| Task Name | Dataset Name | SOTA Result | Trend | |
|---|---|---|---|---|
| Open domain dialogue | Bloom ZS | RSR47.8 | 9 | |
| Red Teaming against BB-3B | Bloom ZS | RSR4,120 | 9 | |
| Red Teaming | Bloom ZS (filtered hard positive) | RSR15.6 | 7 | |
| Open-domain dialogue red teaming | Bloom ZS (filtered) (test) | RSR16.3 | 7 | |
| Language Identification | BLOOM | Macro F195.76 | 5 | |
| Language Modeling | BLOOM-1b7 (train) | Training PPL15.1 | 3 | |
| Attention Head Health Analysis | BLOOM-1b7 internal attention heads | Healthy Heads Count379 | 3 |