Search papers, labs, and topics across Lattice.
3
0
6
Discrete diffusion models can solve lookahead planning tasks more efficiently than autoregressive models by exploiting the asymmetry between forward and reverse generation.
Naive causal estimates of text effects are significantly biased due to text conflating treatment and covariate information, but this can be mitigated with covariate residualization.
Ditch the randomness: STATe-of-Thoughts lets you steer LLMs through reasoning tasks with interpretable actions, boosting diversity and quality while revealing the secrets to better outputs.
