---
title: Trident-Qwen3VL-2B
canonical_url: "https://www.modelscope.cn/models/OpenBMB/Trident-Qwen3VL-2B"
md_url: "https://www.modelscope.cn/models/OpenBMB/Trident-Qwen3VL-2B.md"
repository: OpenBMB/Trident-Qwen3VL-2B
last_updated: 2026-10-15
license: mit
pipeline_tag: feature-extraction
tasks:
- feature-extraction
model_type:
- qwen3_vl
architectures:
- Qwen3VLForEmbedding
base_model:
- Qwen/Qwen3-VL-2B-Instruct
base_model_relation: finetune
parameters: 2.1B
tensor_type:
- BF16
library_name:
- safetensors
language:
- en
downloads: 42
stars: 0
tags:
- multimodal-embedding
- mixed-modality-retrieval
- multimodal-retrieval
- visual-document-retrieval
- qwen3-vl
---
# Trident-Qwen3VL-2B
> Trident-Qwen3VL-2B - OpenBMB 在 ModelScope 开源的模型。Official model for Chaos in the Text: Revealing the Modality Preference in Mixed-Modality Retrievers
OpenBMB/Trident-Qwen3VL-2B 是 ModelScope 魔搭社区上的 2.1B 参数feature-extraction模型，采用 mit 许可，基于 Qwen/Qwen3-VL-2B-Instruct 构建。
- **Repository**: OpenBMB/Trident-Qwen3VL-2B
- **License**: mit
- **Tasks**: feature-extraction
- **Parameters**: 2.1B
- **Base model**: Qwen/Qwen3-VL-2B-Instruct
- **Tags**: multimodal-embedding, mixed-modality-retrieval, multimodal-retrieval, visual-document-retrieval, qwen3-vl
- **Downloads**: 42
- **Stars**: 0
- **Last updated**: 2026-10-15
Source: https://www.modelscope.cn/models/OpenBMB/Trident-Qwen3VL-2B
---
Trident-Qwen3VL-2B
Official model for Chaos in the Text: Revealing the Modality Preference in Mixed-Modality Retrievers
**Trident-Qwen3VL-2B** is a multimodal embedding model for **mixed-modality retrieval**. It maps text, images, and fused text-image inputs into one shared space, so a text query can rank text-only, image-only, and fused documents in a single list. It is fine-tuned from [Qwen3-VL-2B-Instruct](https://huggingface.co/Qwen/Qwen3-VL-2B-Instruct) on [Trident-Train](https://huggingface.co/datasets/OpenBMB/Trident-Train), as described in the paper [Chaos in the Text: Revealing the Modality Preference in Mixed-Modality Retrievers](https://arxiv.org/abs/2610.11816). The LoRA adapter is already merged, so no base model or [PEFT](https://github.com/huggingface/peft) is needed.
## Quickstart
**1. Set up the code and download the weights**
```bash
git clone https://github.com/OpenBMB/Trident.git
cd Trident
conda create -n trident python=3.12 -y && conda activate trident
pip install -r requirements.txt # tested with transformers 4.57.1 (Qwen3-VL needs >= 4.57)
hf download OpenBMB/Trident-Qwen3VL-2B --local-dir checkpoints/Trident-Qwen3VL-2B
```
**2. Encode queries and documents** (run from the root of the Trident repository)
```python
import sys
sys.path.insert(0, "src")
from embedding_backends.common import (
ensure_tokenizers_thread_safety,
ensure_transformers_special_tokens_compat,
)
from embedding_backends.trident_qwen3vl import TridentQwen3VLBackend
ensure_transformers_special_tokens_compat()
ensure_tokenizers_thread_safety()
CKPT = "checkpoints/Trident-Qwen3VL-2B"
backend = TridentQwen3VLBackend(
{"model_type": "trident_qwen3vl", "checkpoint": CKPT, "dtype": "bfloat16"},
device="cuda:0",
)
backend.load()
def encode(items, role):
"""items: list of {"text": str | None, "images": [paths]}; role: "query" or "doc"."""
backend.args_dict["role"] = role
instruction = backend.default_instruction(role)
inputs = [{**item, "instruction": instruction} for item in items]
embeddings, _ = backend.process(inputs, role=role)
return embeddings # [N, dim] numpy array, L2-normalised
queries = [
{"text": "How many stores did Saint Laurent operate in Western Europe in 2020?", "images": []},
]
documents = [
{"text": "A bar chart showing the number of Saint Laurent stores by region and year.", "images": []}, # text-only
{"text": None, "images": ["figs/example_chart.png"]}, # image-only
{"text": "Number of stores by region.", "images": ["figs/example_chart.png"]}, # text + image
]
q = encode(queries, role="query") # [1, dim]
d = encode(documents, role="doc") # [3, dim]
scores = q @ d.T # cosine similarity, comparable across all document types
print(scores)
print(scores[0].argsort()[::-1]) # document ranking for the query
```
Each input is a dict with an optional `text` string and an `images` list of local image paths. Use `text` only, `images` only, or both for fused inputs. Queries and documents use different default instructions, which are part of the training format.
## License
The Trident-specific contributions, including the fine-tuned weights, are released under the [MIT License](https://github.com/OpenBMB/Trident/blob/main/LICENSE). The model is fine-tuned from [Qwen3-VL-2B-Instruct](https://huggingface.co/Qwen/Qwen3-VL-2B-Instruct), which is released under Apache-2.0, and the base model's license terms continue to apply to the weights derived from it. The training images come from third-party datasets (see [Trident-Train](https://huggingface.co/datasets/OpenBMB/Trident-Train)) and remain subject to their original licenses.
## Citation
If you find this work useful, please cite:
```bibtex
@article{sun2026chaos,
title={Chaos in the Text: Revealing the Modality Preference in Mixed-Modality Retrievers},
author={Sun, Yubo and Peng, Chunyi and Yan, Yukun and Liu, Zhenghao and Xu, Zhipeng and Mei, Sen and Xin, Linlin and Zeng, Zheni and Sun, Maosong},
journal={arXiv preprint arXiv:2610.11816},
year={2026}
}
```