Qwen3-14B-Intuitor-MATH-1EPOCH

This model is an Intuitor-fine-tuned version of Qwen3-14B trained on the MATH dataset, as presented in the paper Learning to Reason without External Rewards.

Intuitor is a reinforcement learning method that fine-tunes large language models (LLMs) using self-certainty—the model’s own internal confidence—as the sole reward. It is built on a novel paradigm called Reinforcement Learning from Internal Feedback (RLIF), which enables LLMs to learn from intrinsic signals without external rewards or labeled data. RLIF offers a scalable and domain-agnostic fine-tuning approach for LLMs in settings where external supervision is expensive or unavailable.

Links

• 📄 Paper: Learning to Reason without External Rewards
• 💻 GitHub Repository

---

Usage

You can use this model with the Hugging Face transformers library.

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

model_id = "sunblaze-ucb/Qwen3-14B-Intuitor-MATH-1EPOCH"

tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    torch_dtype=torch.bfloat16,
    device_map="auto",
    trust_remote_code=True,
)

messages = [
    {"role": "user", "content": "Solve the following problem: $2x+5=112x\; +\; 5\; =\; 11$. What is x?"}
]

text = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True
)

model_inputs = tokenizer([text], return_tensors="pt").to(model.device)

generated_ids = model.generate(
    model_inputs.input_ids,
    max_new_tokens=100,
    do_sample=True,
    temperature=0.7,
    top_p=0.9,
    eos_token_id=tokenizer.eos_token_id,
)

response = tokenizer.batch_decode(generated_ids[:, model_inputs.input_ids.shape[1]:], skip_special_tokens=True)[0]
print(response)

---

Citation

@article{zhao2025learning,
  title   = {Learning to Reason without External Rewards},
  author  = {Zhao, Xuandong and Kang, Zhewei and Feng, Aosong and Levine, Sergey and Song, Dawn},
  journal = {arXiv preprint arXiv:2505.19590},
  year    = {2025}
}