InstructEval - Alignment (HHH): leaderboard
Metric: Average (%, harmless/helpful/honest). Source: declare-lab.github.io. 28 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | flan-ul2 | 78.2 |
| 2 | flan-t5-xxl | 76.4 |
| 3 | flan-t5-xl | 75 |
| 4 | flan-t5-large | 70.4 |
| 5 | flan-t5-base | 59 |
Interactive version: theaggregate.ai/benchmark?slug=instructeval-alignment-hhh · How It Works · Data refreshed daily, snapshot 2026-09-05.