Reasoning Segmentation on EarthReason (test)
76.86gIoUGRTO
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| GRTO2026.05 | 76.86 | 79.3 | — | — | |
| B-GRTO2026.05 | 76.27 | 78.54 | — | — | |
| GRPO2026.05 | 74.31 | 76.06 | — | — | |
| Think2Seg-RSModel=Qwen-2.5-VL-7B + SAM2-Small2025.12 | 73.73 | 75.83 | — | — | |
| SegEarth-R2-3BLLM=Phi-2-2.7B2025.12 | 73.5 | 69.5 | — | — | |
| Think2Seg-RS2026.05 | 73.36 | 75.6 | — | — | |
| UniGeoSegdomain-specific=true2026.05 | 73.08 | 74.61 | — | — | |
| Text4Seg++LLM=Qwen2-7B2025.12 | 73 | 65.6 | — | — | |
| Think2Seg-RSModel=Qwen-2.5-VL-3B + SAM2-Small2025.12 | 71.09 | 74.13 | — | — | |
| RemoteReasonerLLM=Qwen2.5-7B2025.12 | 71 | 69.1 | — | — | |
| RemoteReasonerModel=Qwen-2.5-VL-7B + SAM22025.12 | 70.96 | 69.13 | — | — | |
| RemoteReasonerdomain-specific=true2026.05 | 70.96 | 69.02 | — | — | |
| SegEarth-R1LLM=Phi-1.5-1.3B2025.12 | 70.8 | 68.3 | — | — | |
| SegEarth-R1Model=Swin-B (88M) + Phi-1.5-1.3B2025.12 | 70.75 | 68.25 | — | — | |
| SegEarth-R1domain-specific=true2026.05 | 70.75 | 68.6 | — | — | |
| PSALMModel=Swin-B (88M) + Phi-1.5-1.3B2025.12 | 68.3 | 64.61 | — | — | |
| PSALM2026.05 | 68.3 | 64.61 | — | — | |
| Seg-Zero2026.05 | 67.89 | 62.43 | — | — | |
| VisionReasoner2026.05 | 67.19 | 65.41 | — | — | |
| VisionReasonerModel=Qwen-2.5-VL-7B + SAM2-Large2025.12 | 67.17 | 65.29 | — | — | |
| Seg-ZeroModel=Qwen-2.5-VL-7B + SAM2-Large2025.12 | 63.16 | 62.76 | — | — | |
| LISALLM=Vicuna-7B2025.12 | 60.9 | 59.1 | — | — | |
| LISAModel=CLIP-L (304M) + Vicuna-7B2025.12 | 60.88 | 59.1 | — | — | |
| LISA2026.05 | 60.88 | 61.04 | — | — | |
| PixelLMModel=CLIP-L (304M) + Vicuna-7B2025.12 | 60.01 | 59.22 | — | — | |
| PixelLM2026.05 | 60.01 | 59.22 | — | — | |
| PixelLMLLM=Vicuna-7B2025.12 | 60 | 59.2 | — | — | |
| GeoPixeldomain-specific=true2026.05 | 52.53 | 53.9 | — | — | |
| Earth-OneVisionScale=2B2026.06 | — | — | 57.97 | 63.41 | |
| Earth-OneVisionScale=2B, Auxiliary=SAM22026.06 | — | — | 74.2 | 62.34 | |
| GeoPixelScale=7B2026.06 | — | — | 52.53 | 53.9 | |
| GPT-5-SAMLLM=GPT-5, Trained on RS data: LLM=false, Trained on RS data: Mask Decoder=false, Trained on RS data: Extra=false2026.02 | — | — | 47.4 | — | |
| GPT-Image-1LLM=GPT-5, Trained on RS data: LLM=false, Trained on RS data: Mask Decoder=false, Trained on RS data: Extra=false2026.02 | — | — | 41 | — | |
| LISALLM=Vicuna-7B, Trained on RS data: LLM=LoRA, Trained on RS data: Mask Decoder=true, Trained on RS data: Extra=true2026.02 | — | — | 60.9 | — | |
| LISAScale=7B2026.06 | — | — | 60.88 | 61.04 | |
| PixelLMLLM=Vicuna-7B, Trained on RS data: LLM=LoRA, Trained on RS data: Mask Decoder=true, Trained on RS data: Extra=true2026.02 | — | — | 60 | — | |
| PixelLMScale=13B2026.06 | — | — | 60.01 | 57.94 | |
| PSALMLLM=phi-1.5-1.3B, Trained on RS data: LLM=true, Trained on RS data: Mask Decoder=true, Trained on RS data: Extra=true2026.02 | — | — | 68.3 | — | |
| PSALMScale=1.3B2026.06 | — | — | 68.3 | 66.61 | |
| Qwen3-VL-SAMLLM=Qwen3-VL-2B, Trained on RS data: LLM=LoRA, Trained on RS data: Mask Decoder=false, Trained on RS data: Extra=false2026.02 | — | — | 72.7 | — | |
| RemoteReasonerScale=7B, Auxiliary=SAM22026.06 | — | — | 70.96 | 69.02 | |
| SegEarth-R1LLM=phi-1.5-1.3B, Trained on RS data: LLM=true, Trained on RS data: Mask Decoder=true, Trained on RS data: Extra=true2026.02 | — | — | 70.7 | — | |
| SegEarth-R1Scale=1.3B2026.06 | — | — | 70.75 | 68.6 |