Search papers, labs, and topics across Lattice.
School of Computer Science and Technology , University of Science and Technology of China , Hefei , China
2
0
5
3
HarnessRisk reveals that up to 80.9% of adversarial attacks can succeed in agent harnesses, even when risk detection is high.
This work proposes a multi-layered framework, AID-SR, that establishes a closed loop by translating simulator-observed physical states into structured feedback for the LLM designer, and promotes the generation of physically feasible and functionally meaningful robot designs.
