用4比特浮点量化实现推理精度不降反升，解决大模型部署瓶颈。
ReQAT: Achieving Full-Precision Reasoning Accuracy with 4-bit Floating-Point Quantization-Aware Training
- 针对低熵符号错误设计三组件训练框架，聚焦关键推理节点。
- 在相同训练成本下超越全精度微调，推理速度最高提升3.9倍。
- 适合追求高效高精度推理的工业级大模型部署场景。
大型推理模型通过长链式思维实现强问题求解能力，但其部署受限于全精度推理的高成本及不断增长的键值缓存（KV cache）占用。微缩版FP4格式可实现高效的FP4部署；然而，将权重、激活和KV缓存全部量化为4比特（W4A4KV4）会导致严重推理性能下降，现有量化感知训练（QAT）和后训练量化（PTQ）无法恢复。我们发现，FP4失败主要集中在低熵令牌——如数字和运算符等精确符号——这些位置的量化噪声会放大采样误差，并在推理轨迹中持续传播。基于此洞察，我们提出ReQAT：一种以推理为中心的FP4训练框架，包含三个组件：(i) 跟踪对齐量化感知训练（TAQ），通过重访相同推理轨迹，将更新聚焦于关键低熵决策点；(ii) 选择性熵最小化（SEM），强化低熵位置的置信度；(iii) Q-FIT，一种量化友好初始化方法，联合校准旋转位置编码（RoPE）一致的KV缓存变换，以稳定量化训练过程。在相同训练预算下，ReQAT不仅恢复甚至超越了BF16微调的精度，在NVIDIA DGX Spark上实现最高3.9倍的吞吐提速，在B200上实现3.1倍提速。
原文摘要 · Abstract (English)
Large Reasoning Models (LRMs) achieve strong problem-solving through long chain-of-thought, but their deployment is constrained by the high cost of full-precision inference and growing KV cache footprints. Microscaled FP4 formats enable efficient FP4 deployment; however, fully quantizing weights, activations, and KV caches (W4A4KV4) causes severe reasoning degradation that existing PTQ and QAT fail to recover. We identify that FP4 failures concentrate on low-entropy tokens--precise symbolic commitments such as digits and operators--where quantization noise inflates sampling errors that cascade through reasoning traces. Based on this insight, we propose ReQAT, a reasoning-centric FP4 training framework with three components: (i) Trace-Aligned QAT (TAQ), which revisits identical reasoning traces to focus updates on critical low-entropy decisions; (ii) Selective Entropy Minimization (SEM), which reinforces confidence at low-entropy positions; and (iii) Q-FIT, a quantization-friendly initialization that jointly calibrates RoPE-consistent KV cache transformations to stabilize QAT. Under the same training budget, ReQAT not only recovers but surpasses BF16 fine-tuning accuracy, while delivering up to 3.9x throughput speedup on NVIDIA DGX Spark and 3.1x on B200.
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。