Instructions to use nuetee/RA-RFT-Qwen2.5-VL-7B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use nuetee/RA-RFT-Qwen2.5-VL-7B with Transformers:

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

pipe = pipeline("image-text-to-text", model="nuetee/RA-RFT-Qwen2.5-VL-7B")
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)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoProcessor, AutoModelForMultimodalLM

processor = AutoProcessor.from_pretrained("nuetee/RA-RFT-Qwen2.5-VL-7B")
model = AutoModelForMultimodalLM.from_pretrained("nuetee/RA-RFT-Qwen2.5-VL-7B", 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=256)
print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use nuetee/RA-RFT-Qwen2.5-VL-7B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "nuetee/RA-RFT-Qwen2.5-VL-7B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "nuetee/RA-RFT-Qwen2.5-VL-7B",
		"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/nuetee/RA-RFT-Qwen2.5-VL-7B

SGLang

How to use nuetee/RA-RFT-Qwen2.5-VL-7B 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 "nuetee/RA-RFT-Qwen2.5-VL-7B" \
    --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": "nuetee/RA-RFT-Qwen2.5-VL-7B",
		"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 "nuetee/RA-RFT-Qwen2.5-VL-7B" \
        --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": "nuetee/RA-RFT-Qwen2.5-VL-7B",
		"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 nuetee/RA-RFT-Qwen2.5-VL-7B with Docker Model Runner:
```
docker model run hf.co/nuetee/RA-RFT-Qwen2.5-VL-7B
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

RA-RFT-Qwen2.5-VL-7B

RA-RFT (Refusal-Aware Reinforcement Fine-Tuning) enables video temporal grounding models to locate relevant temporal segments and refuse unanswerable queries with explanations and corrected queries.

Paper: Learning to Refuse: Refusal-Aware Reinforcement Fine-Tuning for Hard-Irrelevant Queries in Video Temporal Grounding
Implementation: JINSUBY/RA-RFT
Base model: Boshenxx/Time-R1-7B
Architecture: Qwen2.5-VL-7B
Format: full model weights in Safetensors, with tokenizer and processor files (not a LoRA adapter).

Usage

Follow the RA-RFT README for installation and dataset preparation. Download the checkpoint from the RA-RFT repository directory:

from huggingface_hub import snapshot_download

snapshot_download(
    repo_id="nuetee/RA-RFT-Qwen2.5-VL-7B",
    local_dir="./ckpts/RA-RFT-Qwen2.5-VL-7B",
)

Run inference using the implementation's task prompt and video preprocessing. For example, HI-VTG (ActivityNet):

python inference.py \
  --model_path ./ckpts/RA-RFT-Qwen2.5-VL-7B \
  --test_data_path ./dataset/anno/hi_vtg_activitynet.json \
  --output_dir ./inference_output/hi_vtg_activitynet \
  --preprocessed_data_path ./dataset/preprocessed_video \
  --batch_size 8

Adjust dataset and video paths to your environment. For evaluation, other datasets, and multi-GPU inference, see the RA-RFT evaluation instructions.

Output format

Responses follow the implementation's <think>...</think> <answer>...</answer> <correction>...</correction> format. Answerable queries receive temporal boundaries in seconds; unanswerable queries receive a refusal explanation and a corrected query.

Citation

Please cite the CVPR 2026 paper:

@InProceedings{Lee_2026_CVPR,
  author={Lee, Jin-Seop and Lee, SungJoon and Jung, SeongJun and Li, Boyang and Lee, Jee-Hyong},
  title={Learning to Refuse: Refusal-Aware Reinforcement Fine-Tuning for Hard-Irrelevant Queries in Video Temporal Grounding},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
  month={June},
  year={2026},
  pages={10397-10407}
}

License

Apache 2.0. See LICENSE.

Downloads last month: 117

Safetensors

Model size

8B params

Tensor type

BF16

Model tree for nuetee/RA-RFT-Qwen2.5-VL-7B

Base model

Boshenxx/Time-R1-7B

Finetuned

(1)

this model

Quantizations

2 models

Paper for nuetee/RA-RFT-Qwen2.5-VL-7B

Learning to Refuse: Refusal-Aware Reinforcement Fine-Tuning for Hard-Irrelevant Queries in Video Temporal Grounding

Paper • 2511.23151 • Published Nov 28, 2025