sun and Dan Yang and Junjie Wang and Yue Shen and Jian Wang and Peng Wei and Jinjie Gu and Jiahai Wang},
year={2025},
eprint={2508.07995},
archivePrefix={arXiv},
primaryClass={cs.IR},
url={https://arxiv.org/abs/2508.07995},
}
```