List your company
ResearchFormerly WizWand

Chart Understanding on ChartQA (test)

90.8AccuracyClaude-3.5 Sonnet
69.209674.814880.4286.0252Oct 8, 2024Jan 19, 2025May 3, 2025Aug 15, 2025Nov 27, 2025Mar 11, 2026Jun 23, 2026
Updated 4mo ago

Evaluation results

MethodLinks
Claude-3.5 Sonnet2025.0290.8
Claude-3.5-Sonnet2024.1290.8
iVGR-Qwen3-VL-32BBackbone=Qwen3-VL, Model Size=32B, Training Method=iVGR2026.0590.4
Qwen2.5-VL-72BModel Size=72B2025.0289.5
OptMerge2025.0588.92
iVGR-Qwen2.5-VL-7BBackbone=Qwen2.5-VL, Model Size=7B, Training Method=iVGR2026.0588.5
InternVL2-76BModel Scale=76B2024.1288.4
InternVL2.5-78BModel Size=78B2025.0288.3
Qwen2-VL-72BModel Scale=72B2024.1288.3
iVGR-Qwen3-VL-8BBackbone=Qwen3-VL, Model Size=8B, Training Method=iVGR2026.0587.6
AndesVLNumber of Parameters=2.4B2025.1287.4
Qwen2.5-VL-7BModel Size=7B2025.0287.3
Gemini-1.5 Pro2024.1087.2
Gemini 1.5 Pro2025.0287.2
Gemini-1.5-Pro2024.1287.2
Individual OCR2025.0587
Individual Chart2025.0586.96
Individual Geometry2025.0586.76
GPT-4o2025.0286.7
Qwen2.5-VL-7B + Cont. Rewardreward_type=Continuous2025.1186.7
InternVL3.5-8BParameter Scale=8B2026.0586.7
Individual VQA2025.0586.68
ARIA2024.1086.4
Qwen2.5-VL-7BBackbone=Qwen2.5-VL, Model Size=7B, Training Method=Baseline2026.0586.4
InternVL3.5-4BParameter Scale=4B2026.0586.4
MAmmoTH-VL-8BModel Scale=8B2024.1286.2
Molmo2-4BParameter Scale=4B2026.0586.1
PerceptionLM-8BParameter Scale=8B2026.0586
Molmo2-8BParameter Scale=8B2026.0586
MAmmoTH-VL-8BModel Scale=8B, Single-image variant (SI)=true2024.1285.9
Individual Grounding2025.0585.88
GPT-4o2024.1085.7
GPT-4o2024.1285.7
HyperVL ViTLNumber of Parameters=2.0B2025.1285.7
Gemini-1.5 Flash2024.1085.4
SAIL-VL2Number of Parameters=2.7B2025.1285.4
Zamba2-VL-7BParameter Scale=7B2026.0585.3
PerceptionLM-3BParameter Scale=3B2026.0585.1
Qwen3-VL-32BBackbone=Qwen3-VL, Model Size=32B, Training Method=Baseline2026.0585
LLaVA-OV-72BModel Scale=72B, Single-image variant (SI)=true2024.1284.9
SAIL-VL1.5Number of Parameters=2.5B2025.1284.8
Qwen2.5-VL-7B + Discrete Rewardreward_type=Discrete2025.1184.62
Vision-Zeroexternal supervision=true2025.1184.24
Molmo-7B-DModel Scale=7B2024.1284.1
Qwen2.5-VL-3BModel Size=3B2025.0284
InternVL2.5-4BComplexity=O(n)2025.1284
Qwen2.5VL-3BComplexity=O(n)2025.1284
Qwen2.5-VL-7B (Baseline)2025.1184
HyperVLNumber of Parameters=1.8B2025.1283.8
LLaVA-OV-72BModel Scale=72B2024.1283.7
Qwen2.5-VL-32B-Instruct2025.0583.64
Llama3.2-11B2024.1083.4
Llama-3.2-11B-Vision-InsModel Scale=11B2024.1283.4
InternVL-2-8BModel Scale=8B2024.1283.3
Qwen3-VL-8BBackbone=Qwen3-VL, Model Size=8B, Training Method=Baseline2026.0583.2
Qwen2-VL-7B-InsModel Scale=7B2024.1283
Qwen2-VL-7BParameters=7B2026.0683
Qwen3-VL-8BParameter Scale=8B2026.0582.8
MiniCPM-V-2.6-7BModel Scale=7B2024.1282.4
INXComp-2.5-7BModel Scale=2.5B2024.1282.2
InfiniteVL-4BComplexity=O(1)2025.1282
Pixtral-12B2024.1081.8
Phi-3.5-Vision-4BComplexity=O(n)2025.1281.8
Qwen3-VL-4BParameter Scale=4B2026.0581.8
InternVL3.5-2BParameter Scale=2B2026.0581.6
Qwen3VL-8BParameters=8B2026.0681.2
PreciseDoc-Reasoner2026.0681.2
InternVL3.5Number of Parameters=2.3B2025.1280.7
InternVL3Number of Parameters=2.1B2025.1280.2
LLaVA-OV-7BModel Scale=7B2024.1280
Zamba2-VL-2.7BParameter Scale=2.7B2026.0579.6
Ovis2Number of Parameters=2.5B2025.1279.5
PerceptionLM-1BParameter Scale=1B2026.0579.2
LLaVA-OV-7BModel Scale=7B, Single-image variant (SI)=true2024.1278.8
Qwen3-VL-2BParameter Scale=2B2026.0578.7
GPT-4V2024.1078.4
InternVL3.5-1BParameter Scale=1B2026.0578
Qwen3-VLNumber of Parameters=2.1B2025.1277.6
Zamba2-VL-1.2BParameter Scale=1.2B2026.0577.6
Iso-CMethod Class=Merging2025.0576.96
TA w/ DAREMethod Class=Merging2025.0576.48
TSV MergingMethod Class=Merging2025.0576.44
OptMergeMethod Class=Merging2025.0576.24
TIES w/ DAREMethod Class=Merging2025.0576.04
TIES MergingMethod Class=Merging2025.0575.48
Task ArithmeticMethod Class=Merging2025.0575.28
WUDI MergingMethod Class=Merging2025.0574.36
Qwen2-VLNumber of Parameters=2.2B2025.1273.5
Cambrian-1-8BModel Scale=8B2024.1273.3
TA w/ DAREProtocol=full fine-tuning2025.0573.08
Task ArithmeticProtocol=full fine-tuning2025.0572.88
TIES w/ DAREProtocol=full fine-tuning2025.0572.72
TIES MergingProtocol=full fine-tuning2025.0572.28
Weight AverageProtocol=full fine-tuning2025.0571.64
Qwen3.5-9BParameters=9B2026.0670.7
TSV MergingProtocol=full fine-tuning2025.0570.56
Iso-CProtocol=full fine-tuning2025.0570.56
Mixture TrainingProtocol=full fine-tuning2025.0570.32
DocOwl-1.5-Chat2026.0670.2
Qwen2-VL-InstructMethod Class=Baseline2025.0570.04