wandb: https://wandb.ai/eleutherai/pythia-rlhf/runs/8p0wfi7m?workspace=user-yongzx
Model Evals:
| Task | Version | Filter | Metric | Value | Stderr | |
|---|---|---|---|---|---|---|
| arc_challenge | Yaml | none | acc | 0.2654 | ± | 0.0129 |
| none | acc_norm | 0.2875 | ± | 0.0132 | ||
| arc_easy | Yaml | none | acc | 0.6149 | ± | 0.0100 |
| none | acc_norm | 0.5391 | ± | 0.0102 | ||
| lambada_openai | Yaml | none | perplexity | 5.6120 | ± | 0.1509 |
| none | acc | 0.6146 | ± | 0.0068 | ||
| logiqa | Yaml | none | acc | 0.1951 | ± | 0.0155 |
| none | acc_norm | 0.2796 | ± | 0.0176 | ||
| piqa | Yaml | none | acc | 0.7160 | ± | 0.0105 |
| none | acc_norm | 0.7182 | ± | 0.0105 | ||
| sciq | Yaml | none | acc | 0.8610 | ± | 0.0109 |
| none | acc_norm | 0.7930 | ± | 0.0128 | ||
| winogrande | Yaml | none | acc | 0.5754 | ± | 0.0139 |