Paper Detail
Qihan Ren, Peng Wang, Ruikun Cai, Shuai Shao, Dadi Guo, Yuejin Xie, Yafu Li, Quanshi Zhang, Xia Hu, Jing Shao, Dongrui Liu
A prevailing narrative in LLM post-training holds that supervised finetuning (SFT) memorizes while reinforcement learning (RL) generalizes. We revisit this claim for reasoning SFT with long chain-of-thought (CoT) supervision and find that cross-domain generalization is not absent but conditional, jointly shaped by optimization dynamics, training data, and base-model capability. Some reported failures are under-optimization artifacts: cross-domain performance first degrades before recovering and improving with extended training (a dip-and-recovery pattern), so shorttraining checkpoints can underestimate generalization. Data quality and structure both matter: low-quality solutions broadly hurt generalization,while verified long-CoT traces yield consistent cross-domain gains. Model capability is essential: stronger models internalize transferable procedural patterns (e.g., backtracking) even from a toy arithmetic game, while weaker ones imitate surface verbosity. This generalization is asymmetric, however: reasoning improves while safety degrades, reframing the question from whether reasoning SFT generalizes to under what conditions and at what cost.
No structured notes yet. Add `summary_sections`, `why_relevant`, `claim_impact`, or `next_action` in `papers.jsonl` to enrich this view.
No ranking explanation is available yet.
No tags.
@misc{ren2026rethinking,
title = {Rethinking Generalization in Reasoning SFT: A Conditional Analysis on Optimization, Data, and Model Capability},
author = {Qihan Ren and Peng Wang and Ruikun Cai and Shuai Shao and Dadi Guo and Yuejin Xie and Yafu Li and Quanshi Zhang and Xia Hu and Jing Shao and Dongrui Liu},
year = {2026},
abstract = {A prevailing narrative in LLM post-training holds that supervised finetuning (SFT) memorizes while reinforcement learning (RL) generalizes. We revisit this claim for reasoning SFT with long chain-of-thought (CoT) supervision and find that cross-domain generalization is not absent but conditional, jointly shaped by optimization dynamics, training data, and base-model capability. Some reported failures are under-optimization artifacts: cross-domain performance first degrades before recovering and },
url = {https://huggingface.co/papers/2604.06628},
keywords = {supervised finetuning, reinforcement learning, cross-domain generalization, optimization dynamics, long chain-of-thought, training data, base-model capability, dip-and-recovery pattern, data quality, model capability, reasoning tasks, safety degradation, code available, huggingface daily},
eprint = {2604.06628},
archiveprefix = {arXiv},
}
{}