Task 1: Cross-Camera Temporal Handoff
Named
The red box marks a construction_vehicle in the first frame of Front-left. In which camera view is the same target visible at the end?
Anonymous
The red box marks a target in the first frame of Front-left. In which camera view is the same target visible at the end?
| Model | Named | Anonymous |
|---|---|---|
| gpt-5.5 · non-thinking | 31/80 (38.8%) | 28/80 (35.0%) |
| gpt-5.6-luna · non-thinking | 18/80 (22.5%) | 21/80 (26.3%) |
| gpt-5.6-terra · non-thinking | 36/80 (45.0%) | 35/80 (43.8%) |
| gpt-5.6-sol · non-thinking | 32/80 (40.0%) | 23/80 (28.7%) |
| Qwen2.5-VL-3B-Instruct-4bit | 4/80 (5.0%) | 3/80 (3.8%) |
| Qwen3.5-9B · 8-bit · non-thinking | 26/60 (43.3%) | 23/60 (38.3%) |