List your company
ResearchFormerly WizWand

Reasoning Segmentation on EarthReason (val)

76.35gIoUGRTO
8.843626.369343.89561.4207Sep 26, 2025Nov 4, 2025Dec 14, 2025Jan 23, 2026Mar 3, 2026Apr 12, 2026May 22, 2026
Updated 5mo ago

Evaluation results

MethodLinks
GRTO2026.0576.3578.32——
B-GRTO2026.0576.2577.59——
GRPO2026.0573.7673.32——
UniGeoSegdomain-specific=true2026.0572.5473.3——
Think2Seg-RSModel=Qwen-2.5-VL-7B + SAM2-Small2025.1272.4673.62——
SegEarth-R2-3BLLM=Phi-2-2.7B2025.1272.368.1——
Think2Seg-RS2026.0572.1672.87——
Text4Seg++LLM=Qwen2-7B2025.1271.969.8——
Think2Seg-RSModel=Qwen-2.5-VL-3B + SAM2-Small2025.1269.371.28——
RemoteReasoner2026.0469.29——67.04
RemoteReasonerdomain-specific=true2026.0569.1367.8——
RemoteReasonerModel=Qwen-2.5-VL-7B + SAM22025.1269.0267.8——
RemoteReasonerLLM=Qwen2.5-7B2025.126967.8——
SegEarth-R1LLM=Phi-1.5-1.3B2025.1268.664.1——
SegEarth-R1Model=Swin-B (88M) + Phi-1.5-1.3B2025.1268.664.13——
SegEarth-R1Evaluation Protocol=Full Amount Fine-tune2025.0968.6———
SegEarth-R1domain-specific=true2026.0568.2564.13——
Seg-Zero2026.0567.1565.41——
PSALMModel=Swin-B (88M) + Phi-1.5-1.3B2025.1266.6162.03——
PSALMEvaluation Protocol=Full Amount Fine-tune2025.0966.61———
PSALM2026.0566.6162.03——
VisionReasoner2026.0566.5867.13——
VisionReasonerModel=Qwen-2.5-VL-7B + SAM2-Large2025.1266.567.98——
Seg-ZeroModel=Qwen-2.5-VL-7B + SAM2-Large2025.126361.93——
LISAModel=CLIP-L (304M) + Vicuna-7B2025.1261.0457.39——
LISAEvaluation Protocol=Full Amount Fine-tune2025.0961.04———
LISALLM=Vicuna-7B2025.126157.4——
LISA2026.0559.157.39——
PixelLMModel=CLIP-L (304M) + Vicuna-7B2025.1257.9457.79——
PixelLMEvaluation Protocol=Full Amount Fine-tune2025.0957.94———
PixelLM2026.0557.9457.79——
PixelLMLLM=Vicuna-7B2025.1257.957.8——
Geo-R1 (GRPO)Evaluation Protocol=10-shot Fine-tune, Number of training samples=2402025.0957.78———
SegEarth-R1Evaluation Protocol=10-shot Fine-tune, Number of training samples=2402025.0956.4———
Geo-R1 (DAPO)Evaluation Protocol=10-shot Fine-tune, Number of training samples=2402025.0955.56———
Geo-R1 (GRPO)Evaluation Protocol=5-shot Fine-tune, Number of training samples=1202025.0954.73———
Geo-R1 (DAPO)Evaluation Protocol=5-shot Fine-tune, Number of training samples=1202025.0954.46———
RemoteAgent2026.0452.22——55.6
GeoPixeldomain-specific=true2026.0552.1354.23——
Geo-R1 (GRPO)Evaluation Protocol=1-shot Fine-tune, Number of training samples=242025.0950.3———
Geo-R1 (DAPO)Evaluation Protocol=1-shot Fine-tune, Number of training samples=242025.0950.09———
SegEarth-R1Evaluation Protocol=5-shot Fine-tune, Number of training samples=1202025.0945.37———
SegEarth-R1Evaluation Protocol=1-shot Fine-tune, Number of training samples=242025.0942.47———
Qwen2.5-VL-7B2026.0441.8——38.77
Qwen2.5-VL w/ thinkingEvaluation Protocol=Zero-shot Baseline2025.0919.35———
DeepSeek-VL2-tiny2026.0418.62——17.51
GeoChat2026.0411.44——12.57
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——46—
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——38.4—
LISALLM=Vicuna-7B, Trained on RS data: LLM=LoRA, Trained on RS data: Mask Decoder=true, Trained on RS data: Extra=true2026.02——61—
PixelLMLLM=Vicuna-7B, Trained on RS data: LLM=LoRA, Trained on RS data: Mask Decoder=true, Trained on RS data: Extra=true2026.02——57.9—
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——66.6—
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——70.6—
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——68.6—