Visual Question Answering on MC-Bench
75.4AccuracyEAGLE
Evaluation results
| Method | Links | |
|---|---|---|
| EAGLEAgent=Qwen, GLM, InternVL2026.05 | 75.4 | |
| ReConcileAgent=Qwen, GLM, InternVL2026.05 | 73.6 | |
| DARTAgent=Qwen, GLM, InternVL2026.05 | 73.1 | |
| Debate(Vote)Agent=Qwen, GLM, InternVL2026.05 | 72.8 | |
| Debate(Judge)Agent=Qwen, GLM, InternVL2026.05 | 72.6 | |
| Self-ConsistencyAgent=Qwen2026.05 | 68.6 | |
| Zero-shot CoTAgent=Qwen2026.05 | 68 | |
| Zero-shot CoTAgent=InternVL2026.05 | 62.4 | |
| Zero-shot CoTAgent=GLM2026.05 | 53 | |
| Self-RefineAgent=Qwen2026.05 | 27.1 |