8 papers indexed
Fukui Gao, Haixue Cui, Yifei Wang, Yadan Wang, Baiyu Li, Daokuan Zhong, et al.
Also available via: Industrial crops and products (Print)
Despite strong capabilities in data understanding and decision-making, autonomous data science agents still heavily rely on trial-and-error workflows that involve expensive computation. This bottleneck motivates models that can anticipate the effects of data science operations be…
Hao Liu, Chenghuan Huang, Ye Huang, Zhiying Wen, Hao Liu, Mohan Zhang, et al.
Video Diffusion Transformers process long spatio-temporal sequences, making self-attention the main bottleneck in high-resolution video generation. Training-free sparse attention reduces this cost, but adaptive Top-$p$ routing creates uneven per-head workloads under multi-GPU seq…
Xianzhi Ma, Shujun Wang, Xiaohan Li, Hao Liu, Changhua Pei, Jianhui li
Ultra-High-Resolution (UHR) remote sensing image understanding requires Vision-Language Models (VLMs) to capture both the global scene layout and sparse yet task-critical local details under limited computational budgets. Existing methods mainly follow two paradigms. One is passi…
Lu Dai, Ziyang Rao, Yili Wang, Hanqing Wang, Hao Liu, Hui Xiong
Fine-tuning LLMs to inject new knowledge faces a critical challenge: LLMs can quickly memorize new facts, yet fail to use them for downstream reasoning tasks. We formalize this failure as the \textit{\textbf{Knowing--Using Gap}}, characterized by an accuracy gap and a temporal la…
Deyao Zhu, Xin Zhou, Shengling Qin, Xuekai Zhu, Hangliang Ding, Shu Zhong, et al.
Pretraining scaling laws reveal that model capability improves predictably with data and compute. But learning from real world environments after deployment remains far less understood. Analyzing roughly 38,000 hours of agent interaction with the environment across 134 real world…
Dongbin Zhang, Hao Liu, Binquan Dai, Kangjie Chen, Chuming Wang, Chen Li, et al.
High-fidelity and expressive controllable human animation is essential for content creation and digital avatar applications. However, existing methods face a dilemma between expressiveness and disentanglement. Mainstream 2D pose-conditioned approaches suffer from "motion-shape en…
Ziyang Zhang, Hao Liu, Donghao Han, Te Wang, Jiyi Bian, Bingxu Li, et al.
High-resolution passive microwave imaging is important for numerical weather prediction, disaster monitoring, and oceanographic studies, but kilometer-level spatial resolution remains difficult to achieve because of aperture limitations and the high complexity of large interferom…