SegAgent

Reported on 6 benchmarks across 2 tasks · 1 paper

Note: results are matched by exact model name. Different papers may use the same name for different model variants.

Computer Vision6 results

Instance SegmentationonRefCoCo val
Overall IoU· 2025-03-11
79.7
best: 85.41 (DeRIS-L)
SegAgent: Exploring Pixel Understanding Capabilities in MLLMs by Imitating Human Annotator Trajectories arXiv:2503.08625
Instance SegmentationonRefCOCO+ val
Overall IoU· 2025-03-11
72.49
best: 79.4 (MLCD-Seg-7B)
SegAgent: Exploring Pixel Understanding Capabilities in MLLMs by Imitating Human Annotator Trajectories arXiv:2503.08625
Instance SegmentationonRefCOCOg-val
Overall IoU· 2025-03-11
75.11
best: 79.9 (MLCD-Seg-7B)
SegAgent: Exploring Pixel Understanding Capabilities in MLLMs by Imitating Human Annotator Trajectories arXiv:2503.08625
Referring Expression SegmentationonRefCoCo val
Overall IoU· 2025-03-11
79.7
best: 85.41 (DeRIS-L)
SegAgent: Exploring Pixel Understanding Capabilities in MLLMs by Imitating Human Annotator Trajectories arXiv:2503.08625
Referring Expression SegmentationonRefCOCO+ val
Overall IoU· 2025-03-11
72.49
best: 79.4 (MLCD-Seg-7B)
SegAgent: Exploring Pixel Understanding Capabilities in MLLMs by Imitating Human Annotator Trajectories arXiv:2503.08625
Referring Expression SegmentationonRefCOCOg-val
Overall IoU· 2025-03-11
75.11
best: 79.9 (MLCD-Seg-7B)
SegAgent: Exploring Pixel Understanding Capabilities in MLLMs by Imitating Human Annotator Trajectories arXiv:2503.08625