| Dataset Name | SOTA Method | Metric | Trend | ||
|---|---|---|---|---|---|
| MM-Vet | REVIS | Total Score72.16 | 43 | 1mo ago | |
| MMBench | Qwen3-VL-4B-Instruct | Accuracy88.7 | 42 | 4d ago | |
| MMBench cn | Accuracy60.6 | 14 | 1mo ago | ||
| Vision-Language Evaluation Suite MMB, MMStar, MMMU, Hallusion, AI2D, OCR, SEED, SQA (test val) | Qwen2-VL-7B (Teacher) | MMB Score80.7 | 10 | 1mo ago | |
| SEED-Bench Image | Vanilla | SEEDI69.7 | 7 | 1mo ago | |
| MME | VisionZip | MME Score1,846 | 7 | 1mo ago | |
| MMBench (dev) | Prompt Highlighter | Accuracy69.7 | 4 | 1mo ago | |
| MME Perception | Prompt Highlighter | MME Score1,552.5 | 4 | 1mo ago | |
| Vision-Language Evaluation Suite (ChartQA, DocVQA, AI2D, VQA, AndroidControl, CountBenchQA) | Our Method | ChartQA Accuracy68.1 | 2 | 16d ago | |
| Vision-Language Benchmarks Hard Partition | VISOR | ChartQA Score78.1 | 2 | 24d ago | |
| Vision-Language Benchmarks Easy Partition | Qwen2-VL-2B | RealWorldQA Accuracy61.1 | 2 | 24d ago |
