Lattice AI Research

Research focus

Training Efficiency & Optimization (3)Computer Vision (2)Inference & Quantization (2)Distributed Systems & Hardware (1)

Frequent co-authors

Yujun Lin (2)Shuo Yang (2)Shuo Yang (2)Haocheng Xi (2)

Papers (3)

Jul 7, 2026

FourTune: Towards Fully 4-Bit Efficient Post-Training for Diffusion Models

FourTune slashes memory overhead by 2.25x while matching the performance of full-precision fine-tuning in diffusion models.

Bowen Xue, Zihan Min, Xingyang Li +7

Computer Vision Inference & Quantization Training Efficiency & Optimization

Mar 10, 2026

BAIRMar 10, 2026·also NVIDIA, Tsinghua AI, Soyeon Caren Han is the corresponding

Flash-KMeans: Fast and Memory-Efficient Exact K-Means

K-means, previously relegated to offline processing, gets a 17.9x speed boost on modern GPUs thanks to Flash-KMeans' clever IO and contention optimizations.

Shuo Yang, Shuo Yang, Haocheng Xi +16

Distributed Systems & Hardware Inference & Quantization Training Efficiency & Optimization

Mar 9, 2026

BAIRMar 9, 2026·also Tsinghua AI, Soyeon Caren Han is the corresponding

SVG-EAR: Parameter-Free Linear Compensation for Sparse Video Generation via Error-aware Routing

Get 2x faster video generation from diffusion transformers without sacrificing quality, thanks to a clever parameter-free error compensation technique.

Xuanyi Zhou, Qiuyang Mang, Shuo Yang +10

Architecture Design (Transformers, SSMs, MoE)Computer Vision Training Efficiency & Optimization

Search

Haocheng Xi

Publication activitypapers/week, last 8 weeks

Research focus

Frequent co-authors

Papers (3)