Search papers, labs, and topics across Lattice.
3
0
7
A single round of targeted feedback can boost DRA performance by up to 15 points, but subsequent revisions may undo prior gains.
Fine-tuning time series foundation models on data sub-domains identified via Bayesian mixtures beats fine-tuning on entire datasets, boosting zero-shot forecasting in new domains.
Many popular cooperative MARL benchmarks don't actually require sophisticated reasoning about hidden states or robust coordination, as simple reactive policies often perform just as well.
