English

SFT-GRPO Data Overlap as a Post-Training Hyperparameter for Autoformalization

Machine Learning 2026-04-16 v1 Artificial Intelligence Logic in Computer Science

Abstract

Supervised fine-tuning (SFT) followed by Group Relative Policy Optimization (GRPO) is a common post-training recipe. We conduct a controlled ablation over SFT-GRPO data overlap, evaluating Qwen3-8B (thinking disabled) post-trained for Lean 4 autoformalization under six conditions that differ solely in training recipe: a base model, SFT-only, GRPO-only, and three SFT+GRPO configurations where 0 percent, 30 percent, or 100 percent of the GRPO prompts coincide with the SFT corpus. Keeping SFT and GRPO data disjoint consistently outperforms full overlap at zero additional compute cost. Evaluating on Gaokao-Formal and PutnamBench under both compile pass at k and semantic pass at k assessed by an LLM judge, we find that lower overlap is monotonically associated with higher compilation and semantic accuracy. At 0 percent overlap, GRPO yields a 10.4 percentage point semantic gain over SFT alone on Gaokao, while at 100 percent overlap both metrics remain flat, rendering the GRPO stage effectively redundant. We further show that dual-metric evaluation reveals compile semantic gaps exceeding 30 percentage points for the highest compiling models, a disparity invisible under compile-only benchmarking. To our knowledge, this is the first controlled investigation of SFT-GRPO data overlap as a post-training hyperparameter, demonstrating how model behavior varies based on the degree of data sharing between training stages.

Keywords

Cite

@article{arxiv.2604.13515,
  title  = {SFT-GRPO Data Overlap as a Post-Training Hyperparameter for Autoformalization},
  author = {Xiaole Su and Kasey Zhang and Andy Lyu},
  journal= {arXiv preprint arXiv:2604.13515},
  year   = {2026}
}
R2 v1 2026-07-01T12:10:11.072Z