9 papers
VersaQ-3D: Architecture Support for Visual Geometry Grounded Transformers via Versatile Quantization
Yipu Zhang, Jintao Cheng, Xingyu Liu +8
The paper introduces VersaQ-3D, a co-designed quantization algorithm and reconfigurable accelerator that enables low‑bit (4‑bit) inference of Visual Geometry Grounded Transformers…
Not All Tasks Quantize Equally: Fisher-Guided Quantization for Visual Geometry Transformer
Yipu Zhang, Jintao Cheng, Weilun Feng +5
Feed-forward 3D reconstruction models, represented by Visual Geometry Grounded Transformer (VGGT), jointly predict multiple visual geometry tasks such as depth estimation, camera p…
VLA-IAP: Training-Free Visual Token Pruning via Interaction Alignment for Vision-Language-Action Models
Jintao Cheng, Haozhe Wang, Weibin Li +7
Vision-Language-Action (VLA) models have rapidly advanced embodied intelligence, enabling robots to execute complex, instruction-driven tasks. However, as model capacity and visual…
VIKIN: A Reconfigurable Accelerator for KANs and MLPs with Two-Stage Sparsity Support
Wenhui Ou, Zhuoyu Wu, Yipu Zhang +2
Recently, multi-layer perceptrons (MLPs) widely used in modern AI applications suffer from limited real-time performance due to intensive memory access overhead. Kolmogorov--Arnold…
FLICKER: A Fine-Grained Contribution-Aware Accelerator for Real-Time 3D Gaussian Splatting
Wenhui Ou, Zhuoyu Wu, Yipu Zhang +3
Recently, 3D Gaussian Splatting (3DGS) has emerged as a mainstream rendering technique due to its photorealistic quality and low latency. However, processing massive numbers of non…
FLEX: Leveraging FPGA-CPU Synergy for Mixed-Cell-Height Legalization Acceleration
Xingyu Liu, Jiawei Liang, Linfeng Du +5
In this work, we present FLEX, an FPGA-CPU accelerator for mixed-cell-height legalization tasks. We address challenges from the following perspectives. First, we optimize the task…