List your company
ResearchFormerly WizWand

Logical Reasoning on LogicVista

81.4AccuracyGemini3-pro
41.04851.5246272.476May 20, 2025Jul 27, 2025Oct 3, 2025Dec 10, 2025Feb 16, 2026Apr 25, 2026Jul 2, 2026
Updated 3mo ago

Evaluation results

MethodLinks
Gemini3-proModel Category=Closed-Source SOTA Models2026.0581.4
Gemini-2.5-ProModel Category=Proprietary Model2026.0773.8
Doubao-Seed-1.6Model Category=Closed-Source SOTA Models2026.0572.5
Qwen3-VL-235B-ThinkingModel Category=Open-Source Large Baselines2026.0572.3
GPT-5-20250807Model Category=Closed-Source SOTA Models2026.0570
GPT5Model Category=Proprietary Model2026.0770
GLM-4.5VModel Category=Closed-Source SOTA Models2026.0562.4
OSTBase Model=Qwen3-VL-8B-Instruct, Sampling Ratio=Best-20%2026.0561.5
Qwen3-VL-8B-InstructStrategy=Base2026.0561.3
DEITABase Model=Qwen3-VL-8B-Instruct, Sampling Ratio=Top-20%2026.0561.2
Qwen3-VL-8B-Instruct + LLM-as-a-JudgeSampling Ratio=20%2026.0560.9
Qwen3-VL-8B-Instruct + RandomSampling Ratio=20%2026.0560.6
OSTBase Model=Qwen3-VL-4B-Instruct, Sampling Ratio=Best-20%2026.0560.4
Qwen3-VL-4B-InstructStrategy=Base2026.0560
Qwen3-VL-8B-Instruct + Full SFTSampling Ratio=100%2026.0560
Qwen3-VL-4B-Instruct + LLM-as-a-JudgeSampling Ratio=20%2026.0559.7
DEITABase Model=Qwen3-VL-4B-Instruct, Sampling Ratio=Top-20%2026.0559.6
Qwen3-VL-4B-Instruct + RandomSampling Ratio=20%2026.0559.5
Qwen3-VL-30B A3B-InstructParameters=30B A3B, Type=Instruct2025.1258.2
Qwen3-VL-4B-Instruct + Full SFTSampling Ratio=100%2026.0558.2
InternVL-3.5Model Size=8B2026.0357.3
Qwen3-VL + LOCUSSize=4B2026.0656.7
Qwen3-VL-8B + Qwen3-8B / H-OPDStudent=Qwen3-VL-4B-Instruct2026.0755.9
Qwen3-VLModel Size=8B2026.0355.3
Qwen3-VL-8B2026.0755.3
MiMo-VL + LOCUSSize=7B2026.0654.8
Penguin-VLModel Size=8B2026.0353.8
Step-GUI-8BParameters=8B2025.1253.7
GPT-4oTool Use=false, Param Size=-2026.0353.2
Qwen3-VLSize=4B2026.0652.5
MiMo-VLSize=7B2026.0652.5
Qwen3-VL 8B-InstructParameters=8B, Type=Instruct2025.1252.3
Step-GUI-4BParameters=4B2025.1252.1
LASERModel Category=Open-source Vision-Language Model2026.0751.2
Kimi-vl-A3B-thinkingModel Category=Open-Source Large Baselines2026.0551
VAPO-Thinker-7BModel Category=Open-source Vision-Language Model2026.0750.9
PRPOModel Scale=7B, Base Model=Qwen2.5-VL-7B, Rollout=82026.0649.7
VisionR1-7BModel Category=Open-source Vision-Language Model2026.0749.7
ThymeTool Use=true, Param Size=7B2026.0349
DeepEyesV2Tool Use=true, Param Size=7B2026.0348.7
VRETool Use=false, Param Size=7B2026.0348.7
Qwen2.5-VL-7B + DUPLBase Model=Qwen2.5-VL-7B, RL Algorithm=DUPL2025.1048.7
V-ZeroBackbone=Qwen2.5-VL-7B-Instruct, Iteration=22026.0148.6
Qwen2.5-VL-7B + PGPOModel Size=7B, Optimization Strategy=PGPO2026.0447.93
Qwen2.5-VL-7B + DAPOModel Size=7B, Optimization Strategy=DAPO2026.0447.85
Supervised GRPOBackbone=Qwen2.5-VL-7B-Instruct2026.0147.8
InternVL3.5-2BThinking Mode=true2025.1247.7
DeepEyesParam Size=7B2025.0547.7
InternVL3.5Parameters=2B2026.0347.7
InternVL3.5#Params=2B2026.0347.7
Qwen2.5-VL-7B + VPPOModel Size=7B, Optimization Strategy=VPPO2026.0447.52
NoisyRollout-7BModel Scale=7B, Rollout=82026.0647.5
VPPOModel Scale=7B, Base Model=Qwen2.5-VL-7B, Rollout=82026.0647.5
Qwen2.5-VL-7B + NoisyRolloutBase Model=Qwen2.5-VL-7B, RL Algorithm=NoisyRollout2025.1047.5
NoisyRollout-7BModel Size=7B2026.0447.49
Qwen2.5-VL-7B + PAPOModel Size=7B, Optimization Strategy=PAPO2026.0447.46
V-ZeroBackbone=Qwen2.5-VL-7B-Instruct, Iteration=12026.0147.4
VLAA-Thinker-7BBase Model=VLAA-Thinker-7B2025.1047.3
MM-Eureka-Qwen-7BBase Model=MM-Eureka-Qwen-7B2025.1047.3
Qwen2.5-VL-7B-Instruct (Base)Backbone=Qwen2.5-VL-7B-Instruct2026.0147.2
OpenVLThinker-7BBase Model=OpenVLThinker-7B2025.1047.1
PAPO-7BBase Model=PAPO-7B2025.1047.1
Qwen2.5-VL + LOCUSSize=7B2026.0647.1
OSTBase Model=Qwen3-VL-2B-Instruct, Sampling Ratio=Best-20%2026.0547
Qwen2.5-VL-7B + GRPOModel Size=7B, Optimization Strategy=GRPO2026.0446.81
Qwen3-VL-2B-InstructStrategy=Base2026.0546.8
DEITABase Model=Qwen3-VL-2B-Instruct, Sampling Ratio=Top-20%2026.0546.7
AutoToolSize=7B2026.0546.7
PAPO-DModel Scale=7B, Base Model=Qwen2.5-VL-7B, Rollout=82026.0646.7
Qwen3-VL-2B-Instruct + LLM-as-a-JudgeSampling Ratio=20%2026.0546.5
VL-Rethinker-7BModel Scale=7B, Rollout=82026.0646.5
VL-Rethinker-7BModel Size=7B2026.0446.47
MM-Eureka-7BModel Size=7B2026.0446.3
MM-Eureka-7BModel Scale=7B, Rollout=82026.0646.3
InternVL3-8BSize=8B2026.0546
Qwen2.5-VL-7B + GRPOBase Model=Qwen2.5-VL-7B, RL Algorithm=GRPO2025.1046
Qwen2.5-VL*Param Size=7B2025.0545.9
Qwen2.5-VLTool Use=false, Param Size=7B2026.0345.9
Qwen3-VL-2B-Instruct + RandomSampling Ratio=20%2026.0545.9
R1-ShareVL-7BModel Scale=7B, Rollout=82026.0645.8
PAPO-GModel Scale=7B, Base Model=Qwen2.5-VL-7B, Rollout=82026.0645.8
R1-ShareVL-7BModel Size=7B2026.0445.76
Qwen3-VL-2B-Instruct + Full SFTSampling Ratio=100%2026.0545.6
GRPOModel Scale=7B, Base Model=Qwen2.5-VL-7B, Rollout=82026.0645.6
R1-Onevision-7BModel Category=Open-source Vision-Language Model2026.0745.6
Qwen2.5-VL-7BSize=7B2026.0545.5
R1-Onevision-7BBase Model=R1-Onevision-7B2025.1045.5
Qwen2.5-VL-7BBase Model=Qwen2.5-VL-7B2025.1045.5
DeepEyesSize=7B2026.0545.3
OpenVLThinker-7BBackbone=Qwen2.5-VL-7B-Instruct2026.0145.1
Qwen2.5-VL-3B + DUPLBase Model=Qwen2.5-VL-3B, RL Algorithm=DUPL2025.1045
PRPOModel Scale=3B, Base Model=Qwen2.5-VL-3B, Rollout=82026.0644.9
CFPOGBackbone=Qwen3-VL-2B-Thinking2026.0644.45
BAGEL#Params=7B+7B2026.0344.3
Qwen2.5-VLParam Size=7B2025.0544.1
Qwen2.5-VL-3B + PGPOModel Size=3B, Optimization Strategy=PGPO2026.0443.2
Qwen2.5-VLSize=7B2026.0643.1
Qwen2.5-VL-7BModel Size=7B2026.0443.01
Qwen2.5-VL-3B + PAPOModel Size=3B, Optimization Strategy=PAPO2026.0442.7
Qwen2.5-VL-7BModel Category=Open-source Vision-Language Model2026.0742.6