Search papers, labs, and topics across Lattice.
University of Science and Technology of China
5
0
11
Omni-modal RL post-training just got a whole lot faster: Relax delivers up to 2x speedups over existing systems, even for massive MoE models, without sacrificing reward convergence.
Ventricular dysfunction can be surprisingly well-predicted in a zero-shot manner from ECG diagnostic probabilities, suggesting a structured encoding of cardiac function within these representations.
LLMs can't even reproduce published physics papers end-to-end, with the best model scoring only 34% on a new benchmark designed for this purpose.
Finally, a gripper-in-hand data collection system, FeasibleCap, gives real-time feedback on robot trajectory feasibility, eliminating the need for AR/VR headsets or robot-in-the-loop hardware.
A Transformer-based ranking model can boost e-commerce orders by 6.35% while halving latency, thanks to optimizations targeting feature sparsity and low label density.
