Co-rewarding
Collection
Co-rewarding is a novel self-supervised RL framework that improves training stability by seeking complementary supervision from another views. • 75 items • Updated • 1
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM
tokenizer = AutoTokenizer.from_pretrained("TMLR-Group-HF/Entropy-Qwen3-4B-Base-OpenRS")
model = AutoModelForCausalLM.from_pretrained("TMLR-Group-HF/Entropy-Qwen3-4B-Base-OpenRS", 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=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))Quick Links
Entropy Minimization: Qwen3-4B-Base trained on OpenRS
This is the Qwen3-4B-Base model trained by Entropy Minimization using OpenRS training set, as presented in the paper Co-rewarding: Stable Self-supervised RL for Eliciting Reasoning in Large Language Models.
If you are interested in Co-rewarding, you can find more details on our Github Repo [https://github.com/tmlr-group/Co-rewarding].
- Downloads last month
- 13
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="TMLR-Group-HF/Entropy-Qwen3-4B-Base-OpenRS") messages = [ {"role": "user", "content": "Who are you?"}, ] pipe(messages)