Search papers, labs, and topics across Lattice.
Affiliation:
3
0
6
This work proposes SAVLA, an end-to-end symmetry-aware VLA model for robust and data-efficient policy learning, which keeps the pretrained vision-language backbone entirely frozen while combining it with an equivariant flow-matching action head and a learned canonicalizer.
Forget discrete time labels – RoMem's continuous phase rotation lets knowledge graphs naturally shadow obsolete facts in vector space, outperforming existing methods on temporal reasoning and zero-shot generalization.
Steer LLM attention without the memory bottleneck: SEKA unlocks prompt highlighting and other control capabilities even with FlashAttention.
