Search papers, labs, and topics across Lattice.
Great Bay University, The University of Hong Kong
7
0
13
Vision-based tactile sensors could revolutionize robotic interaction by providing high-resolution tactile data that enhances perception and manipulation capabilities.
Ternary LLMs can now achieve efficient attention computation without the overhead of high-precision K/V processing, revolutionizing their performance.
Achieving 97.2% accuracy on N-MNIST while consuming only 0.8 pJ/SOP could redefine energy-efficient neuromorphic computing.
Ditch online optimization for robot catching: this RL-trained trajectory manifold lets robots snatch fast-moving objects out of the air with compliant finesse.
Forget training from scratch: Nexusformer lets you scale Transformers by nonlinearly expanding attention, inheriting knowledge and slashing compute by up to 41.5%.
LLMs get *more* creative at generating molecules when you add *more* constraints, defying the intuition that creativity thrives on freedom.
GPT-4's mobile proactivity is so bad (7.4% success) that a fine-tuned Qwen2 model more than doubles its performance, revealing a critical gap in current MLLMs and a path to improvement.
