Search papers, labs, and topics across Lattice.
3
0
8
Turns out, the secret to Transformer efficiency might be hiding in plain sight: flatter loss landscapes implicitly encourage activation sparsity.
Correcting errors in long-video understanding doesn't have to be a nightmare: IMPACT-CYCLE slashes human arbitration costs by 4.8x while boosting VQA accuracy by intelligently decomposing the task and focusing human effort where it matters most.
LLMs can now generate Innovus Tcl scripts for physical design with higher accuracy, thanks to a new domain-adapted model and benchmark that tackles the data scarcity problem.
