Token-level Reconstruction on codeparrot-clean In-Domain (val)
99.47Reconstruction Accuracy (TRA)87M parameter Transformer Autoencoder
Evaluation results
| Method | Links | |
|---|---|---|
| 87M parameter Transformer AutoencoderNumber of Parameters=87M, Validation Tokens=3,170,816, Data Distribution=In-Domain (Code)2025.12 | 99.47 |