Seven Papers Accepted by DAC'2026

Seven papers on efficient AI are accepted by DAC'2026 as regular papers.

Last updated on Feb 28, 2026

Efficient AI Accelerators:

  • EdgeSC: Universal Stochastic Computing Architecture for Efficient Edge Detection (collaboration w/ Prof. Ngai Wong from HKU)
  • DRIFT: Harnessing Inherent Fault Tolerance for Efficient and Reliable Diffusion Model Inference
  • NASiC: 3D NAND-based CAM-Selected Multibit CIM Architecture for Efficient On-Device Mixture-of-Experts LLM Inference
  • S2CIM: A Secure-Computation and Secure-Storage Compute-in-Memory Architecture with Circuit-Algorithm Co-Design for Efficient and Trustworthy Edge Inference (led by Prof. Kechao Tang from PKU)

Efficient AI Algorithms:

  • KEEP: A KV-Cache-Centric Memory Management System for Efficient Embodied Planning
  • Orchestrating Dual-Boundaries: An Arithmetic Intensity Inspired Acceleration Framework for Diffusion Language Models
  • DySL-VLA: Efficient Vision-Language-Action Model Inference via Dynamic-Static Layer-Skipping for Robot Manipulation
Conference
var dimensionValue = 'SOME_DIMENSION_VALUE'; ga('set', 'dimension1', dimensionValue);