🏢 University of Science and Technology of China

EgoChoir: a novel framework harmonizes visual appearance, head motion, and 3D objects to accurately estimate 3D human contact and object affordance from egocentric videos, surpassing existing methods.

DN-4DGS: Real-time dynamic scene rendering is revolutionized by a denoised deformable network with temporal-spatial aggregation, achieving state-of-the-art quality.

This research introduces a novel plug-and-play module that efficiently integrates prior partial order constraints into differentiable structure learning, significantly improving structure recovery qua…

DeAR: A novel framework lets LLMs solve complex problems with human-like iterative reasoning.

Instance-wise LoRA (iLoRA) boosts LLM sequential recommendation accuracy by customizing model parameters for each user, mitigating negative transfer and improving performance.

CaPaint: a novel causal spatio-temporal prediction framework that uses causal reasoning and diffusion inpainting to boost model accuracy and generalizability, especially in data-scarce settings.

Controllable long-tailed learning achieved via hypernetwork-generated diverse experts, adapting to user preferences and distribution shifts.

DDSR: a novel sequential recommendation model uses fuzzy sets and discrete diffusion to capture user behavior randomness, outperforming existing methods.

ViSu boosts semi-supervised scene text recognition by using an online generation strategy for diverse synthetic data and a novel character alignment loss to improve model generalization and robustness…

MMStar benchmark tackles flawed LVLMs evaluation by focusing on vision-critical samples, minimizing data leakage, and introducing new metrics for fair multi-modal gain assessment.