AutomationBench (Public Set) - release-reported, metric unspecified
Agents
Release-reported AutomationBench results labeled as the public split when the source states neither the version nor whether the number is the strict all-assertions pass rate or the repository's partial-credit assertion score. Kept diagnostic rather than assigned to the strict 600-task public-set pass rate.
Top models (higher is better)
| Model | Score |
|---|---|
| LFM3-1.7B | 41.3 |
| Qwen3.5-9B | 15.1 |
| Qwen3.5-4B | 8.7 |
| LFM2.5-2.6B | 0.9 |