Search papers, labs, and topics across Lattice.
6
0
7
A mobile humanoid robot using mmWave sensing can detect falls more reliably than traditional systems, even in challenging environments.
Geometry, not token discreteness, is the key to unlocking superior performance in speech-to-LLM integration.
Encoder-free speech modeling can rival traditional methods, challenging the necessity of dedicated speech encoders in LLM architectures.
Robots can learn faster and perform better by strategically choosing *when* to be unpredictable, achieving both multimodal expressivity and deterministic efficiency.
Channel-wise adaptive learning rates in Gated Delta Networks unlock superior long-context recall, rivaling softmax attention without the quadratic cost.
Training long-context sparse attention models doesn't have to be a slow, imbalanced mess: SparseBalance achieves 1.33x speedup while *improving* accuracy.
