Track trials, findings, and clinical evidence as it lands.
Search papers, labs, and topics across Lattice.
5
0
8
Diffusion models can learn new objectives by stochastically routing attention within their existing architecture, achieving 4x faster RL training without sacrificing generation quality.
Native 8-bit LLM training silently destabilizes at billion-parameter scale due to softmax gradient drift, but restoring the zero-row-sum invariant completely eliminates the FP8 attention gap without architecture modifications.
ActWorld bridges the navigation-interaction gap in interactive world models, enabling rich object interactions that were previously unattainable.
Offloading avatar reconstruction can enable over 2.3 times more users in VR while maintaining privacy and minimizing quality loss.
General-purpose LLM agents stumble badly when faced with the messy reality of diverse, multi-domain tasks, and simply scaling interactions or parallel sampling doesn't fix it.