language, tags, license, datasets
| language |
tags |
license |
datasets |
|
|
|
apache-2.0 |
|
Infos
Pythia-1b supervised finetuned with Anthropic-hh-rlhf dataset for 1 epoch.
wandb log
See Pythia-1b for model details (paper).
Benchmark raw results:
Results for the base model are taken from the Pythia paper.
Zero shot
| Task |
1B_base |
1B_sft |
| Lambada (OpenAI) |
0.562 ± 0.007 |
0.563 ± 0.007 |
| PIQA |
0.707 ± 0.011 |
0.711 ± 0.011 |
| WinoGrande |
0.537 ± 0.014 |
0.534 ± 0.014 |
| WSC |
0.365 ± 0.047 |
0.365 ± 0.047 |
| ARC - Easy |
0.569 ± 0.010 |
0.583 ± 0.010 |
| ARC - Challenge |
0.244 ± 0.013 |
0.248 ± 0.013 |
| SciQ |
0.840 ± 0.012 |
0.847 ± 0.011 |
| LogiQA |
0.223 ± 0.016 |
-- |
Five shot
| Task |
1B_base |
1B_sft |
| Lambada (OpenAI) |
0.507 ± 0.007 |
0.4722 ± 0.007 |
| PIQA |
0.705 ± 0.011 |
0.7165 ± 0.0105 |
| WinoGrande |
0.532 ± 0.014 |
0.5343 ± 0.014 |
| WSC |
0.365 ± 0.047 |
0.5000 ± 0.0493 |
| ARC - Easy |
0.594 ± 0.010 |
0.6010 ± 0.010 |
| ARC - Challenge |
0.259 ± 0.013 |
0.2679 ± 0.0129 |
| SciQ |
0.920 ± 0.009 |
0.9100 ± 0.0091 |
| LogiQA |
0.227 ± 0.016 |
N/A |
