AA-Omniscience Public Accuracy
General QA · 2025-11-17
Share of the 600 AA-Omniscience-Public questions answered correctly, whether or not the model abstains on others. The public set is a 10% subset of the original 6,000-question pool and is reported alongside the public-set index and non-hallucination rate.
Top models (higher is better)
| Model | Score |
|---|---|
| Qwen3.5-4B | 15.7 |
| Qwen3.5-9B | 13.2 |
| LFM2.5-2.6B | 8.8 |
| LFM3-1.7B | 7.1 |