List your company
ResearchFormerly WizWand

Mathematical Multimodal Reasoning on MathVista (Accuracy)

85.6AccuracySeed-1.5-thinking
71.35275.05178.7582.449Jun 5, 2025Aug 7, 2025Oct 9, 2025Dec 11, 2025Feb 12, 2026Apr 16, 2026Jun 18, 2026
Updated 4mo ago

Evaluation results

MethodLinks
Seed-1.5-thinkingAccess Type=Closed-Source2025.1285.6
Gemini-2.5-Pro-ThinkingAccess Type=Closed-Source2025.1283.8
Gemini 2.5 ProTool=Code, Param Size=-2025.1183
MM-Eureka-32B-R-TAPModel Category=Open-Source Reasoning Models2026.0380.2
InternVL3.5-8B-MPOTraining=DF-GRPO2026.0379.9
Qwen2.5-VL-7B-Instruct + R-TAPModel Size=7B2026.0379.4
MM-Eureka-7B-R-TAPModel Category=Open-Source Reasoning Models2026.0379.3
InternVL3-78BModel Category=Open-Source2025.0679
Qwen2.5-VL-32BTraining=DF-GSPO2026.0378.8
InternVL3.5-8B-MPOTraining=PRIME2026.0378.5
InternVL3.5-8B-MPOTraining=GRPO2026.0378.4
VL-Rethinker-72BModel Category=Open-Source2025.0678.2
R1-ShareVL-32B2026.0377.6
InternVL-3.5-4BModel Scale=4B2026.0277.1
Qwen2.5-VL-7B (Bo8)Training Strategy=GRPO, Sampling Strategy=Best-of-82025.0676.8
Qwen2.5-VL-32BReasoning Framework=Standard2025.0676.8
Qwen2.5-VL-32B w/ RAPIDReasoning Framework=RAPID, Reasoner Model=Qwen3-8B2025.0676.8
InternVL2.5-78B-MPOModel Category=Open-source Models (>70B)2025.0676.6
DenseMLLM-4BModel Scale=4B2026.0276.5
Vision-G1Access Type=Open-Source, Model Scale=7B2025.1276.1
Ovis2-34BModel Category=Open-Source2025.0676.1
Qwen2.5-VL-7B w/ RAPIDReasoning Framework=RAPID, Reasoner Model=Qwen3-8B2025.0676.1
Vision-G12026.0376.1
Qwen2.5-VL-7BTraining=DF-GSPO2026.0376.1
PeBR-R1Access Type=Open-Source, Model Scale=7B2025.1276
ETTCAggregation Strategy=ETTC, Model Scale=7B-12B Ensemble2026.0575.93
Qwen2.5-VL-32B w/ RAPIDReasoning Framework=RAPID, Reasoner Model=GPT-OSS-120B-A5B2025.0675.9
InternVL3.5-8B-MPOTraining=MPO2026.0375.9
SRPO-7BMethod Category=Verification-augmented2025.0675.8
Sora-2 AudioInput Modality=Audio, LLM-as-a-Judge=GPT-4o2025.1175.7
VAPO-ThinkerAccess Type=Open-Source, Model Scale=7B2025.1275.6
Qwen2.5-VL-32BTraining=GSPO2026.0375.5
R1-ShareVL-7BEvaluation Source=original paper2026.0375.4
R1-ShareVL-7B2026.0375.4
ThinkLite-VLAccess Type=Open-Source, Model Scale=7B2025.1275.1
InternVL3-14BModel Category=Open-Source2025.0675.1
InternVL3-38BModel Category=Open-Source2025.0675.1
Qwen2.5-VL-72B w/ RAPIDReasoning Framework=RAPID, Reasoner Model=GPT-OSS-120B-A5B2025.0675.1
VL-RethinkerAccess Type=Open-Source, Model Scale=7B2025.1274.9
VL-Rethinker-7BModel Category=Open-Source2025.0674.9
Qwen2.5-VL-72BModel Category=Larger MLLMs without Reasoning2025.1274.8
Qwen-2.5-VL-72B2026.0374.8
Qwen-2.5-VL-72BModel Category=Open-Source General Models2026.0374.8
MM-Eureka-32BModel Category=Open-Source Reasoning Models2026.0374.8
Qwen-2.5-VL-32BEvaluation Source=original paper2026.0374.7
Qwen-2.5-VL-32BModel Category=Open-Source General Models2026.0374.7
MM-Eureka-32BModel Category=Open-Source2025.0674.7
Qwen2.5-VL-32BTraining=Base2026.0374.7
SPOT-EBase Model=InternVL3-8B, Source Type=Open-Source, Test-time Adaptation=SPOT-E2026.0674.4
ThinkLite-7B2026.0374.3
DIVA-GRPO-7B2026.0374.2
Qwen2.5-VL-72BModel Category=Open-Source2025.0674.2
Qwen2.5-VL-72BReasoning Framework=Standard2025.0674.2
ToR-DAPOData Size=39K (ViRL-39K), Base Model=Qwen-2.5-VL-7B2026.0374.2
DR-MMSearchAgent (w/ SPAI)Model Category=Open-source Models2026.0474.2
Qwen2.5-VL-72BModel Category=Open-source Models (>70B)2025.0674.2
o12026.0373.9
o1Model Category=Closed-Source Models2026.0373.9
InternVL2.5-38B-MPO2026.0373.8
InternVL2.5-38B-MPOModel Category=Open-Source Reasoning Models2026.0373.8
Qwen3-VL-4BModel Scale=4B2026.0273.7
VL-RethinkerTool=✗, Param Size=7B2025.1173.7
Ovis2-16BModel Category=Open-Source2025.0673.7
VL-Rethinker-7BModel Category=Open-source Models2026.0473.7
InternVL3-8BModel Category=Open-Source2025.0673.6
Qwen2.5-VL-7BTraining=GSPO2026.0373.6
VisonR1 7BModel Category=Reasoning MLLMs2026.0173.5
Adora-7B2026.0373.5
R1-ShareVL-7BEvaluation Source=re-evaluation2026.0373.5
ADORA-7BModel Category=Open-Source Reasoning Models2026.0373.5
Vision-R1-7BModel Size=7B2026.0373.5
Vision-R1-7BData Size=200K+10K, Base Model=Qwen-2.5-VL-7B2026.0373.5
Vision-R12026.0373.5
OursAccess Type=Open-Source, Model Scale=7B2025.1273.4
MiniCPM-o-2.6-8BModel Category=Open-source Models (~7B)2025.0673.3
SPOT-EBase Model=Qwen3-VL-8B, Source Type=Open-Source, Test-time Adaptation=SPOT-E2026.0673.3
Revisual-R1Access Type=Open-Source, Model Scale=7B2025.1273.1
MM-EurekaAccess Type=Open-Source, Model Scale=7B2025.1273
MathBookAccess Type=Open-Source, Model Scale=7B2025.1273
MM-Eureka-7BEvaluation Source=original paper2026.0373
MM-Eureka-7BModel Category=Open-Source Reasoning Models2026.0373
MM-Eureka-7BModel Category=Open-Source2025.0673
Qwen2.5-VL-7B-Instruct + NoisyRolloutModel Size=7B2026.0372.9
ThinkLite-7B-VLModel Size=7B2026.0372.7
ThinkLite-7B-VLData Size=11K, Base Model=Qwen-2.5-VL-7B2026.0372.7
SFTTraining Stage=SFT, RL Alignment (GRPO)=true2025.0872.7
MM-EurekaTool=✗, Param Size=7B2025.1172.6
NoisyRollout-7BData Size=2.1K, Base Model=Qwen-2.5-VL-7B2026.0372.6
ToR-DAPOData Size=2.1K (Geo3K), Base Model=Qwen-2.5-VL-7B2026.0372.6
Claude Sonnet 4.5Input Modality=Multimodal, LLM-as-a-Judge=GPT-4o2025.1172.5
ThinkLite-VL-7BModel Category=Open-source Models2026.0472.4
OpenVLThinkerAccess Type=Open-Source, Model Scale=7B2025.1272.3
Look-BackAccess Type=Open-Source, Model Scale=7B2025.1272.3
InternVL2.5-VL-78B2026.0372.3
InternVL2.5-VL-78BModel Category=Open-Source General Models2026.0372.3
PSFTTraining Stage=PSFT, RL Alignment (GRPO)=true2025.0872.2
Qwen-7BModel Backbone=Qwen, Model Scale=7B, Aggregation Strategy=Single Model2026.0572.08
GPT-4.1Access Type=Closed-Source2025.1272
InternVL2.5-VL-38B2026.0371.9
InternVL2.5-VL-38BModel Category=Open-Source General Models2026.0371.9