Table 2.
Category distribution, per ruleset.
| Category | Claude | Sparrow | Llama 2 | STELA | ||||
|---|---|---|---|---|---|---|---|---|
| % | N | % | N | % | N | % | N | |
| Considerate | 15.9% | 13 | 4.2% | 1 | 0.0% | 0 | 12.8% | 6 |
| Deferential | 11.0% | 9 | 0.0% | 0 | 0.0% | 0 | 0.0% | 0 |
| Factual | 2.4% | 2 | 8.3% | 2 | 6.7% | 1 | 10.6% | 5 |
| Formal | 3.7% | 3 | 4.2% | 1 | 0.0% | 0 | 2.1% | 1 |
| Harmless | 30.5% | 25 | 37.5% | 9 | 73.3% | 11 | 10.6% | 5 |
| Helpful | 4.9% | 4 | 16.7% | 4 | 6.7% | 1 | 19.1% | 9 |
| Honest | 15.9% | 13 | 20.8% | 5 | 0.0% | 0 | 10.6% | 5 |
| Humble | 4.9% | 4 | 4.2% | 1 | 0.0% | 0 | 12.8% | 6 |
| Impartial | 0.0% | 0 | 4.2% | 1 | 0.0% | 0 | 21.3% | 10 |
| Rights-respecting | 11.0% | 9 | 0.0% | 0 | 13.3% | 2 | 0.0% | 9 |
| Grand Total | 100.0% | 82 | 100.0% | 24 | 100.0% | 15 | 100.0% | 47 |
Some rules are counted in multiple categories. Thus the grand totals listed can exceed the number of rules in each set.