用可信证据筛选节能大模型部署方案，避免盲目优化。
TokenPowerSandbox: Evidence-Gated CPU-First Screening for Energy-Aware LLM Serving
- CPU上运行可解释投影器，结合短时GPU探针校准工作负载。
- 能源预测平均绝对误差仅6.23%，排名相关性达0.976。
- 自动拒绝低并发场景，防止错误结论误导部署决策。
节能的大模型服务需要在真实请求模式下比较不同配置，但全面的目标GPU性能测试成本高昂，而廉价预测器在其测量范围外可能过度自信。本文提出TokenPowerSandbox，一种证据门控工作流，结合可解释的CPU端投影器、短时目标GPU探测、完整工作负载验证以及防篡改的冻结前测量溯源机制。在单块NVIDIA H100 80GB上对Qwen2.5-7B-Instruct与vLLM进行测试，通过三个锚点重复和六个开发工作负载完成工作负载迁移校准。同一冻结模型在盲测集与独立预声明的无重调确认集上共执行51次后续评估。能源预测平均绝对百分比误差（MAPE）分别为6.23%和7.35%，斯皮尔曼等级相关系数为0.976和0.933。然而，预声明的TTFT门控在并发度为4时通过（9.27% MAPE），并在并发低于4时触发拒绝（64.80%），表明能源精度无法保证延迟表现。
原文摘要 · Abstract (English)
Energy-aware LLM serving requires comparing configurations under realistic request shapes, yet exhaustive target-GPU profiling is costly and a cheap predictor can be dangerously confident outside its measured scope. We present TokenPowerSandbox, an evidence-gated workflow that combines an interpretable CPU-resident projector, short target-GPU probes, full-workload verification, and tamper-evident freeze-before-measurement provenance. On one NVIDIA H100 80GB serving Qwen2.5-7B-Instruct with vLLM, three anchor repeats and six development workloads calibrate workload transfer. The same frozen model is evaluated on a blind holdout and a separately predeclared no-refit confirmation totaling 51 post-freeze runs. Energy MAPE is 6.23% and 7.35%, with Spearman rank correlations of 0.976 and 0.933. However, a predeclared TTFT gate passes at concurrency four (9.27% MAPE) and triggers abstention below four (64.80%), showing why energy accuracy cannot certify latency.
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。