Instructions to use XinNUS/CycleGRPO-4B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use XinNUS/CycleGRPO-4B with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("image-text-to-text", model="XinNUS/CycleGRPO-4B")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
pipe(text=messages)

# Load model directly
from transformers import AutoProcessor, AutoModelForMultimodalLM

processor = AutoProcessor.from_pretrained("XinNUS/CycleGRPO-4B")
model = AutoModelForMultimodalLM.from_pretrained("XinNUS/CycleGRPO-4B", device_map="auto")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
inputs = processor.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use XinNUS/CycleGRPO-4B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "XinNUS/CycleGRPO-4B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "XinNUS/CycleGRPO-4B",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker

docker model run hf.co/XinNUS/CycleGRPO-4B

SGLang

How to use XinNUS/CycleGRPO-4B with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "XinNUS/CycleGRPO-4B" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "XinNUS/CycleGRPO-4B",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "XinNUS/CycleGRPO-4B" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "XinNUS/CycleGRPO-4B",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Docker Model Runner
How to use XinNUS/CycleGRPO-4B with Docker Model Runner:
```
docker model run hf.co/XinNUS/CycleGRPO-4B
```

Link model card to paper and project page

by nielsr HF Staff - opened Jul 14

base: refs/heads/main

←

from: refs/pr/1

Discussion Files changed

+20

-3

Files changed (1) hide show

README.md +20 -3

README.md CHANGED Viewed

@@ -1,13 +1,17 @@
 ---
 license: apache-2.0
 pipeline_tag: image-text-to-text
-library_name: transformers
-base_model:
-  - zhouyik/Qwen3-VL-4B-SAMTok
 ---
 # CycleGRPO-4B
 CycleGRPO-4B is post-trained from
 [zhouyik/Qwen3-VL-4B-SAMTok](https://huggingface.co/zhouyik/Qwen3-VL-4B-SAMTok)
 with **caption ↔ grounding cycle-consistent reinforcement learning**: a caption is
@@ -83,3 +87,16 @@ from the base model [Qwen3-VL-4B-SAMTok](https://huggingface.co/zhouyik/Qwen3-VL
 Released under Apache-2.0. Derived from Qwen3-VL-4B-SAMTok; use is also subject to the
 base model's license and terms.

 ---
+base_model:
+- zhouyik/Qwen3-VL-4B-SAMTok
+library_name: transformers
 license: apache-2.0
 pipeline_tag: image-text-to-text
 ---
 # CycleGRPO-4B
+This is the official checkpoint for **CycleGRPO-4B**, presented in the paper [Actor as Its Own Critic: Unifying Region Understanding and Localization via CycleGRPO](https://huggingface.co/papers/2607.11581).
+**[📜 Paper](https://huggingface.co/papers/2607.11581)** | **[💻 Project Page](https://devinxzhang.github.io/CycleGRPO-Page/)** | **[🐙 Code](https://github.com/devinxzhang/CycleGRPO)**
 CycleGRPO-4B is post-trained from
 [zhouyik/Qwen3-VL-4B-SAMTok](https://huggingface.co/zhouyik/Qwen3-VL-4B-SAMTok)
 with **caption ↔ grounding cycle-consistent reinforcement learning**: a caption is
 Released under Apache-2.0. Derived from Qwen3-VL-4B-SAMTok; use is also subject to the
 base model's license and terms.
+## Citation
+```bibtex
+@inproceedings{cyclegrpo2026,
+  title     = {Actor as Its Own Critic: Unifying Region Understanding and
+               Localization via CycleGRPO},
+  author    = {Zhang, Xin and Wang, Haochen and Zhou, Yikang and Wang, Zhuochen
+               and Li, Jason and Tan, Robby T.},
+  booktitle = {European Conference on Computer Vision (ECCV)},
+  year      = {2026}
+}
+```