Instructions to use zhenliuu/K12-Reasoner-4B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use zhenliuu/K12-Reasoner-4B with Transformers:

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

pipe = pipeline("image-text-to-text", model="zhenliuu/K12-Reasoner-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)

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

processor = AutoProcessor.from_pretrained("zhenliuu/K12-Reasoner-4B")
model = AutoModelForMultimodalLM.from_pretrained("zhenliuu/K12-Reasoner-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=256)
print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use zhenliuu/K12-Reasoner-4B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "zhenliuu/K12-Reasoner-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": "zhenliuu/K12-Reasoner-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/zhenliuu/K12-Reasoner-4B

SGLang

How to use zhenliuu/K12-Reasoner-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 "zhenliuu/K12-Reasoner-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": "zhenliuu/K12-Reasoner-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 "zhenliuu/K12-Reasoner-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": "zhenliuu/K12-Reasoner-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 zhenliuu/K12-Reasoner-4B with Docker Model Runner:
```
docker model run hf.co/zhenliuu/K12-Reasoner-4B
```

K12-Reasoner-4B

K12-Reasoner-4B 是基于 Qwen3.5-4B 进行 DAPO 后训练的多模态模型，使用本项目整理的数学与物理训练集，面向图文题目理解、学科推理与答案生成。

项目主页 · 训练数据 · 训练代码

模型概况

项目	配置
基础模型	Qwen3.5-4B
参数规模	4B 级
输入 / 输出	文本与图片 / 文本
后训练	DAPO，基于 TRL
训练数据	数学与物理训练集，1,160 题
权重格式	BF16，Safetensors

快速使用

安装支持本模型的 Transformers 与 PyTorch 环境，下载模型后使用随附的infer.py：

pip install "torch==2.10.0" "transformers==5.15.0" "accelerate==1.13.0" "Pillow==12.2.0"
hf download zhenliuu/K12-Reasoner-4B --local-dir K12-Reasoner-4B
python K12-Reasoner-4B/infer.py --model ./K12-Reasoner-4B \
  --subject 数学 --question "求方程 2x + 3 = 11 的解。"

图片题通过--image传入本地图片，题干与选项通过--question传入；重复--image可提供多张图片：

python K12-Reasoner-4B/infer.py --model ./K12-Reasoner-4B \
  --subject 物理 --image question.png --question "根据图示，判断物体所受合力的方向。"

默认使用单张CUDA GPU、BF16与贪心解码，关闭thinking，最大输出8,192 tokens。脚本支持--device cpu以及本地模型路径。PyTorch的CUDA构建需与运行环境匹配。模型使用基础架构的processor与chat template，输出包含简洁解题过程和末行“最终答案：”。

评测结果

我们在以下多模态数学与物理基准上，将 K12-Reasoner-4B 与基础模型 Qwen3.5-4B 进行对比。指标为准确率，提升以百分点（pp）表示。

基准	可计分题数 / 测试题数	Qwen3.5-4B	K12-Reasoner-4B	提升（pp）
OlympiadBench	4,055 / 4,869	31.89%	40.12%	+8.24
MMSciBench	710 / 1,167	63.24%	69.58%	+6.34
LiveK12Bench	1,168 / 1,168	48.72%	54.71%	+5.99
MM-MATH	5,881 / 5,901	67.40%	71.96%	+4.56
PhyX	6,000 / 6,000	43.30%	47.78%	+4.48
PhysUniBench	1,247 / 3,304	38.89%	41.94%	+3.05

评测设置： 两模型使用相同输入模板、解码设置和自动判分程序；关闭 thinking，采用贪心解码，最大输出长度为 8,192 tokens。准确率按可计分题目计算：MMSciBench、PhysUniBench 使用选择题，OlympiadBench 使用非证明题。结果基于项目统一评测协议。

完整提示词和计分口径见评测说明，训练配置见项目训练文档。

使用范围

模型面向教育图文推理研究与应用开发。涉及重要教学结论时，应结合参考答案与专业知识复核模型输出。

Downloads last month: 13

Safetensors

Model size

5B params

Tensor type

BF16

Model tree for zhenliuu/K12-Reasoner-4B

Base model

Qwen/Qwen3.5-4B-Base

Finetuned

Qwen/Qwen3.5-4B

Finetuned

(895)

this model

zhenliuu
/

K12-Reasoner-4B