QuantVLA：面向视觉-语言-动作模型的尺度校准训练后量化
QuantVLA: Scale-Calibrated Post-Training Quantization for Vision-Language-Action Models
为视觉-语言-动作模型提供无需训练的训练后量化方案，可复用于扩散Transformer动作头的低比特部署，并给出注意力温度匹配与输出头平衡等校准技巧。
原文依据 QuantVLA, a training-free post-training quantization (PTQ) framework that, to our knowledge, is the first PTQ approach for VLA systems and the first to successfully quantize a diffusion transformer (DiT) action head.
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论视觉-语言-动作（VLA）模型统一了具身智能体的感知、语言与控制，但由于计算和内存需求迅速增长，尤其是在模型扩展到更长时域和更大骨干网络时，其实际部署面临重大挑战。为解决这些瓶颈，我们提出 QuantVLA，一个无需训练的训练后量化（PTQ）框架，据我们所知，这是首个面向 VLA 系统的 PTQ 方法，也是首个成功量化扩散 Transformer（DiT）动作头的方法。QuantVLA 包含三个尺度校准组件：（1）选择性量化布局，将语言骨干网络和 DiT 中的所有线性层整数化，同时保持注意力投影为浮点，以保留原始算子调度；（2）注意力温度匹配，一种轻量级的逐头缩放机制，用于稳定注意力 logits，并在推理时折叠到反量化尺度中；（3）输出头平衡，一种逐层残差接口校准，用于缓解投影后的能量漂移。该框架无需额外训练，仅使用小型无标签校准缓冲区，支持低比特权重和激活的整数核，同时保持架构不变。在 LIBERO 上的代表性 VLA 模型中，QuantVLA 超越了全精度基线的任务成功率，在量化组件上实现了约 70% 的相对内存节省，为在严格计算、内存和功耗约束下实现可扩展的低比特具身智能提供了实用路径。
英文原文摘要 1605 字符
Vision-language-action (VLA) models unify perception, language, and control for embodied agents but face significant challenges in practical deployment due to rapidly increasing compute and memory demands, especially as models scale to longer horizons and larger backbones. To address these bottlenecks, we introduce QuantVLA, a training-free post-training quantization (PTQ) framework that, to our knowledge, is the first PTQ approach for VLA systems and the first to successfully quantize a diffusion transformer (DiT) action head. QuantVLA incorporates three scale-calibrated components: (1) a selective quantization layout that integerizes all linear layers in both the language backbone and the DiT while keeping attention projections in floating point to preserve the original operator schedule; (2) attention temperature matching, a lightweight per-head scaling mechanism that stabilizes attention logits and is folded into the dequantization scales at inference; and (3) output head balancing, a per-layer residual interface calibration that mitigates post-projection energy drift. The framework requires no additional training, uses only a small unlabeled calibration buffer, and supports integer kernels for low-bit weights and activations while leaving the architecture unchanged. Across representative VLA models on LIBERO, QuantVLA exceeds the task success rates of full-precision baselines, achieves about 70% relative memory savings on the quantized components, providing a practical pathway toward scalable low-bit embodied intelligence under strict compute, memory, and power constraints.