List your company

Geometry Problem Solving on Geometry3K (test)

95.2Choice AccuracyBiNSGPS
26.35244.22662.179.974Feb 22, 2023Sep 13, 2023Apr 3, 2024Oct 24, 2024May 15, 2025Dec 4, 2025Jun 26, 2026

Evaluation results

MethodLinks
BiNSGPSBase Model=Qwen3-VL-Plus, Solver Type=Neural-Symbolic Solvers2026.0695.290.5—
Human Expert2023.0290.9——
SD-GPSType=Symbolic and Neuro-symbolic Solver2026.0690.486.4—
BiNSGPSBase Model=Qwen3-VL-32B, Solver Type=Neural-Symbolic Solvers2026.0690.387—
Kimi-K2.5Type=MLLM2026.0687.282.9—
GPT-5.2Solver Type=MLLMs2026.0684.577.9—
LANSModel Type=Neural Solver2025.0582.37,130—
LANSSolver Type=Neural Solvers2026.0682.372.1—
LANSType=Neural Solver, Diagram GT=true2026.0682.372.1—
AutoGPSAgent Backbone=GPT-4o2025.0581.67,540—
AutoGPSBase Model=GPT-4o, Solver Type=Neural-Symbolic Solvers2026.0681.675.4—
Auto-GPSType=Symbolic and Neuro-symbolic Solver2026.0681.675.4—
AutoGPSBase Model=Qwen3-VL-Plus, Solver Type=Neural-Symbolic Solvers2026.0681.174.8—
AutoGPSAgent Backbone=Qwen2.52025.0578.27,100—
PGPSNet2023.0277.96580.7
PGPSNetModel Type=Neural Solver2025.0577.96,500—
PGPSNetSolver Type=Neural Solvers2026.0677.965.2—
PGPSNetType=Neural Solver2026.0677.965—
Pi-GPSSolver Type=Neural-Symbolic Solvers2026.0677.870.6—
Pi-GPSType=Symbolic and Neuro-symbolic Solver2026.0677.870.6—
AutoGPSAgent Backbone=InternVL32025.0577.67,020—
PGPSNet-v2-SType=Neural Solver2026.0676.465.2—
InterGPS (All GT)Input Mode=ground truth text and diagram, re-produced=true2023.0275.969—
Auto-GPSType=Symbolic and Neuro-symbolic Solver, OCR=true2026.0675.265.6—
InternVL3-78BModel Type=MLLM2025.0574.55,740—
InternVL3-78BSolver Type=MLLMs2026.0674.557.4—
InternVL3-78BType=MLLM2026.0674.557.4—
Pi-GPSType=Symbolic and Neuro-symbolic Solver, OCR=true2026.0673.166.1—
InterGPS (Diagram GT)Input Mode=text parser and ground truth diagram, re-produced=true2023.0271.764.2—
GOLDGround Truth diagram annotations=true2024.0569.1——
GeoDRLModel Type=Symbolic Solver2025.0568.4——
GeoDRLSolver Type=Symbolic Solvers2026.0668.4——
GeoDRLType=Symbolic and Neuro-symbolic Solver2026.0668.457.9—
E-GPSModel Type=Symbolic Solver2025.0567.9——
E-GPSSolver Type=Symbolic Solvers2026.0667.9——
E-GPSType=Symbolic and Neuro-symbolic Solver2026.0667.9——
Qwen2.5-VL-32BModel Type=MLLM2025.0567.64,190—
Qwen2.5-VL-32BType=MLLM2026.0667.641.9—
Qwen3-VL-PlusSolver Type=MLLMs2026.0667.459.8—
GLM-4.6VType=MLLM2026.0665.353.7—
PGPSNetGround Truth diagram annotations=true2024.0565——
InterGPSGround Truth diagram annotations=true2024.0564.2——
InterGPSModel Type=Symbolic Solver2025.0563.55,060—
InterGPSSolver Type=Symbolic Solvers2026.0663.550.6—
GOLDGround Truth diagram annotations=false2024.0562.7——
GOLDType=Neural Solver2026.0662.7——
InternVL2.5-78BModel Type=MLLM2025.0560.93,610—
InternVL2.5-78BSolver Type=MLLMs2026.0660.936.1—
InternVL2.5-78BType=MLLM2026.0660.936.1—
PGPSNetGround Truth diagram annotations=false2024.0559.5——
Geoformerre-implemented=true2023.0259.336.862.5
GeoformerModel Type=Neural Solver2025.0559.33,680—
GeoformerSolver Type=Neural Solvers2026.0659.336.8—
GeoformerType=Neural Solver2026.0659.336.8—
NGSre-implemented=true2023.0258.835.362
NGSModel Type=Neural Solver2025.0558.83,530—
NGSSolver Type=Neural Solvers2026.0658.835.3—
NGSType=Neural Solver2026.0658.835.3—
Qwen2.5-VL-32BSolver Type=MLLMs2026.0658.646.3—
GPT-4oType=MLLM2026.0658.634.8—
Inter-GPSType=Symbolic and Neuro-symbolic Solver2026.0657.543.4—
Vision-R1-7BModel Type=MLLM2025.0557.14,380—
GPT-4oModel Type=MLLM2025.0557.14,630—
Vision-R1-7BSolver Type=MLLMs2026.0657.143.8—
GPT-4oSolver Type=MLLMs2026.0657.146.3—
Vision-R1-7BType=MLLM2026.0657.143.8—
InterGPS (Predict)Input Mode=diagram and text formal language are predicted, re-produced=true2023.0256.944.6—
GNSBase Model=LLaVA-1.5-13B, Solver Type=Neural-Symbolic Solvers2026.0653.8——
QwenVL3-2BType=MLLM2026.0652.536.6—
Qwen3-VL-7BSolver Type=MLLMs2026.0650.144.8—
InterGPSGround Truth diagram annotations=false2024.0548.6——
GeoformerGround Truth diagram annotations=false2024.0536.8——
Baseline (Neural Solver)2023.0235.9——
NGSGround Truth diagram annotations=false2024.0535.3——
G-LLaVA-13BModel Type=MLLM2025.052930—
G-LLaVA-13BType=MLLM2026.06290.3—
SCA-GPSType=Neural Solver2026.06—76.7—