Situational Grounding on GROUND-Eval
53Sorting Fruits (1P)MERGE
Evaluation Results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| MERGEVLM=Gemini 2.5 Flash, Method=Ours2026.03 | 53 | 43 | 48 | 20 | 22 | 26 | 36 | 36 | 35 | 0.56 | |
| MERGEVLM=GPT-4o, Method=Ours2026.03 | 46 | 42 | 48 | 47 | 38 | 22 | 44 | 50 | 42 | 0.36 | |
| MERGEVLM=GPT-5, Method=Ours2026.03 | 46 | 36 | 48 | 29 | 42 | 40 | 50 | 0 | 38 | 1.39 | |
| MERGEVLM=Gemini 2.5 Flash Video, Method=Ours2026.03 | 45 | 43 | 55 | 19 | 45 | 24 | 50 | 22 | 39 | 0.56 | |
| GPT-5VLM=GPT-5, Method=VLM-only2026.03 | 11 | 29 | 32 | 15 | 17 | 35 | 38 | 0 | 24 | 5.52 | |
| GPT-4oVLM=GPT-4o, Method=VLM-only2026.03 | 0 | 29 | 37 | 22 | 7 | 22 | 25 | 15 | 23 | 0.66 | |
| Gemini 2.5 FlashVLM=Gemini 2.5 Flash, Method=VLM-only2026.03 | 0 | 19 | 13 | 7 | 7 | 7 | 13 | 4 | 10 | 3.23 | |
| Gemini 2.5 Flash VideoVLM=Gemini 2.5 Flash Video, Method=VLM-only2026.03 | 0 | 24 | 18 | 10 | 8 | 15 | 29 | 17 | 16 | 2.77 |