Kejing Xia

Kejing Xia

Computer Science Ph.D. student at Georgia Tech

Georgia Institute of Technology · GTISC Lab · Atlanta

Kejing Xia 夏可敬

I am a first-year Computer Science Ph.D. student at Georgia Tech, advised by Prof. Wenke Lee in the GTISC Lab. My research centers on LLM post-training and agentic AI, with particular interests in reasoning, diffusion language models, and systems that make language models more useful in interactive environments.

Before joining the Ph.D. program, I earned my B.Eng. in Communication Engineering (Artificial Intelligence Track) from Wuhan University. Earlier, I studied at the No. 1 Middle School Affiliated to Central China Normal University.

CV Email Google Scholar GitHub
  • Ph.D. StudentComputer Science
  • Research FocusLLM Reasoning and Agents
  • Based inAtlanta, GA

Research Interests

LLM Reasoning and Agents

Reasoning, computer-use agents, alignment.

Diffusion LLMs

Decoding acceleration, hybrid-continuous dLLMs.

Efficient LLMs

Long-context modeling, training-free optimization.

News

  • 2025.09 Our paper D2GS was accepted by NeurIPS 2025.
  • 2025.05 Our paper LAMB was accepted by ACL 2025.
  • 2025.02 Our paper LongMamba was accepted by ICLR 2025.
  • 2024.06 Our paper A Spatio-temporal Event Transformer on Versatile Tasks for Human Behavior Analysis was accepted by IJCAI-W MiGA 2024.
  • 2024.05 I won National Second Prize and 5th Place (top 9%) in RoboCup China Open as team leader.
  • 2023.11 I won Province First Prize in the China Undergraduate Mathematical Contest in Modeling.

Publications

MetaState paper teaser
MetaState: Persistent Working Memory Enhances Reasoning in Discrete Diffusion Language Models
arXiv Preprint

Persistent memory for stronger reasoning in discrete diffusion language models.

Paper Code Project
R2-dLLM paper teaser
R2-dLLM: Accelerating Diffusion Large Language Models via Spatio-Temporal Redundancy Reduction
arXiv Preprint

Spatio-temporal redundancy reduction for faster diffusion LLM decoding.

Paper Code Project
EDITOR paper teaser
EDITOR: Effective and Interpretable Prompt Inversion for Text-to-Image Diffusion Models
arXiv Preprint

Interpretable prompt inversion for text-to-image diffusion models.

Paper Code Project
D2GS paper teaser
D2GS: Dense Depth Regularization for LiDAR-free Urban Scene Reconstruction
NeurIPS 2025

Dense depth regularization for urban scene reconstruction without LiDAR supervision.

Paper Code Project
LAMB paper teaser
LAMB: A Training-Free Method to Enhance the Long-Context Understanding of SSMs via Attention-Guided Token Filtering
ACL 2025

Attention-guided token filtering for longer-context state space models.

Paper Code Project
LongMamba paper teaser
LongMamba: Enhancing Mamba's Long-Context Capabilities via Training-Free Receptive Field Enlargement
ICLR 2025

A training-free receptive-field enlargement method for long-context Mamba models.

Paper Code Project

Research Experience

Jan. 2026 – Present

GTISC Lab, Georgia Tech

I work on MetaState: persistent working memory for reasoning in discrete diffusion language models.

May 2024 – Feb. 2026

EIC Lab, Georgia Tech

Led and contributed to R2-dLLM, Soft Token Interpolation, LongMamba, and LAMB, focusing on training-free optimization and efficient decoding.

Oct. 2023 – Apr. 2024

Signal Processing Lab, Wuhan University

Co-led BRS-E2NeRF, developing event-enhanced NeRFs for sharp reconstruction under rolling shutter blur.

2023 – 2024

CMVS, University of Oulu

Led research on event-based human behavior analysis with spatio-temporal transformers, presented at IJCAI 2024 MiGA Workshop.

Professional Experience

Jul. 2024 – May 2025

Bosch Research & Innovation Software Development Co., Ltd.

Lead Research Intern for D2GS, combining diffusion depth enhancement with 3D Gaussian models for LiDAR-free scene reconstruction.

Competition

  • RoboCup China Open (2024)Captain, National Second Prize.
  • Student Innovation & Entrepreneurship Project (2023–2024)National Prize for All-in-Focus Seeing Through Occlusions with Event and Frame.
  • China Undergraduate Mathematical Contest in Modeling (2023)Hubei Province First Prize.
  • Huazhong Cup Mathematical Modeling Challenge (2023)National First Prize.

Awards

  • Wuhan University First-Class Academic Scholarship2023–2024
  • "Luojia" Distinguished Exchange Student Scholarship2023–2024
  • Wuhan University Merit Students2021–2022, 2023–2024
  • Wuhan University Outstanding Student2022–2023
  • Wuhan University Merit and Outstanding Academic Scholarships2021–2023