Instructions to use DtYXs/SCAPO-Qwen3-1.7B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use DtYXs/SCAPO-Qwen3-1.7B with Transformers:

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

pipe = pipeline("text-generation", model="DtYXs/SCAPO-Qwen3-1.7B")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("DtYXs/SCAPO-Qwen3-1.7B")
model = AutoModelForCausalLM.from_pretrained("DtYXs/SCAPO-Qwen3-1.7B", device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.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(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use DtYXs/SCAPO-Qwen3-1.7B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "DtYXs/SCAPO-Qwen3-1.7B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "DtYXs/SCAPO-Qwen3-1.7B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/DtYXs/SCAPO-Qwen3-1.7B

SGLang

How to use DtYXs/SCAPO-Qwen3-1.7B 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 "DtYXs/SCAPO-Qwen3-1.7B" \
    --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": "DtYXs/SCAPO-Qwen3-1.7B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

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 "DtYXs/SCAPO-Qwen3-1.7B" \
        --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": "DtYXs/SCAPO-Qwen3-1.7B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use DtYXs/SCAPO-Qwen3-1.7B with Docker Model Runner:
```
docker model run hf.co/DtYXs/SCAPO-Qwen3-1.7B
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

SCAPO-Qwen3-1.7B

SCAPO-Qwen3-1.7B is a full-parameter RL fine-tune of Qwen3-1.7B-Base using Semifactual Credit-Augmented Policy Optimization (SCAPO) on DAPO-Math-17K. SCAPO uses answer-preserving prompt perturbations to refine token-level credit assignment during training.

Results

Full results and evaluation details are available in the paper and code repository.

Method	AIME 24–26	AMC 23–25	HMMT 25–26	ID Avg.	OOD Avg.
Base	3.61	23.33	0.69	12.45	9.31
GRPO	7.71	32.97	1.98	17.86	13.70
SCAPO	11.88	41.39	4.96	23.08	17.22

Usage

Use the checkpoint's chat template with the R1-style instruction shown below.

from transformers import pipeline

model = pipeline(
    "text-generation",
    model="DtYXs/SCAPO-Qwen3-1.7B",
    torch_dtype="auto",
    device_map="auto",
)
messages = [{
    "role": "user",
    "content": "What is the remainder when 7^100 is divided by 13?\n\n"
               "Please reason step by step, and put your final answer within \\boxed{}.",
}]
output = model(
    messages,
    max_new_tokens=16384,
    do_sample=True,
    temperature=0.7,
    top_p=0.9,
    top_k=0,
)
print(output[0]["generated_text"][-1]["content"])

Citation

@misc{pan2026semifactualcreditaugmentedpolicyoptimization,
  title         = {Semifactual Credit-Augmented Policy Optimization},
  author        = {Junshu Pan and Zhizhang Fu and Shulin Huang and Yiran Ding and Zifan Cheng and Wenqi Shao and Qiaosheng Zhang and Yue Zhang},
  year          = {2026},
  eprint        = {2609.40360},
  archivePrefix = {arXiv},
  primaryClass  = {cs.LG},
  url           = {https://arxiv.org/abs/2609.40360}
}

Downloads last month: 17

Safetensors

Model size

2B params

Tensor type

BF16

Model tree for DtYXs/SCAPO-Qwen3-1.7B

Base model

Qwen/Qwen3-1.7B-Base

Finetuned

(482)

this model

Quantizations

1 model

Dataset used to train DtYXs/SCAPO-Qwen3-1.7B

Paper for DtYXs/SCAPO-Qwen3-1.7B

Semifactual Credit-Augmented Policy Optimization

Paper • 2609.40360 • Published 29 days ago • 34