A two-stage Transformer-plus-PPO feature generation framework reports improved downstream performance on 13 tabular datasets, but the described pretraining mechanism is not technically specified and results lack error bars.
On calibration of modern neural net- works
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
Two-Stage Feature Generation with Transformer and Reinforcement Learning
A two-stage Transformer-plus-PPO feature generation framework reports improved downstream performance on 13 tabular datasets, but the described pretraining mechanism is not technically specified and results lack error bars.