Google TabFM predicts from tables without extra training

Google Research's TabFM is a for data arranged in rows and columns. It can handle and regression when a table mixes number columns and category columns. It does not need or search.

Example rows are given as context, and the model makes a prediction in one . This can reduce the work of training and testing a separate model for tabular prediction tasks.

Key points

  • TabFM is a Google Research for tabular data.
  • It supports and regression on mixed number and category columns.
  • It uses example data as context instead of requiring .
  • It avoids search, according to the description.
  • It produces predictions in one .
Read original