List your company

Geometric Question Answering on GEOQA (test)

82.74Total AccuracyVeriGeo-beam5
24.156839.365954.57569.7841Jun 2, 2026Jun 3, 2026Jun 5, 2026Jun 7, 2026Jun 8, 2026Jun 10, 2026Jun 12, 2026

Evaluation results

MethodLinks
VeriGeo-beam5Params=7B, Data Scale=8.7k, Inference=beam52026.0682.74————
VeriGeo-greedyParams=7B, Data Scale=8.7k, Inference=greedy2026.0679.31————
TR-CoT-Qwen2.5-VL-7BParams=7B, Data Scale=65K + Geo170K2026.0679.2————
GeoGen-SFT-7BParams=7B, Data Scale=224K2026.0678————
TR-CoT-InternVL2.5-8BParams=8B, Data Scale=65K + Geo170K2026.0676.7————
TR-CoT-8BParams=8B, Data Scale=87K2026.0675.9————
GeoGen-SFT-3BParams=3B, Data Scale=224K2026.0675.6————
Qwen2.5-VL-7B-InstructParams=7B, Data Scale=-2026.0672.07————
MAVIS-7BParams=7B, Data Scale=834K2026.0668.3————
G-LLaVA-13BParams=13B, Data Scale=170K2026.0667————
G-LLaVA-7BParams=7B, Data Scale=170K2026.0662.8————
GPT-4oParams=Closed, Data Scale=-2026.0661.4————
InternVL2.5-8BParams=8B, Data Scale=-2026.0659————
InternVL2-8BParams=8B, Data Scale=-2026.0656.5————
Qwen2-VL-7B/8BParams=7B/8B, Data Scale=-2026.0655.7————
FGRPOModel=Qwen3-4B2026.0653.9353.8155.458.5950.83
FedProxModel=Qwen3-4B2026.0652.9356.1952.1757.3150.28
SCAFFOLDModel=Qwen3-4B2026.0651.9560.4852.4254.6749.75
FedAvgModel=Qwen3-4B2026.0651.453.3345.3453.7951.15
FGRPOModel=Qwen2.5-7B2026.0651.366055.459.1645.34
FedProxModel=Qwen2.5-7B2026.0648.9556.6751.6852.4745.84
FedAvgModel=Qwen2.5-7B2026.0648.8758.5750.0655.3344.2
SCAFFOLDModel=Qwen2.5-7B2026.0648.4257.6252.1751.9444.99
Math-LLaVA-13BParams=13B, Data Scale=360K2026.0647.8————
GPT-4VParams=Closed, Data Scale=-2026.0643.4————
FGRPOModel=Qwen2.5-3B2026.0640.7647.6245.3446.6535.88
FedProxModel=Qwen2.5-3B2026.0639.2847.6240.9944.8935.09
SCAFFOLDModel=Qwen2.5-3B2026.0638.0851.941.3740.7934.98
FedAvgModel=Qwen2.5-3B2026.0635.5941.4336.8940.9331.78
DeepSeek-VL-7BParams=7B, Data Scale=-2026.0633.7————
FGRPOModel=Llama-3.2-11B2026.0628.4236.1931.1830.1326.37
FedAvgModel=Llama-3.2-11B2026.0626.8233.8130.0627.4925.34
SCAFFOLDModel=Llama-3.2-11B2026.0626.7535.2429.8129.3824.04
FedProxModel=Llama-3.2-11B2026.0626.4132.8630.6827.4924.49