General AI Assistant on GAIA Level-1/2 seed 42, n=139 per condition
9.4Accuracydirect
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| directCondition=direct (no probe), Backbone=gpt-4o-mini2026.06 | 9.4 | 0.92 | 0 | 9.4 | 9.3 | |
| AURA (probe)Condition=probe (Explore on), Backbone=gpt-4o-mini2026.06 | 8.6 | 20.5 | 3.5 | 7.5 | 9.3 |