General MCQA on MMBench
76.2AccuracyBase VLM 7B + Joint FFT
Evaluation results
| Method | Links | |
|---|---|---|
| Base VLM 7B + Joint FFTTrain Data=0.7M + 1.6M2026.06 | 76.2 | |
| Scaled base VLM 7B + MERIT (Proposed, 2D)Train Data=3.6M + 1.6M2026.06 | 75.8 | |
| Base VLM 7B + MERIT (Proposed, 2D)Train Data=0.7M + 1.6M2026.06 | 75.7 | |
| Scaled base VLM 7B + Joint FFTTrain Data=3.6M + 1.6M2026.06 | 75.3 | |
| Scaled base VLM 7BTrain Data=3.6M2026.06 | 74.5 | |
| LLaVA-1.5-13B + Joint FFTTrain Data=0.7M + 0.2M2026.06 | 69.8 | |
| LLaVA-1.5-13BTrain Data=0.7M2026.06 | 67.7 | |
| Base VLM 7BTrain Data=0.7M2026.06 | 67.2 | |
| LLaVA-1.5-7BTrain Data=0.7M2026.06 | 64.3 | |
| LLaVA-7BTrain Data=0.6M2026.06 | 38.7 |