Search papers, labs, and topics across Lattice.
3
0
9
7
Even near-perfect planning ability in LLMs doesn't ensure safety in robotic tasks, with the best models still generating dangerous plans almost 30% of the time.
LLM agents can appear to reason well (high entropy) while completely ignoring the input, and mutual information is a far better metric for catching this failure.
Forget end-to-end image generation – this work shows that explicitly interleaving textual reasoning with visual refinement, guided by dense supervision, yields more controllable and interpretable image synthesis.
