一句话：提出 LLM 推理碳排估计的参考框架设计原则，并实现多基准驱动的逐提示碳排估算原型 SEAL。
大语言模型在软件工程中快速普及，其不断增长的碳足迹引发可持续性担忧。虽然训练阶段的排放可观，但由于推理需处理的提示数量庞大，推理排放很快超过训练阶段，因此亟需在推理阶段进行准确到提示级别的碳排测量，以支持面向可持续性的决策。针对现有方法的局限，本文提出了一套面向 LLM 推理碳排估计的新型参考框架的设计指导原则，为未来工具的设计提供指引，并为该领域可持续性研究提供系统化基础。作者同时给出这些原则的早期实例化系统 SEAL，它采用多基准驱动的方式实现逐提示碳排估算。初步验证结果展现出良好前景，使 SEAL 有望成为整个 LLM 生态中标准化可持续性评估的基础。
Large Language Models (LLMs) are rapidly gaining traction in software engineering, yet their growing carbon footprint raises pressing sustainability concerns. While training emissions are substantial, inference quickly surpasses them due to the sheer volume of prompts processed. This shift underscores the urgent need for accurate, prompt-level carbon measurement during inference to enable informed, sustainability-focused decision-making. To address the limitations of existing approaches, in this paper, we outline the guiding principles for a novel reference framework for LLM inference carbon estimation that can guide the design of future tools and provide a systematic foundation for advancing sustainability research in this domain. We also introduce SEAL, an early embodiment of these principles that leverages a multi-benchmark-driven approach for per-prompt carbon estimation. Its initial validation shows promising results, positioning SEAL as a foundation for standardized sustainability assessment across the LLM ecosystem.
@inproceedings{paperbot12007,
title = {SEALing the Gap: A Reference Framework for LLM Inference Carbon Estimation via Multi-Benchmark Driven Embodiment},
author = {Priyavanshi Pathania and Rohit Mehra and Vibhu Saujanya Sharma and Vikrant Kaulgud and Tiffani Nevels and Sanjay Podder and Adam P. Burden},
booktitle = {ICSE},
year = {2026},
doi = {10.1145/3786582.3786846}
}