List your company
ResearchFormerly WizWand

Question Answering on TAT-QA (held-out)

59.33AccuracyClaude-3.5-Sonnet
-1.967613.946229.8645.7738Sep 21, 2025
Updated 7mo ago

Evaluation results

MethodLinks
Claude-3.5-SonnetResolution=UNK2025.0959.33
Gemini-2.5-ProResolution=UNK2025.0956.29
OpenAI-o4-miniResolution=UNK2025.0956.16
QVQ-72B-PreviewResolution=Dyn.2025.0955.48
GPT-4oResolution=UNK2025.0953.85
Qwen2-VL-72BResolution=Dyn.2025.0952.23
Qwen2-VL-7B-GRPOResolution=Dyn.2025.0951.68
Qwen2-VL-7B-Table-R1Resolution=Dyn.2025.0948.06
Qwen2-VL-7BResolution=Dyn.2025.0946.5
Qwen2-VL-7B-SFTResolution=Dyn.2025.0937.82
Qwen2-VL-2BResolution=Dyn.2025.0930.44
Qwen2-VL-2B-GRPOResolution=Dyn.2025.0929.02
Qwen2-VL-2B-Table-R1Resolution=Dyn.2025.0926.42
DeepSeek-VL2 4.5BResolution=Dyn.2025.0924.11
Qwen2-VL-7B-InsResolution=Dyn2025.0920.85
Qwen2-VL-2B-SFTResolution=Dyn.2025.0919.69
Table-LLaVA 13BResolution=336, Trained on dataset=true2025.0915.67
Table-LLaVA 7BResolution=336, Trained on dataset=true2025.0912.82
Monkey 7BResolution=8962025.0912.31
LLaVA-v1.5 7BResolution=3362025.092.97
mPLUG-Owl2 7BResolution=4482025.090.39