Search papers, labs, and topics across Lattice.
Beijing Academy of Artificial Intelligence, Technical University of Munich
5
0
7
Force-aware evaluation metrics reveal hidden performance gaps in humanoid robot control that conventional benchmarks miss.
A fragmented landscape of LLM agent communication protocols reveals a surprising trend toward hybrid payloads and runtime schema negotiation, but no single protocol can achieve all desired efficiencies.
Task knowledge can be efficiently reused across heterogeneous agents, slashing tracking errors by up to 99.79% while using significantly less data.
Energy consumption in NLP could plummet by up to 93% with the SpikeDecoder, a fully spiking neural network implementation of the Transformer decoder.
$\omega$-EVA's innovative Envision--Verify--Act loop allows policies to foresee the consequences of their actions, enhancing decision-making in embodied AI.
