Captioning on RefCap pt-PT
13.5AccuracyQwen3.5-9B
Evaluation Results
| Method | Links | |
|---|---|---|
| Qwen3.5-9BOpen weights=true2026.06 | 13.5 | |
| Qwen3-VL-8BOpen weights=true2026.06 | 12.2 | |
| AMALIA-VLOpen weights=true, Fully open vision data=true, Fully open LLM=true2026.06 | 11.5 | |
| AMALIA-VL-DPOOpen weights=true, Fully open vision data=true, Fully open LLM=true2026.06 | 11.1 | |
| GLM-4.6V-FlashOpen weights=true2026.06 | 9.8 | |
| LLaVA-OV1.5-8BOpen weights=true, Fully open vision data=true2026.06 | 8.2 | |
| Salamandra-VL-7BOpen weights=true, Fully open vision data=true, Fully open LLM=true2026.06 | 8.2 | |
| MiniCPM-V4Open weights=true2026.06 | 6 | |
| Gemma-3-12BOpen weights=true2026.06 | 5.8 | |
| InternVL3.5-8BOpen weights=true2026.06 | 5.6 | |
| EuroVLM-9BOpen weights=true, Fully open LLM=true2026.06 | 5.3 | |
| TowerVision-9BOpen weights=true, Fully open vision data=true2026.06 | 4.9 | |
| Ministral-3-8BOpen weights=true2026.06 | 4 | |
| Molmo2-8BOpen weights=true, Fully open vision data=true, Fully open LLM=true2026.06 | 3.8 | |
| Perception-LM-8BOpen weights=true, Fully open vision data=true, Fully open LLM=true2026.06 | 3.1 | |
| Nemotron-Nano-12BOpen weights=true2026.06 | 2.9 |