Question Answering on MMLU-Pro (Adversarial Setting, test)
98.9AccuracyMA-Reasoning (ECL E)
Evaluation Results
| Method | Links | |
|---|---|---|
| MA-Reasoning (ECL E)Backbone=Gemini 3 Pro2026.01 | 98.9 | |
| MA-Outcome (ECL I)Backbone=Gemini 3 Flash2026.01 | 96.7 |
| Method | Links | |
|---|---|---|
| MA-Reasoning (ECL E)Backbone=Gemini 3 Pro2026.01 | 98.9 | |
| MA-Outcome (ECL I)Backbone=Gemini 3 Flash2026.01 | 96.7 |