Long-context Multimodal Understanding on HELMET
67.6AccuracyQwen3 VL
Evaluation results
| Method | Links | |
|---|---|---|
| Qwen3 VLCheckpoint=235B A22B2026.02 | 67.6 | |
| Qwen3 VL Plain DistillationCheckpoint=Short Stage2026.02 | 65.7 | |
| Qwen3 VLCheckpoint=32B2026.02 | 63 | |
| LongPOCheckpoint=Short Stage2026.02 | 62.9 | |
| Mistral Plain Distillation*2026.02 | 53.1 | |
| Mistral 3.1 SmallCheckpoint=24B2026.02 | 37 |