List your company
ResearchFormerly WizWand

Mathematical Reasoning on WeMath

84AccuracyOctopus
63.09668.52373.9579.377Sep 30, 2025Nov 15, 2025Jan 1, 2026Feb 17, 2026Apr 5, 2026May 22, 2026Jul 8, 2026
Updated 3mo ago

Evaluation results

MethodLinks
OctopusParameters=8B2026.0784——
Gemini 2.5 ProAccess=Close-source2026.0280.6——
Qwen3-VL + LOCUSSize=4B2026.0678.8——
MiMo-VL + LOCUSSize=7B2026.0678.3——
Gemini-2.5-Pro*Model Category=Close-source, Reference Source=OpenCompass leaderboard2025.0978——
Gemini-2.5-Pro*Model Category=Closed-Source Models2026.047882.7—
Gemini-2.5-ProModel Category=Proprietary Model2026.0778——
MiMo-VLSize=7B2026.0677.2——
Qwen3-VLSize=4B2026.0676.3——
Qwen2.5-VL-32B-Instruct + NoisyRolloutStudent Model=Qwen2.5-VL-32B-Instruct, Distillation/Training Method=NoisyRollout2026.0375.51——
RLR³Training Source=OpenMMR2026.0574.3——
Vision-R1Parameters=7B2026.0773.9——
MiMo-VLtraining=7B-SFT, parameters=7B2026.0173.79——
RLR³Training Source=DeepVision2026.0573.6——
Claude-3.5-Sonnet2026.0273.05——
RLVRTraining Source=DeepVision2026.0572.9——
EASE-7BModel Scale=7B2026.0572.9——
Claude 3.7Tool Use=false, Param Size=-2026.0372.6——
VGPO-7BModel Scale=7B2026.0572.5——
SRPOParameters=7B2026.0771.6——
RuCL2026.0271.49——
Semantic-backTool Use=false, Param Size=7B2026.0371.3——
BUS-8BParameters=8B2026.0771.3——
GPT-5-Thinking*Model Category=Close-source, Reference Source=OpenCompass leaderboard2025.0971.1——
GPT-5-Thinking*Model Category=Closed-Source Models2026.0471.178.4—
Qwen2.5-VL-7B + DUPLBase Model=Qwen2.5-VL-7B, RL Algorithm=DUPL2025.1071.1——
GPT5Model Category=Proprietary Model2026.0771.1——
Innovator-VLvariant=8B-Thinking, parameters=8B2026.0170.86——
PRPOModel Scale=7B, Base Model=Qwen2.5-VL-7B, Rollout=82026.0670.8——
Solution-backParameters=7B2026.0770.8——
MaLoRAModel=Qwen3-VL-8B, Training examples=5.8k2025.1070.75——
VPPO-RL-7BModel Scale=7B2026.0570.6——
NoisyRollout-7BParameters=7B2026.0270.57——
NoisyRolloutTeacher Model=Qwen2.5-VL-32B-Instruct + NoisyRollout, Student Model=Qwen2.5-VL-7B-Instruct, Distillation/Training Method=NoisyRollout2026.0370.57——
RLR³Training Source=ViRL2026.0570.4——
GPT-5 miniVersion=high2026.0570.2——
VLAA-ThinkerParameters=7B2026.0770.2——
RKLTeacher Model=Qwen2.5-VL-32B-Instruct + NoisyRollout, Student Model=Qwen2.5-VL-7B-Instruct, Distillation/Training Method=RKL2026.0370.06——
Official thinkingSource=Qwen3-VL2026.0570——
REOPOLDTeacher Model=Qwen2.5-VL-32B-Instruct + NoisyRollout, Student Model=Qwen2.5-VL-7B-Instruct, Distillation/Training Method=REOPOLD2026.0369.77——
NoisyRollout-7BModel Scale=7B, Rollout=82026.0669.6——
Qwen2.5-VL-7B + NoisyRolloutBase Model=Qwen2.5-VL-7B, RL Algorithm=NoisyRollout2025.1069.6——
GeoFocusModel Scale=7B2026.0269.4——
PAPO_D-7BModel Scale=7B2026.0569.4——
VL-Rethinker-7BModel Scale=7B, Rollout=82026.0669.3——
Qwen2.5-VL-32BParameters=32B2026.0269.1——
VL-RethinkerParameters=7B2026.0769.1——
GPT4oAccess=Close-source2026.0269——
GPT-4oTool Use=false, Param Size=-2026.0369——
GRPOModel Scale=7B2026.0268.9——
Qwen2.5-VL-7B-Instruct + Faithful-MR1Backbone model=Qwen2.5-VL-7B-Instruct, Training Data=19.2K2026.0568.9——
VPPOModel Scale=7B, Base Model=Qwen2.5-VL-7B, Rollout=82026.0668.9——
GPT-4o2026.0268.8——
VRETool Use=false, Param Size=7B2026.0368.7——
R1-ShareVL-7BModel Scale=7B, Rollout=82026.0668.7——
InternVL2.5-38BParameters=38B2026.0268.61——
Qwen2.5-VL-7B + GRPOBase Model=Qwen2.5-VL-7B, RL Algorithm=GRPO2025.1068.5——
Perception-R1-7BTraining Data=1.4K2026.0568.3——
RLVRTraining Source=ViRL2026.0568.3——
PAPO-DModel Scale=7B, Base Model=Qwen2.5-VL-7B, Rollout=82026.0668.3——
CFPOGBackbone=Qwen3-VL-2B-Thinking2026.0668.25——
VL-Rethinker-7BParameters=7B2026.0268.22——
LoRAModel=Qwen3-VL-8B, Training examples=5.8k2025.1068.1——
GRPOModel Scale=7B, Base Model=Qwen2.5-VL-7B, Rollout=82026.0668.1——
VLAA-Thinker-7BBase Model=VLAA-Thinker-7B2025.1067.7——
PAPO-7BBase Model=PAPO-7B2025.1067.6——
VL-Rethinker-7BModel Scale=7B2026.0567.5——
GRPOTeacher Model=Qwen2.5-VL-32B-Instruct + NoisyRollout, Student Model=Qwen2.5-VL-7B-Instruct, Distillation/Training Method=GRPO2026.0367.4——
RLVRTraining Source=OpenMMR2026.0567.3——
OpenVLThinker-7BBase Model=OpenVLThinker-7B2025.1067.1——
GRPOModel Scale=8B2026.0667.05——
MiMo-VLtraining=7B-RL, parameters=7B2026.0167.01——
PAPO-GModel Scale=7B, Base Model=Qwen2.5-VL-7B, Rollout=82026.0666.8——
OpenVLThinkerParameters=7B2026.0766.7——
OpenVLThinker-7BParameters=7B2026.0266.63——
Semantic-back-7BParameters=7B2026.0266.61——
MM-Eureka-7BModel Scale=7B2026.0566.6——
MM-Eureka-7BModel Scale=7B, Rollout=82026.0666.1——
Qwen3-VLparameters=8B2026.0166.03——
Qwen3-VL-8BParameters=8B2026.0766——
Qwen2.5-VL-7B-Instruct + VPPOBackbone model=Qwen2.5-VL-7B-Instruct, Training Data=19.2K2026.0565.7——
MM-Eureka-Qwen-7BBase Model=MM-Eureka-Qwen-7B2025.1065.6——
Perception-R1-7BParameters=7B2026.0265.57——
ThinkLite-VL-7BParameters=7B2026.0265.52——
MM-Eureka-7BParameters=7B2026.0265.34——
Qwen2.5-VL + LOCUSSize=7B2026.0665.1——
Innovator-VLvariant=8B-Instruct, parameters=8B2026.0165——
Vision-R1-7BParameters=7B2026.0264.98——
SAYO-Qwen-8BBase Model=Qwen3-VL-8B2026.0264.83——
Qwen2.5-VL-7B-InstructStudent Model=Qwen2.5-VL-7B-Instruct2026.0364.77——
REOPOLDTeacher Model=Qwen2.5-VL-32B-Instruct + NoisyRollout, Student Model=Qwen2.5-VL-3B-Instruct, Distillation/Training Method=REOPOLD2026.0364.6——
OpenVLThinker-7BParameters=7B2026.0264.48——
RKLTeacher Model=Qwen2.5-VL-32B-Instruct + NoisyRollout, Student Model=Qwen2.5-VL-3B-Instruct, Distillation/Training Method=RKL2026.0364.48——
NoisyRollout-7BModel Scale=7B2026.0564.4——
Qwen2.5-VLSize=7B2026.0664.4——
Vision-SR1-7BTraining Data=56K2026.0564.2——
Qwen2.5-VL-3B + DUPLBase Model=Qwen2.5-VL-3B, RL Algorithm=DUPL2025.1064.1——
SAYO-Qwen-4BBase Model=Qwen3-VL-4B2026.0263.97——
MaLoRAModel=Qwen2.5-VL-7B, Training examples=5.8k2025.1063.97——
PRPOModel Scale=3B, Base Model=Qwen2.5-VL-3B, Rollout=82026.0663.9——