List your company
ResearchFormerly WizWand

Visual Reasoning on BLINK

85.2AccuracyQwen3-VL-8B-Inst.
44.32854.93965.5576.161Sep 29, 2025Nov 14, 2025Dec 30, 2025Feb 14, 2026Apr 1, 2026May 17, 2026Jul 3, 2026
Updated 3mo ago

Evaluation results

MethodLinks
Qwen3-VL-8B-Inst.Model Type=Open-source general-purpose MLLM2026.0385.2
Vlaser-8BModel Type=Embodied Brain MLLM, Evaluation Protocol=Our evaluation framework2026.0384.9
RoboBrain2.5-8BModel Type=Embodied Brain MLLM, Evaluation Protocol=Our evaluation framework2026.0384.3
InternVL3.5-8BModel Type=Open-source general-purpose MLLM, Evaluation Protocol=Our evaluation framework2026.0384.1
ACE-Brain-0-8BModel Type=Embodied Brain MLLM2026.0383.9
InternVL3-8BModel Type=Open-source general-purpose MLLM, Evaluation Protocol=Our evaluation framework2026.0382.9
Qwen2.5-VL-7B-Inst.Model Type=Open-source general-purpose MLLM, Evaluation Protocol=Our evaluation framework2026.0382.5
Gemini-2.5-ProModel Type=Closed-source MLLM2026.0381.8
RoboBrain2.0-7BModel Type=Embodied Brain MLLM, Evaluation Protocol=Our evaluation framework2026.0381.4
GPT-5-miniGrounded=false2025.1281
VeBrain-7BModel Type=Embodied Brain MLLM2026.0379.7
Claude-4-SonnetModel Type=Closed-source MLLM2026.0378.1
GPT-4oModel Type=Closed-source MLLM2026.0377.9
GPT-5-miniGrounded=true2025.1276
Qwen3-VL-2B-Inst.Model Type=Open-source general-purpose MLLM2026.0374.9
Gemini-2.5-ProData Scale=-, Training Phases=-2026.0170.6
GRITBase Model=Qwen2.5-VL-3B2025.1270.3
VALOR2025.1269.2
VALORBase Model=Qwen3-8B2025.1269.2
VALORModel Category=Open-Source, Tool Use=true2025.1269.2
ViGoRLBase Model=Qwen2.5-VL-7B2025.1268.4
GPT-4oReasoning Paradigm=Zero-Shot VLMs2026.0168
GPT-4omodel_category=Proprietary Model2025.0968
VALOR-RLBase Model=Qwen3-8B2025.1267.3
VALOR-RLModel Category=Open-Source, Tool Use=true2025.1267.3
Qwen3-VLScale=4B2026.0566.6
o4-miniModel Category=Proprietary, Tool Use=true2025.1266.5
GPT-4oData Scale=-, Training Phases=-2026.0165.9
Claude-3.5-HaikuModel Category=Proprietary, Tool Use=true2025.1264.6
GPT-4oModel Category=Proprietary, Tool Use=true2025.1264.2
Qwen3-8BModel Category=Open-Source, Tool Use=true2025.1263.9
GPT-4oModel Type=Proprietary2026.0563.55
Gemini2.0-Flashmodel_category=Proprietary Model2025.0963.5
Qwen3.5-VLScale=4B2026.0563.5
SmoothOp-7BData Scale=50K, Training Phases=1 (R)2026.0162.6
VST-7B-SFTData Scale=4.1M, Training Phases=1 (S)2026.0162.1
Qwen3.5-VLScale=2B2026.0562
Gemini-2.5-FlashModel Category=Proprietary, Tool Use=true2025.1261.5
RIS+VLPOBackbone=Qwen2.5-VL-7B, Latent Tokens=5, Optimization=Visual-latent Policy Optimization (VLPO)2026.0560.95
RISBackbone=Qwen2.5-VL-7B, Latent Tokens=52026.0560.6
VST-3B-SFTData Scale=4.1M, Training Phases=1 (S)2026.0159.1
InternVL3.5Scale=4B2026.0557.8
CoVTReasoning Paradigm=Latent-Space Reasoning, Reasoning Space=Latent2026.0757.49
ProLaViT (Base)Reasoning Paradigm=Progressive Latent Visual Thought, Reasoning Space=Latent2026.0757.49
Gemma-3-12BModel Category=Open-Source, Tool Use=true2025.1257.4
CoVTBackbone=Qwen2.5-VL-7B, Latent Tokens=52026.0557.4
ProLaViT (Full)Reasoning Paradigm=Progressive Latent Visual Thought, Reasoning Space=Latent, Training Strategy=Distance-Weighted Diversity Loss2026.0757.23
VisionZero-Qwen-7B (Real-World)training_context=Real-World, base_model=Qwen2.5-VL-7B2025.0957.2
LaserReasoning Paradigm=Latent Reasoning2026.0156.92
SpatialLadder-3BData Scale=26K, Training Phases=3 (S+R)2026.0156.9
SmoothOp-3BData Scale=50K, Training Phases=1 (R)2026.0156.9
Pelican-VL-7BModel Type=Embodied Brain MLLM2026.0356.8
VisionZero-Qwen-7B (Chart)training_context=Chart, base_model=Qwen2.5-VL-7B2025.0956.8
LVRBackbone=Qwen2.5-VL-7B, Latent Tokens=52026.0556.79
MonetBackbone=Qwen2.5-VL-7B, Latent Tokens=52026.0556.7
CoT SFTReasoning Paradigm=Text-Space Reasoning, Reasoning Space=Text2026.0756.54
Qwen2.5-VL-7BData Scale=-, Training Phases=-2026.0156.4
ProLaViT + L_div^marginReasoning Paradigm=Progressive Latent Visual Thought, Reasoning Space=Latent2026.0756.33
One-step Latent Pred.Reasoning Paradigm=Latent-Space Reasoning, Reasoning Space=Latent2026.0756.28
Qwen2.5-VL-7B+GLSDBackbone=Qwen2.5-VL-7B, Training=Grounded Latent Supervision Dataset (GLSD)2026.0556.25
ViLaSR-7BData Scale=81K, Training Phases=2 (S+R)2026.0156.2
VisionZero-Qwen-7B (CLEVR)training_context=CLEVR, base_model=Qwen2.5-VL-7B2025.0956
Qwen3-VLScale=2B2026.0555.7
ViGaL-Snake+Rotationbase_model=Qwen2.5-VL-7B2025.0955.6
SFTReasoning Paradigm=Text-Space Reasoning, Reasoning Space=Text2026.0755.6
VL-RethinkerReasoning Paradigm=Tool-use & RL Enhanced Reasoning2026.0155.55
InternVL3-8BData Scale=-, Training Phases=-2026.0155.5
SpaceR-7BData Scale=151K, Training Phases=2 (S+R)2026.0155.4
Qwen2.5-VL-7Bbase_model=Qwen2.5-VL-7B2025.0955.2
Qwen2.5-VL-7BBackbone=Qwen2.5-VL-7B2026.0554.94
InternVL3.5-8BReasoning Paradigm=Zero-Shot VLMs2026.0154.81
Qwen2.5-VL-InstructReasoning Paradigm=Base Model, Reasoning Space=N/A2026.0754.49
Llama-3.2-11BModel Category=Open-Source, Tool Use=true2025.1254.3
MM-Eureka-Qwen-7Bbase_model=Qwen2.5-VL-7B2025.0954
Qwen2.5-VL-7BReasoning Paradigm=Zero-Shot VLMs2026.0153.6
LVRReasoning Paradigm=Latent Reasoning2026.0153.6
LVRReasoning Paradigm=Latent-Space Reasoning, Reasoning Space=Latent2026.0753.02
VLAA-Thinker-7Bbase_model=Qwen2.5-VL-7B2025.0953
Vision-R1Reasoning Paradigm=Tool-use & RL Enhanced Reasoning2026.0152.71
LLaVALLM=Qwen2-7B2025.1252.7
PAPOReasoning Paradigm=Tool-use & RL Enhanced Reasoning2026.0152.66
InternVL3.5Scale=2B2026.0552.6
Gemini-2.0-FlashModel Category=Proprietary, Tool Use=true2025.1251.7
VISTA-SFTTraining Dataset=SLAKE, Model=Qwen2.5-3B-VL, Training Protocol=single-round self-improvement SFT2026.0551.23
DeepEyesReasoning Paradigm=Tool-use & RL Enhanced Reasoning2026.0151.08
MonetReasoning Paradigm=Latent Reasoning2026.0150.71
OpenVLThinker-7Bbase_model=Qwen2.5-VL-7B2025.0949.9
STaRTraining Dataset=SLAKE, Model=Qwen2.5-3B-VL, Training Protocol=single-round self-improvement SFT2026.0549.78
Ours w/o MaskingLLM=Qwen2-7B2025.1249.6
SFT-SeedTraining Dataset=SLAKE, Model=Qwen2.5-3B-VL, Training Protocol=single-round self-improvement SFT2026.0549.35
LLaVA-OneVisionReasoning Paradigm=Zero-Shot VLMs2026.0149.34
VIRALLLM=Qwen2-7B2025.1249.3
JARVISLLM=Qwen2-7B2025.1249.3
R1-OneVision-7Bbase_model=Qwen2.5-VL-7B2025.0948.7
JARVISLLM=Vicuna-7B2025.1248.3
LLava-OneVision-7BData Scale=-, Training Phases=-2026.0148.2
ReSTEMTraining Dataset=SLAKE, Model=Qwen2.5-3B-VL, Training Protocol=single-round self-improvement SFT2026.0547.62
Qwen2.5-VL-3BData Scale=-, Training Phases=-2026.0147.6
LLaVALLM=Vicuna-7B2025.1246.8
DatologyScale=4B2026.0545.9