针对大模型推理能耗问题，提出智能压缩方法，保护关键推理模块以节能。
Reasoning-Aware Compression: Identifying and Protecting Vulnerable Reasoning Circuits for Energy-Efficient LLM Deployment
- 按任务测试各模块对量化敏感度，识别脆弱推理路径
- 选择性恢复关键模块至高精度，实现能效与性能双赢
- 适合追求低功耗部署的复杂推理模型应用
大型推理模型在部署中消耗大量能源，而现有压缩方法对所有组件统一量化，可能损害关键推理电路。本文提出一种感知推理的压缩框架，基于五个推理基准（GSM8K、FOLIO、MATH-500、ProofWriter、MuSiQue）进行量化条件评估，并通过硬件级GPU能耗测量，对196-224个（层，投影）组合进行扰动扫描，识别每模块的INT4敏感度，再对最敏感电路恢复为FP16。三个发现：第一，INT4量化可能导致推理链延长，使原本25%的能耗降低转为净能耗增加（在GSM8K上）；第二，敏感度具有任务依赖性：数学推理中注意力投影更关键，逻辑推理中敏感模式随架构变化；第三，选择性压缩达成均匀方法无法企及的帕累托最优：在ProofWriter上R1-Qwen-7B Top-10%表现较FP16提升12个百分点，同时能耗降低9.7%，在五个基准的保留数据上验证有效。
原文摘要 · Abstract (English)
Large Reasoning Models (LRMs) impose substantial energy costs during deployment, yet current compression methods apply uniform quantization across all components, risking damage to critical reasoning circuits. We present a reasoning-aware compression framework that benchmarks quantization conditions across five reasoning benchmarks, GSM8K, FOLIO, MATH-500, ProofWriter, and MuSiQue, with hardware-level GPU energy measurement; profiles per-module INT4 vulnerability across all 196-224 (layer, projection) pairs via a perturbation sweep on a held-out calibration split, then selectively restores the most sensitive circuits to FP16. Three findings emerge. First, INT4 quantization can increase energy by extending reasoning chains; a 25% power reduction becomes a net energy increase on GSM8K. Second, vulnerability is task-dependent: attention projections are more critical for mathematical reasoning, and sensitivity patterns differ by architecture in logical inference. Third, selective compression achieves Pareto-optimal points inaccessible to uniform methods: R1-Qwen-7B Top-10% on ProofWriter gains +12 pp over FP16 at -9.7% energy, validated on held-out data across five reasoning benchmarks.
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。