推理的能源成本：利用测试时计算分析大语言模型的能源使用
The Energy Cost of Reasoning: Analyzing Energy Usage in LLMs with Test-time Compute
AI总结：
本文研究测试时计算（TTC）相比增大模型规模在准确率与能源效率上的优势，发现TTC在复杂推理任务中表现更优，且根据查询复杂度动态调整推理计算资源可显著提升效率，为语言模型的可持续部署提供新方向。
摘要：
扩展大语言模型（LLMs）已推动显著进步，但面临边际收益递减和能源需求不断上升的问题。本工作探讨测试时计算（TTC）如何作为传统扩展策略的节能补充，通过在推理时而非训练期间分配额外计算资源来实现。具体而言，我们研究采用TTC是否能比单纯增大模型规模实现更优的准确率-能源权衡。我们的实证分析表明，TTC在准确率/能源效率上超越了传统模型扩展，在需要复杂推理而非单纯事实回忆的任务中收益尤为显著。此外，我们识别出TTC性能与输出序列长度之间的关键交互关系，证明根据查询复杂度在推理时策略性地调整计算资源可大幅提升效率。我们的研究结果主张将TTC作为一个有前景的方向，使未来语言模型的部署更加可持续、准确且具有适应性。
英文摘要：
Scaling large language models (LLMs) has driven significant advancements, yet it faces diminishing returns and escalating energy demands. This work explores how test-time compute (TTC) can serve as an energy-efficient complement to conventional scaling strategies by allocating additional computational resources at inference time rather than during training. Specifically, we investigate whether employing TTC can achieve superior accuracy-energy trade-offs compared to simply increasing model size. Our empirical analysis reveals that TTC surpasses traditional model scaling in accuracy/energy efficiency, with notable gains in tasks demanding complex reasoning rather than mere factual recall. Further, we identify a critical interaction between TTC performance and output sequence length, demonstrating that strategically adjusting compute resources at inference time according to query complexity can substantially enhance efficiency. Our findings advocate for TTC as a promising direction, enabling more sustainable, accurate, and adaptable deployment of future language models.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
