List your company

Chart Question Answering on ChartQAPro 67% held-out (test)

46.88Factoid AccuracyCOGS
6.559217.027127.49537.9629Oct 16, 2025

Evaluation results

MethodLinks
COGSBase Model=Qwen2.5-VL-7B, Data Synthesis Approach=COGS2025.1046.8865.7351.1661.8558.2552.02
In-Context Q ExampleBase Model=Qwen2.5-VL-7B, Data Synthesis Approach=In-Context Q Example2025.1046.3362.9446.9161.1161.7250.58
GPT-5-nanoModel Category=Proprietary Models2025.1045.9563.6449.463.5849.8250.74
GPT-4o-miniModel Category=Proprietary Models2025.1043.6366.4345.4859.8845.248.32
Claude Haiku 3.5Model Category=Proprietary Models2025.1043.4465.0339.8461.7938.7746.74
Self-ConsistencyBase Model=Qwen2.5-VL-7B, Prompting Strategy=Self-Consistency2025.1043.4461.544459.8241.7647.22
InternVL3.5-GPT-OSSModel Category=Opensource Models (7B+)2025.1043.0258.7442.8658.0254.4846.86
Chart-R1Base Model=Qwen2.5-VL-7B, Data Synthesis Approach=Chart-R12025.1042.1746.8550.5361.1155.5547.32
Decompositional CoTBase Model=Qwen2.5-VL-7B, Prompting Strategy=Decompositional CoT2025.1042.0865.0342.5756.5345.5546.36
Qwen2.5-VL-7B (base)Model Category=Opensource Models (7B+)2025.1042.0762.5944.8860.7850.7247.36
Gemini 2.5 Flash-LiteModel Category=Proprietary Models2025.1040.4219.9648.7737.4316.6638.72
Tree of ThoughtsBase Model=Qwen2.5-VL-7B, Prompting Strategy=Tree of Thoughts2025.1040.0157.9441.5554.1353.3544.44
ChartQA-TrainBase Model=Qwen2.5-VL-7B, Data Synthesis Approach=ChartQA-Train2025.1038.7760.1449.7261.1153.1246.64
Phi-4-14BModel Category=Opensource Models (7B+)2025.1023.1834.2740.9346.9136.3131.61
ChartMoEModel Category=Chart Specialist Models2025.1019.0335.6632.9745.6827.0827.28
ChartLLaMAModel Category=Chart Specialist Models2025.108.1123.0818.3745.0629.5517.19