List your company
ResearchFormerly WizWand

Mathematical Reasoning on MathVerse V

67.8AccuracyEASE-7B
28.59238.77148.9559.129May 1, 2026May 9, 2026May 18, 2026May 27, 2026Jun 4, 2026Jun 13, 2026Jun 22, 2026
Updated 4mo ago

Evaluation results

MethodLinks
EASE-7BModel Scale=7B2026.0567.8
VPPO-RL-7BModel Scale=7B2026.0567.6
VGPO-7BModel Scale=7B2026.0567.6
PAPO_D-7BModel Scale=7B2026.0566.6
PRPOModel Scale=7B, Base Model=Qwen2.5-VL-7B, Rollout=82026.0666.3
VPPOModel Scale=7B, Base Model=Qwen2.5-VL-7B, Rollout=82026.0665.5
VL-Rethinker-7BModel Scale=7B, Rollout=82026.0665
PAPO-DModel Scale=7B, Base Model=Qwen2.5-VL-7B, Rollout=82026.0664.9
VL-Rethinker-7BModel Scale=7B2026.0564.8
R1-ShareVL-7BModel Scale=7B, Rollout=82026.0664.3
PAPO-GModel Scale=7B, Base Model=Qwen2.5-VL-7B, Rollout=82026.0664.1
NoisyRollout-7BModel Scale=7B, Rollout=82026.0663.8
MM-Eureka-7BModel Scale=7B2026.0563.6
MM-Eureka-7BModel Scale=7B, Rollout=82026.0662.4
GRPOModel Scale=7B, Base Model=Qwen2.5-VL-7B, Rollout=82026.0661.7
PRPOModel Scale=3B, Base Model=Qwen2.5-VL-3B, Rollout=82026.0660.7
PVM-8B (SFT + GRPO)Backbone=8B, Training Strategy=SFT + GRPO, Model Architecture=Persistent Visual Memory2026.0559.8
PEARL-8BBackbone=8B, Training Strategy=RL-tuned2026.0558.9
Qwen3-VL-8B (SFT + GRPO)Backbone=8B, Training Strategy=SFT + GRPO2026.0558.5
CFPOGBackbone=Qwen3-VL-2B-Thinking2026.0658.43
VPPOModel Scale=3B, Base Model=Qwen2.5-VL-3B, Rollout=82026.0658.1
Euclid-8BBackbone=8B, Training Strategy=RL-tuned2026.0557.9
PAPO-DModel Scale=3B, Base Model=Qwen2.5-VL-3B, Rollout=82026.0657.7
Qwen3-VL-8B (LoRA-SFT + GRPO)Backbone=8B, Training Strategy=LoRA-SFT + GRPO2026.0557.6
PVM-8B (SFT)Backbone=8B, Training Strategy=SFT, Model Architecture=Persistent Visual Memory2026.0557.5
OneThinker-8BBackbone=8B, Training Strategy=RL-tuned2026.0557.4
Qwen3-VL-8B (SFT)Backbone=8B, Training Strategy=SFT2026.0556.9
ICoTTraining Strategy=Visual Injection2026.0556.1
CoMemoTraining Strategy=Visual Injection2026.0555
Qwen3-VL-8B (LoRA-SFT)Backbone=8B, Training Strategy=LoRA-SFT2026.0555
PVM-4B (SFT + GRPO)Backbone=4B, Training Strategy=SFT + GRPO, Model Architecture=Persistent Visual Memory2026.0555
Qwen3-VL-4B (SFT + GRPO)Backbone=4B, Training Strategy=SFT + GRPO2026.0554.6
PVM-4B (SFT)Backbone=4B, Training Strategy=SFT, Model Architecture=Persistent Visual Memory2026.0554.4
Qwen3-VL-4B (LoRA-SFT + GRPO)Backbone=4B, Training Strategy=LoRA-SFT + GRPO2026.0554.2
PAPO-GModel Scale=3B, Base Model=Qwen2.5-VL-3B, Rollout=82026.0653.6
DAPOModel Scale=3B, Base Model=Qwen2.5-VL-3B, Rollout=82026.0653.1
MemVRTraining Strategy=Visual Injection2026.0552.9
Qwen3-VL-8B-InstructBackbone=8B2026.0552.9
Qwen3-VL-4B (SFT)Backbone=4B, Training Strategy=SFT2026.0552.7
Qwen3-VL-4B-InstructBackbone=4B2026.0552.4
Qwen3-VL-4B (LoRA-SFT)Backbone=4B, Training Strategy=LoRA-SFT2026.0552.4
GRPOModel Scale=3B, Base Model=Qwen2.5-VL-3B, Rollout=82026.0652.3
NoisyRollout-7BModel Scale=7B2026.0551.7
DAPOModel Scale=7B, Base Model=Qwen2.5-VL-7B, Rollout=82026.0651
PAPOGBackbone=Qwen3-VL-2B-Thinking2026.0650.05
GRPOBackbone=Qwen3-VL-2B-Thinking2026.0645.41
Qwen2.5-VL-7BModel Scale=7B, Rollout=82026.0634.3
ThinkLite-VL-7BModel Scale=7B2026.0530.9
Qwen2.5-VL-3BModel Scale=3B, Rollout=82026.0630.1