Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Memorize Theorems, Not Instances: Probing SFT Generalization through Mathematical Reasoning

About

Supervised Fine-Tuning (SFT) is widely used for task-specific adaptation, yet recent work shows it systematically undermines reasoning generalization. We argue the root cause is not memorization itself, but its target: vanilla SFT drives models to exploit and memorize spurious surface correlations in problem-solution pairs, leaving them brittle to superficial input variations. To address this, we propose Theorem-SFT, which reorients supervision toward explicit theorem application by teaching models how rules are invoked rather than what answers look like. Theorem-SFT yields consistent gains across benchmarks and model families: +8.8% on MATH (LLaMA3.2-3B-Instruct) and +20.27% on GeoQA (Qwen2.5-VL-7B-Instruct) without modality-specific re-training. Fine-tuning MLP layers alone matches full-layers performance, implicating feed-forward components as the primary locus of reasoning rules. Our findings reframe the debate: Generalization failures stem not from memorization as a mechanism, but from memorizing the wrong inductive targets.

Ruiying Peng, Mengyu Yang, Jing Lei, Xiaohui Li, Xueyu Wu, Xinlei Chen• 2026

Related benchmarks

TaskDatasetResultRank
Mathematical ReasoningAGIEval MATH
Accuracy45.6
99
Mathematical ReasoningOmni-MATH
Accuracy18.4
23
Geometric ReasoningGeoQA
Accuracy69.73
15
Mathematical ReasoningGSM1K
Accuracy76.1
12
Mathematical ReasoningLiveBench
Accuracy53.6
12
Showing 5 of 5 rows

Other info

Follow for update