Text-rendered-as-image Retrieval on Flickr30K I2I
90.2R@1InternVL3-8B
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| InternVL3-8BEvaluation Protocol=Zero-shot, Tuning Strategy=SLQ2026.04 | 90.2 | 99.1 | 99.6 | |
| InternVL3-8BEvaluation Protocol=Zero-shot, Tuning Strategy=LoRA2026.04 | 83.1 | 96.2 | 97.9 | |
| E5-V-7BEvaluation Protocol=Zero-shot2026.04 | 79.5 | 95.2 | 97.8 | |
| InternVL3-8BEvaluation Protocol=Zero-shot, Tuning Strategy=Full FT2026.04 | 79.4 | 94.9 | 98.1 | |
| InternVL3-1BEvaluation Protocol=Zero-shot, Tuning Strategy=SLQ2026.04 | 76.9 | 95 | 97.2 | |
| InternVL3-1BEvaluation Protocol=Zero-shot, Tuning Strategy=Full FT2026.04 | 73.5 | 91.3 | 96 | |
| InternVL3-1BEvaluation Protocol=Zero-shot, Tuning Strategy=LoRA2026.04 | 73.3 | 91.8 | 96.3 |