works on

From the 2 of 7 linked papers with an AI index.

collaborators

7 papers

cs.CV2026

FeatFix: Reuse What You Verify through Local Exact-Feature Correction for Faster Cached Diffusion Inference

Hanshuai Cui, Zhiqing Tang, Zhi Yao +3

FeatFix reuses exact intermediate features computed for verification to locally correct draft outputs in cached diffusion inference, speeding up image and video generation while pr…

cs.AI2026

MemTxn: A Transaction Boundary for Source-Supported Updates and Complete-State Recovery in Agent Memory

Hanshuai Cui, Zhiqing Tang, Zhi Yao +3

MemTxn is a governance layer for large language model agents that adds a transaction boundary to verify source-supported memory updates, resolve conflicting facts, and recover a co…

cs.CV2026

Not All Frames Deserve Full Computation: Accelerating Autoregressive Video Generation via Selective Computation and Predictive Extrapolation

Hanshuai Cui, Zhiqing Tang, Zhi Yao +3

Autoregressive (AR) video diffusion models enable long-form video generation but remain expensive due to repeated multi-step denoising. Existing training-free acceleration methods…

cs.CV2026

BWCache: Accelerating Video Diffusion Transformers through Block-Wise Caching

Hanshuai Cui, Zhiqing Tang, Zhifei Xu +3

Recent advancements in Diffusion Transformers (DiTs) have established them as the state-of-the-art method for video generation. However, their inherently sequential denoising proce…

cs.CV2026

Predict to Skip: Linear Multistep Feature Forecasting for Efficient Diffusion Transformers

Hanshuai Cui, Zhiqing Tang, Qianli Ma +3

Diffusion Transformers (DiT) have emerged as a widely adopted backbone for high-fidelity image and video generation, yet their iterative denoising process incurs high computational…

cs.NI2025

Semantic-Aware Caching for Efficient Image Generation in Edge Computing

Hanshuai Cui, Zhiqing Tang, Zhi Yao +2

Text-to-image generation employing diffusion models has attained significant popularity due to its capability to produce high-quality images that adhere to textual prompts. However…