Segmentation on PlantSeg OOD 1.0 (test)
39.38Dice (All)VTOS
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| VTOSLLM=Sonnet 4.6, Analyzer loop diagnostics=true2026.06 | 39.38 | 50 | 31.03 | 13.48 | |
| VTOS BaseLLM=Sonnet 4.6, Analyzer loop diagnostics=false2026.06 | 36.79 | 45.3 | 30.1 | 12.84 | |
| Grounded-SAM 2LLM=—, Protocol=zero-shot2026.06 | 36.56 | 51.24 | 25.03 | 9.75 | |
| VisProgLLM=Sonnet 4.62026.06 | 36.33 | 45.79 | 28.9 | 14.27 | |
| ReflexionLLM=Sonnet 4.6, K=32026.06 | 31.62 | 38.14 | 26.5 | 10.57 | |
| VLM-directLLM=Sonnet 4.62026.06 | 30.45 | 42.18 | 21.24 | 9.22 | |
| SAM 2 + CLIPLLM=—, Protocol=vanilla zero-shot2026.06 | 23.11 | 32.04 | 16.09 | 5.75 | |
| SAM 2 (segment-everything)LLM=—2026.06 | 20.05 | 30.76 | 11.64 | 3.71 | |
| VLM-directLLM=GPT-4o-mini2026.06 | 17.3 | 21.53 | 13.98 | 4.61 |