xiaolesu/OsmosisProofling-SFT-NT-GRPO-NT-No-Overlap

Experimental checkpoint from "Data Overlap as a Post-Training Hyperparameter for Autoformalization." This is the SFT+GRPO with 0% overlap variant (Qwen3-8B, thinking disabled) -- the best-performing condition, where SFT and GRPO data are fully disjoint. See the paper repo for details, results, and all artifacts.

📄 Paper

This model is part of the experiments in:

SFT-GRPO Data Overlap as a Post-Training Hyperparameter for Autoformalization
Xiaole Su, Kasey Zhang, Andy Lyu
https://arxiv.org/abs/2604.13515