Vision Opd Qwen3.5 4b Reproduction
This repository contains an unofficial reproduction of Vision-OPD based on Qwen/Qwen3.5-4B. It is not an official model release from the Vision-OPD authors.
Modellquelle
- Art
- Text model
- Dateigröße
- 9,64 GB
- Basismodell
- Qwen/Qwen3.5-4B
- Version
- vBF16
- Aktualisiert
- 30. Okt. 2026
- Downloads
- 14
- Quelle
- Hugging Face
Quellenbeschreibung
Vision-OPD Qwen3.5-4B Reproduction
This repository contains an unofficial reproduction of Vision-OPD based on
Qwen/Qwen3.5-4B. It is not an official model release from the Vision-OPD
authors.
Vision-OPD applies regional-to-global on-policy self-distillation to improve
fine-grained visual perception. This checkpoint was produced with the public
Vision-OPD implementation and the
public yuanqianhao/Vision-OPD-6K
training set.
Training
| Setting | Value |
|---|---|
| Base model | Qwen/Qwen3.5-4B |
| Training data | yuanqianhao/Vision-OPD-6K (6,241 samples) |
| Epochs | 1 |
| Final step | 65 |
| Batch size | 96 |
| Rollouts per prompt | 8 |
| Learning rate | 2e-6 |
| Hardware | 8 x NVIDIA B200 |
The uploaded artifact is the merged Transformers checkpoint, not an FSDP training shard.
Evaluation
Generation used non-thinking mode with a maximum output length of 32,768
tokens. The answer judge was openai/gpt-oss-120b, served locally with a
65,536-token context window and a 2,048-token judge output limit.
| Benchmark | Qwen3.5-4B baseline | This reproduction | Paper Vision-OPD-4B |
|---|---|---|---|
| V* | 84.29 | 90.05 | 92.15 |
| ZoomBench | 47.69 | 59.64 | 59.76 |
| HRBench-4K | 84.38 | 81.75 | 84.50 |
| HRBench-8K | 80.13 | 80.00 | 80.38 |
| MME-RealWorld-EN | 63.86 | 71.96 | 74.88 |
| MME-RealWorld-CN | 63.70 | 69.56 | 70.76 |
| Average | 70.68 | 75.49 | 77.07 |
The reproduction improves the six-benchmark average by 4.81 points over the reported Qwen3.5-4B baseline and remains 1.58 points below the paper result.
Usage
Use a recent Transformers release with Qwen3.5 support (transformers>=5.5.0).
from transformers import AutoModelForMultimodalLM, AutoProcessor
model_id = "Hugo0713/vision-opd-qwen3.5-4b-reproduction"
processor = AutoProcessor.from_pretrained(model_id)
model = AutoModelForMultimodalLM.from_pretrained(
model_id,
torch_dtype="auto",
device_map="auto",
)
The model can also be served with vLLM:
vllm serve Hugo0713/vision-opd-qwen3.5-4b-reproduction \
--served-model-name Vision-OPD-4B \
--gpu-memory-utilization 0.85
Limitations
This is a single reproduction run. The reported results depend on the local inference and judge configuration described above and should not be treated as an official Vision-OPD release.
Citation
@article{yuan2026vision,
title={Vision-OPD: Learning to See Fine Details for Multimodal LLMs via On-Policy Self-Distillation},
author={Yuan, Qianhao and Lou, Jie and Yu, Xing and Lin, Hongyu and Sun, Le and Han, Xianpei and Lu, Yaojie},
journal={arXiv preprint arXiv:2605.18740},
year={2026}
}
License
Apache-2.0. See LICENSE for details.
Quellen
1 Quelle
Quellen
1 QuelleVerifiziert 30. Okt.
Download files
Repository files
File inventory only; completeness is unverified.
Modellartefakte
1 Artefakt
Modellartefakte
1 Artefaktmodel.safetensors
safetensors · BF16 (tensor header) · 9,64 GB · SHA-256 f4cca3fc8860…492e · Hugging Face
Download fileQuellenauszüge
3 Auszüge
Quellenauszüge
3 Auszüge--- license: apache-2.0 library_name: transformers pipeline_tag: image-text-to-text base_model: Qwen/Qwen3.5-4B datasets: - yuanqianhao/Vision-OPD-6K tags: - vision-language - multimodal - vision-opd - reproduction --- # Vision-OPD Qwen3.5-4B Reproduction This repository contains an **unofficial reproduction** of Vision-OPD based on `Qwen/Qwen3.5-4B`. It is not an official model release from the Vision-OPD authors. Vision-OPD applies regional-to-global on-policy self-distillation to improve fine-grained visual perception. This checkpoint was produced with the public [Vision-OPD implementation](https://github.com/VisionOPD/Vision-OPD) and the public [`yuanqianhao/Vision-OPD-6K`](https://huggingface.co/datasets/yuanqianhao/Vision-OPD-6K) training set. ## Training | Setting | Value | | --- | --- | | Base model | `Qwen/Qwen3.5-4B` | | Training data | `yuanqianhao/Vision-OPD-6K` (6,241 samples) | | Epochs | 1 | | Final step | 65 | | Batch size | 96 | | Rollouts per prompt | 8 | | Learning rate | 2e-6 | | Hardware | 8 x NVIDIA B200 | The uploaded artifact is the merged Transformers checkpoint, not an FSDP training shard. ## Evaluation Generation used non-thinking mode with a maximum output length of 32,768 tokens. The answer judge was `openai/gpt-oss-120b`, served locally with a 65,536-token context window and a 2,048-token judge output limit. | Benchmark | Qwen3.5-4B baseline | This reproduction | Paper Vision-OPD-4B | | --- | ---: | ---: | ---: | | V* | 84.29 | 90.05 | 92.15 | | ZoomBench | 47.69 | 59.64 | 59.76 | | HRBench-4K | 84.38 | 81.75 | 84.50 | | HRBench-8K | 80.13 | 80.00 | 80.38 | | MME-RealWorld-EN | 63.86 | 71.96 | 74.88 | | MME-RealWorld-CN | 63.70 | 69.56 | 70.76 | | Average | 70.68 | 75.49 | 77.07 | The reproduction improves the six-benchmark average by 4.81 points over the reported Qwen3.5-4B baseline and remains 1.58 points below the paper result. ## Usage Use a recent Transformers release with Qwen3.5 support (`transformers>=5.5.0`). ```python from transformers import AutoModelForMultimodalLM, AutoProcessor model_id = "Hugo0713/vision-opd-qwen3.5-4b-reproduction" processor = AutoProcessor.from_pretrained(model_id) model = AutoModelForMultimodalLM.from_pretrained( model_id, torch_dtype="auto", device_map="auto", ) ``` The model can also be served with vLLM: ```bash vllm serve Hugo0713/vision-opd-qwen3.5-4b-reproduction...
Source context: 14 downloads · 0 likes · Pipeline image-text-to-text · Library transformers · Repo Hugo0713/vision-opd-qwen3.5-4b-reproduction
Source context: 32 downloads · 0 likes · Pipeline image-text-to-text · Library transformers · Repo Hugo0713/vision-opd-qwen3.5-4b-reproduction
