Search papers, labs, and topics across Lattice.
Affiliation:
4
0
6
6
A staggering 42.5% of correct answers in long-document VQA can't be derived from terminal working memory alone, revealing a crucial oversight in current evaluation methods.
BehaviorWorldGen transforms how we generate realistic multi-agent interactions, leading to substantial improvements in action model performance.
Multimodal perception is no longer just an add-on: GLM-5V-Turbo bakes it directly into the core of reasoning, planning, and action.
Today's best visual coding agents still struggle to build complete websites from scratch, as revealed by a new benchmark spanning UI generation to full-stack development.
