用四阶段循序渐进训练，让小模型推理能力显著提升
Can an Easy-to-Hard Curriculum Make Reasoning Emerge in Small Language Models? Evidence from a Four-Stage Curriculum on GPT-2
- 按词汇匹配到多步符号推理分四阶段逐步训练
- 推理头激活数量增加10倍，优化步数减半达目标精度
- 适合想低成本提升小模型推理能力的研究者
我们证明，按发展顺序设计的课程能显著提升小语言模型（SLMs）的推理透明度和样本效率。具体而言，我们在一个从词汇匹配到多步符号推理的四阶段课程上训练了参数量为124M的GPT-2模型（Cognivolve），并在不进行任何任务微调的情况下评估其表现。Cognivolve在优化步数仅为单阶段基线一半时即达到目标准确率，激活的梯度显著推理头数量增加一个数量级，并将这些头移向深层网络，产生更高熵注意力，更好地平衡局部与长距离上下文。若课程顺序错乱或重置优化器，则无法复现该效果，证实进步源于阶段性推进而非额外算力。此外，最终答案成功率仍比传统训练低约30%，且我们的显著性探针在最难阶段对语言知识头检测不足，提示未来可探索混合阶段微调和探针扩展。
原文摘要 · Abstract (English)
We demonstrate that a developmentally ordered curriculum markedly improves reasoning transparency and sample-efficiency in small language models (SLMs). Concretely, we train Cognivolve, a 124 M-parameter GPT-2 model, on a four-stage syllabus that ascends from lexical matching to multi-step symbolic inference and then evaluate it without any task-specific fine-tuning. Cognivolve reaches target accuracy in half the optimization steps of a single-phase baseline, activates an order-of-magnitude more gradient-salient reasoning heads, and shifts those heads toward deeper layers, yielding higher-entropy attention that balances local and long-range context. The same curriculum applied out of order or with optimizer resets fails to reproduce these gains, confirming that progression--not extra compute--drives the effect. We also identify open challenges: final-answer success still lags a conventional run by about 30%, and our saliency probe under-detects verbal-knowledge heads in the hardest stage, suggesting directions for mixed-stage fine-tuning and probe expansion.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。