Search papers, labs, and topics across Lattice.
The paper introduces QuantVLA, a post-training quantization (PTQ) framework designed to reduce the computational and memory demands of Vision-Language-Action (VLA) models. QuantVLA employs a selective quantization layout, attention temperature matching, and output head balancing to enable low-bit quantization of VLA models, including the DiT action head, without requiring retraining. Experiments on LIBERO demonstrate that QuantVLA achieves comparable or better task success rates than full-precision baselines, reduces memory usage by approximately 70%, and improves inference latency by 1.22x.
VLA models can now be efficiently quantized *without* retraining, even surpassing full-precision performance, thanks to a new post-training method that carefully calibrates scales across attention and output heads.
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, and delivers a 1.22x speedup in end-to-end inference latency, providing a practical pathway toward scalable low-bit embodied intelligence under strict compute, memory, and power constraints.
