Naigang Wang

Publications

• A Self-Pruning Transformer: Extreme KV-Cache Compression with Universal Attention

  • • Davis Wertheimer
    • Haochen Shen
    • et al.
  • 2026
  • COLM 2026

  Conference paper

• DropKV: Decoupling Residual-Output Perturbation for Near-Optimal KV-Cache Eviction

  • • Aozhong Zhang
    • Selcuk Gurses
    • et al.
  • 2026
  • ICML 2026

  Workshop paper

• Universal Position Interpolation: Unified Context Scaling for Hybrid Mamba-Transformer Models

  • • Haochen Shen
    • Davis Wertheimer
    • et al.
  • 2026
  • ICLR 2026

  Conference paper

• Is Finer Better? The Limits of Microscaling Formats in Large Language Models

  • • Andrea Fasoli
    • Monodeep Kar
    • et al.
  • 2026
  • ICLR 2026

  Conference paper

• Frayed RoPE and Long Inputs: A Geometric Perspective

  • • Davis Wertheimer
    • Aozhong Zhang
    • et al.
  • 2026
  • ICLR 2026

  Conference paper

• Advancing Fluorescence Light Detection and Ranging in Scattering Media with a Physics-Guided Mixture-of-Experts and Evidential Critics

  • • Ismail Erbas
    • Ferhat Demikiran
    • et al.
  • 2025
  • NeurIPS 2025

  Workshop paper

• Accelerating LLM Inference via Dynamic KV Cache Placement in Heterogeneous Memory System

  • • Yunhua Fang
    • Rui Xie
    • et al.
  • 2025
  • IEEE Computer Architecture Letters

  Paper

• CLoQ: Enhancing Fine-Tuning of Quantized LLMs via Calibrated LoRA Initialization

  • • Yanxia Deng
    • Aozhong Zhang
    • et al.
  • 2025
  • TMLR

  Paper

• Generative AI Through CAS Lens: An Integrated Overview of Algorithmic Optimizations, Architectural Advances, and Automated Designs

  • • Chuan Zhang
    • You You
    • et al.
  • 2025
  • IEEE JESTCS

  Paper

• Compressed Decentralized Momentum Stochastic Gradient Methods for Nonconvex Optimization

  • • Wei Liu
    • Anweshit Panda
    • et al.
  • 2025
  • TMLR

  Paper