Search papers, labs, and topics across Lattice.
4
6
7
Looping discrete embeddings with continuous hidden states enables near-perfect accuracy in multi-hop reasoning with fewer training steps than traditional methods.
Multi-token prediction is the key to stabilizing long-horizon forecasts of oscillatory signals, revealing a threshold where traditional methods fail.
Transformers can provably internalize chain-of-thought reasoning, matching the sample efficiency of explicit CoT while eliminating its inference overhead.
A global consensus on AI safety risks and capabilities has emerged from a panel of 100+ independent experts, representing a landmark effort in international collaboration.
