[论文] CritICL: Inference-Time Weak-to-Strong Generalization from Small Langu...

论文概要

研究领域: NLP
作者: Yufan Wu, Yinghui He, Zhengyi Hu
发布时间: 2026-08-28
arXiv: 2508.11372

中文摘要

近期推理时扩展技术的进展显著提升了大语言模型（LLMs）的推理性能。然而，这些方法通常依赖重复生成或外部验证。为解决这一局限，我们提出CritICL——一种新颖的推理时框架，在保持高效率的同时提升推理能力。我们的核心洞察是：LLM的失败模式在同一家族的不同规模模型间呈现出结构化规律。CritICL不将失败视为不受欢迎的输出，而是将其作为指导来源加以利用。具体而言，我们利用从较弱模型中提取的失败模式，并通过基于批判的上下文示例将其融入推理过程。我们提出两个变体：CritICL-dynamic自适应预测输入特定的失败模式并检索批判，CritICL-static使用全局失败模式配置文件提供稳定指导。实验结果表明，CritICL始终优于标准上下文学习，并达到与测试时扩展方法相当或更优的性能，同时需要显著更少的生成次数和更低的token成本。

原文摘要

Recent advances in inference-time scaling have significantly improved the reasoning performance of large language models (LLMs). However, these methods typically rely on repeated generation or external verification. To address this limitation, we introduce CritICL, a novel inference-time framework that improves reasoning while maintaining high efficiency. Our key insight is that LLM failure modes exhibit structured patterns across model scales within the same family. Instead of treating failures as undesirable outputs, CritICL leverages them as a source of guidance. Specifically, we utilize failure modes derived from weaker models and incorporate them into inference through critique-based in-context examples. We propose two variants: CritICL-dynamic, which adaptively predicts input-specifi...

自动采集于 2026-08-29

#论文 #arXiv #NLP #小凯