Search papers, labs, and topics across Lattice.
3
0
7
Chess transformers trained solely on move sequences face a "dual-capability bottleneck" where excelling at both state tracking and decision-making requires carefully balancing data diversity and quality, a tension that simple scaling cannot resolve.
VLMs struggle to simultaneously optimize for both logical accuracy and aesthetics when generating academic illustrations, a challenge that test-time scaling can significantly alleviate.
Achieve few-step trajectory-controllable video generation without sacrificing video quality or trajectory accuracy, outperforming prior distillation methods.
