Rethinking Easy-to-Hard: Limits of Curriculum Learning in Post-Training for Deductive Reasoning
For researchers and practitioners using curriculum learning in LLM post-training, this work challenges its practical utility for deductive reasoning tasks.
This paper investigates curriculum learning (CL) for post-training LLMs on deductive reasoning tasks, finding that difficulty-based sequencing offers no advantage over random sampling in accuracy or response length across multiple models and training methods.
Curriculum learning (CL), motivated by the intuition that learning in increasing order of difficulty should ease generalization, is commonly adopted both in pre-training and post-training of large language models (LLMs). The intuition of CL is particularly compelling for compositional reasoning, where complex problems are built from elementary inference rules; however, the actual impact of CL on such tasks remains largely underexplored. We present a systematic empirical study of CL for post-training of LLMs, using synthetic arithmetic and logical benchmarks where difficulty is characterized by reasoning complexity rather than surface-level proxies. Surprisingly, across multiple model families and curriculum schedules, we find no robust advantage in difficulty-based sequencing over standard random sampling in either accuracy or response length. These findings persist across both supervised fine-tuning (SFT) and reinforcement learning (RL) methods. Our study suggests that, in the context of deductive reasoning, the specific ordering of training examples plays a negligible role in achieving compositional generalization, challenging the practical utility of curriculum-based post-training.