Abstract
Large language models (LLMs) with reasoning capabilities have fueled a compelling narrative that reasoning universally improves performance across language tasks. We test this claim through a comprehensive evaluation of 504 configurations across seven model families--including adaptive, conditional, and reinforcement learning-based reasoning architectures--on sentiment analysis datasets of varying granularity (binary, five-class, and 27-class emotion). Our findings reveal that reasoning effectiveness is strongly task-dependent, challenging prevailing assumptions: (1) Reasoning shows task-complexity dependence--binary classification degrades up to -19.9 F1 percentage points (pp), while 27-class emotion recognition gains up to +16.0pp; (2) Distilled reasoning variants underperform base models by 3-18 pp on simpler tasks, though few-shot prompting enables partial recovery; (3) Few-shot learning improves over zero-shot in most cases regardless of model type, with gains varying by architecture and task complexity; (4) Pareto frontier analysis shows base models dominate efficiency-performance trade-offs, with reasoning justified only for complex emotion recognition despite 2.1x-54x computational overhead. We complement these quantitative findings with qualitative error analysis revealing that reasoning degrades simpler tasks through systematic over-deliberation, offering mechanistic insight beyond the high-level overthinking hypothesis.
Abstract (translated)
具有推理能力的大规模语言模型(LLMs)推动了一种引人注目的观点,即推理在各种语言任务中普遍提高了性能。我们通过评估七类不同模型家族的504种配置——包括自适应、条件和基于强化学习的推理架构——在二元分类、五类以及27类情感识别等具有不同粒度的情感分析数据集上进行了全面测试,以验证这一观点。 我们的研究发现表明,推理的有效性强烈依赖于具体任务: 1. **任务复杂性的依赖关系**:对于简单的二元分类任务,性能下降最多可达到-19.9个F1百分点(pp),而对于复杂的27类情感识别任务,则可以提升高达+16.0pp。 2. **蒸馏推理变体的表现较基础模型在简单任务上表现较差**:这些变体通常比基础模型低3-18 pp,尽管通过少量样本提示可以在某种程度上恢复性能损失。 3. **少量示例学习的改进与零示例方法相比**:无论模型类型如何,在大多数情况下,少量示例学习都能超越零示例学习方法的表现,其提升量因架构和任务复杂性的不同而异。 4. **帕累托前沿分析显示基础模型在效率-性能权衡中占主导地位**:尽管推理带来了2.1倍至54倍的计算开销,但仅对于复杂的感情识别任务而言,引入推理是合理的。 此外,我们通过定性错误分析补充了这些定量发现,揭示出推理对简单任务的表现下降主要是由于过度思考(over-deliberation),这为关于高阶过度思考假设提供了一种超越表面的理解机制。
URL
https://arxiv.org/abs/2602.24060