Instructions to use UCSC-VLAA/VLM-CapCurriculum-InternVL3.5-8B-Staged with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use UCSC-VLAA/VLM-CapCurriculum-InternVL3.5-8B-Staged with Transformers:

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

pipe = pipeline("image-text-to-text", model="UCSC-VLAA/VLM-CapCurriculum-InternVL3.5-8B-Staged")
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, AutoModelForImageTextToText

processor = AutoProcessor.from_pretrained("UCSC-VLAA/VLM-CapCurriculum-InternVL3.5-8B-Staged")
model = AutoModelForImageTextToText.from_pretrained("UCSC-VLAA/VLM-CapCurriculum-InternVL3.5-8B-Staged")
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

vLLM

How to use UCSC-VLAA/VLM-CapCurriculum-InternVL3.5-8B-Staged with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "UCSC-VLAA/VLM-CapCurriculum-InternVL3.5-8B-Staged"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "UCSC-VLAA/VLM-CapCurriculum-InternVL3.5-8B-Staged",
		"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/UCSC-VLAA/VLM-CapCurriculum-InternVL3.5-8B-Staged

SGLang

How to use UCSC-VLAA/VLM-CapCurriculum-InternVL3.5-8B-Staged 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 "UCSC-VLAA/VLM-CapCurriculum-InternVL3.5-8B-Staged" \
    --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": "UCSC-VLAA/VLM-CapCurriculum-InternVL3.5-8B-Staged",
		"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 "UCSC-VLAA/VLM-CapCurriculum-InternVL3.5-8B-Staged" \
        --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": "UCSC-VLAA/VLM-CapCurriculum-InternVL3.5-8B-Staged",
		"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 UCSC-VLAA/VLM-CapCurriculum-InternVL3.5-8B-Staged with Docker Model Runner:
```
docker model run hf.co/UCSC-VLAA/VLM-CapCurriculum-InternVL3.5-8B-Staged
```

VLM-CapCurriculum-InternVL3.5-8B-Staged / README.md

Chtholly17

Update README: add author list, real GitHub/Collection/project-page URLs

acf8f8f verified about 12 hours ago

preview code

raw

history blame contribute delete

4.4 kB

	---
	license: apache-2.0
	language:
	- en
	library_name: transformers
	pipeline_tag: image-text-to-text
	base_model: OpenGVLab/InternVL3_5-8B-HF
	tags:
	- vision-language-model
	- vlm
	- reasoning
	- perception
	- rlvr
	- grpo
	- icml-2026
	---

	# VLM-CapCurriculum-InternVL3.5-8B-Staged

	A vision-language model post-trained from OpenGVLab/InternVL3_5-8B-HF with the staged, capability-dimension curriculum from
	"From Seeing to Thinking: Decoupling Perception and Reasoning Improves Post-Training of Vision-Language Models"
	(ICML 2026).

	This release is the Stage-3 step-279 checkpoint.

	> TL;DR. Visual perception — not reasoning length — is the dominant bottleneck for visual reasoning in VLMs. We fix this by post-training along a capability axis (perception → textual reasoning → visual reasoning) rather than mixing all data together.

	\| Resource \| Link \|
	\|---\|---\|
	\| 📄 Paper \| <TODO_PAPER_URL> \|
	\| 💻 Code \| https://github.com/UCSC-VLAA/VLM-CapCurriculum \|
	\| 🌐 Project page \| https://ucsc-vlaa.github.io/VLM-CapCurriculum \|
	\| 🤗 Collection (model + data + eval) \| https://huggingface.co/collections/UCSC-VLAA/vlm-capcurriculum-from-seeing-to-thinking-icml-2026-6a07691f944148ccb2b183b8 \|

	## Headline numbers (extended benchmark suite, AVG over 10 benchmarks)

	\| Setting \| Extended AVG \|
	\|---\|:---:\|
	\| InternVL3.5-8B (base) \| 37.33 \|
	\| InternVL3.5-8B + Merged training \| 52.76 \|
	\| InternVL3.5-8B + Staged (this model) \| 53.71 \|

	See Appendix Table 9 of the paper for the full per-benchmark breakdown.

	## How it was trained

	Three RLVR stages with GRPO (on top of [EasyR1](https://github.com/hiyouga/EasyR1)):

	1. Stage 1 — visual perception on `UCSC-VLAA/VLM-CapCurriculum-Perception` (synthesised + filtered DOCCI MCQs).
	2. Stage 2 — textual reasoning on `UCSC-VLAA/VLM-CapCurriculum-TextReasoning` (ORZ-Math-13k).
	3. Stage 3 — visual reasoning on `UCSC-VLAA/VLM-CapCurriculum-VisualReasoning` (CLEVR-Math + GeoQA170K + Math PUMA + DocVQA + ArxivQA mix). Released checkpoint: step 279 of Stage 3.

	All three stages share one system / format prompt — see [Inference](#inference) below.

	Detailed launch scripts: [`training/examples/internvl3_5_8b/`](https://github.com/UCSC-VLAA/VLM-CapCurriculum/tree/main/training/examples/internvl3_5_8b) in the code repo.

	## Inference

	The model expects the unified system prompt that it was trained against:

	```
	You FIRST think about the reasoning process as an internal monologue and then
	provide the final answer. The reasoning process MUST BE enclosed within
	<think> </think> tags. The final answer MUST BE put in \boxed{}.
	i.e. <think> reasoning here </think> \boxed{final answer here}
	```

	Quick start with LMDeploy:

	```bash
	lmdeploy serve api_server UCSC-VLAA/VLM-CapCurriculum-InternVL3.5-8B-Staged \
	--server-port 23343 --tp 4
	```

	For VLMEvalKit-style benchmark eval, plug it in via the `InternVL3_5_8B_Staged` alias defined in [`evaluation/configs/models.py`](https://github.com/UCSC-VLAA/VLM-CapCurriculum/blob/main/evaluation/configs/models.py).

	## Intended use & limitations

	Intended for research on vision-language reasoning, post-training methodology, and capability-dimension curriculum learning. Inherits the safety / bias profile of the underlying InternVL3.5-8B backbone; we have not added additional alignment fine-tuning. Not recommended for high-stakes deployments without further evaluation.

	Trained at the 8B parameter scale with 4096-token max prompt length and a fixed group size of 5. Behaviour at much longer contexts or substantially different prompt formats has not been characterised.

	## License & citation

	Released under Apache-2.0, matching the upstream backbone. If you use this model, please cite:

	```bibtex
	@inproceedings{vlmcapcurriculum2026,
	title = {From Seeing to Thinking: Decoupling Perception and Reasoning Improves Post-Training of Vision-Language Models},
	author = {Juncheng Wu and Hardy Chen and Haoqin Tu and Xianfeng Tang and Freda Shi and Hui Liu and Hanqing Lu and Cihang Xie and Yuyin Zhou},
	booktitle = {Proceedings of the International Conference on Machine Learning (ICML)},
	year = {2026}
	}
	```

	## Acknowledgements

	Built on top of [EasyR1](https://github.com/hiyouga/EasyR1), [VLMEvalKit](https://github.com/open-compass/VLMEvalKit), and the [InternVL](https://huggingface.co/OpenGVLab) family.