RLVR-SvS/SvS-Qwen-32B

Model Card for SvS-32B (from Qwen2.5-32B-Instruct)

[🌐 Website] • [🤗 Dataset] • [🤖 Models] • [📜 Paper] • [🐱 GitHub] • [🐦 Twitter] • [📕 Rednote]

The official model checkpoints for SvS. The SvS model is trained on the DAPO-Math-17k dataset along with 8k open-ended problems from DeepMath-103k (included in this repository as DAPO_17k_DeepMath_8k.parquet).

Inference

We recommend using our official inference template, which is inspired by the chat template of the Qwen2.5-Math series.

model_name = "RLVR-SvS/SvS-Qwen-32B"
device = "cuda" # the device to load the model onto

model = AutoModelForCausalLM.from_pretrained(
    model_name,
    torch_dtype="auto",
    device_map="auto"
)
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "Find the value of $x$ that satisfies the equation $4x+5 = 6x+7$."

messages = [
    {"role": "system", "content": "Please reason step by step, and put your final answer within \\boxed{}."},
    {"role": "user", "content": prompt}
]

text = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True
)
model_inputs = tokenizer([text], return_tensors="pt").to(device)

generated_ids = model.generate(
    **model_inputs,
    max_new_tokens=8192
)
generated_ids = [
    output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)
]

response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]

Cite Us

If you find the model helpful, please consider citing our paper:

@misc{liang2025pass1selfplayvariationalproblem,
      title={Beyond Pass@1: Self-Play with Variational Problem Synthesis Sustains RLVR}, 
      author={Xiao Liang and Zhongzhi Li and Yeyun Gong and Yelong Shen and Ying Nian Wu and Zhijiang Guo and Weizhu Chen},
      year={2025},
      eprint={2508.14029},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2508.14029}, 
}