Tydra: An Efficient Hybrid Model for Tabular Data
Tydra is presented as a middle path between TabPFN’s accuracy and Hydra’s speed.
The paper describes a hybrid tabular in-context model that interleaves Transformer attention layers with state space model layers. Across 30 OpenML datasets, it reports 30% faster inference than TabPFN while keeping much of TabPFN’s predictive performance. It also beats an approximately ten-times-larger Hydra model while running faster. ArXiv · AI/CL/LG's note
The paper describes a hybrid tabular in-context model that interleaves Transformer attention layers with state space model layers. Across 30 OpenML datasets, it reports 30% faster inference than TabPFN while keeping much of TabPFN’s predictive performance. It also beats an approximately ten-times-larger Hydra model while running faster. ArXiv · AI/CL/LG's note
score 5