Search papers, labs, and topics across Lattice.
6
0
9
VLA models are surprisingly vulnerable to adversarial attacks via physically plausible 3D textures, causing near-total task failure in robotic manipulation.
StreamingVLA achieves a remarkable 2.4x speedup and 6.5x reduction in execution halting by asynchronously parallelizing observation, action generation, and execution stages in vision-language-action models.
Humanoid robots can now handle heavy, unknown payloads in the real world thanks to a system that identifies mass distribution via differentiable simulation.
GLM-5 doesn't just code; it engineers, showcasing unprecedented capability in tackling end-to-end software engineering challenges.
Key contribution not extracted.
Forget static datasets – RL-based co-training unlocks +20% real-world VLA performance by interactively leveraging simulation while preserving real-world capabilities.
