This model is fine-tuned using SimKO (Simple Pass@K Policy Optimization) with TS2 (Top-Sampling Strategy) on mathematical reasoning tasks.
1from transformers import AutoTokenizer, AutoModelForCausalLM
2
3model_name = "stuziyang/MATH-Llama-3.2-3B-SimKO-TS2-nosquare"
4tokenizer = AutoTokenizer.from_pretrained(model_name)
5model = AutoModelForCausalLM.from_pretrained(model_name)
6
7# Your inference code here
This model was trained using the SimKO reinforcement learning framework with the TS2 strategy for improved pass@K performance on mathematical reasoning tasks.
If you use this model, please cite the SimKO paper and repository.
