Search papers, labs, and topics across Lattice.
Adelaide University
4
0
6
Boundary-region entanglement is a critical bottleneck for GNNs, and our adaptive approach boosts classification accuracy by over 3% while maintaining model stability.
The choice between adjoint methods and PINNs hinges on parameter representation, with a hybrid approach offering a cost-effective path to high accuracy in time-dependent inverse problems.
LVSA achieves up to 3.33x compute savings while maintaining quality in long video generation, enabling efficient inference without retraining.
LLM safety classifiers can be made dramatically more robust against jailbreaks by teaching them to "think twice" via lightweight, self-reflection fine-tuning.
