Table Question Answering on HiTab (Acc., Avg.) benchmark leaderboard · SOTA2 Research

Table Question Answering on HiTab (Acc., Avg.)

64.71AccuracyTableLlama + Oracle

Evaluation Results

Method	Links
TableLlama + Oracle2025.02		64.71	55.33
MiniCPM-V-2.6 + HIPPO2025.02		63	65.46
MiniCPM-V-2.6 + Vanilla SFT2025.02		62.88	61.93
Qwen2.5-VL-Instruct2025.02		61.99	58.58
Qwen2.5-VL-Instruct + HIPPO2025.02		60.97	62.9
Qwen2.5-VL-Instruct + Vanilla SFT2025.02		59.49	62.15
Qwen2.5-Instruct2025.02		58.69	61.54
MiniCPM-V-2.62025.02		58.56	62.12
Qwen2.5-VL-Instruct (Image Only)2025.02		56.82	56.85
MiniCPM-V-2.6 (Image Only)2025.02		56.53	60.42
GPT-4o2025.02		56.02	55.58
GPT-4o (Image Only)2025.02		54.82	54.26
Table-LLaVA + HIPPO2025.02		47.01	55.52
TableLlama + Markdown2025.02		46.57	37.86
Table-LLaVA2025.02		44.79	50.98
Qwen2-72B-Instruct2025.02		39.21	57.11
Table-LLaVA (Image Only)2025.02		10.85	38.09
Llama22025.02		10.72	17.53
Table-LLaVA (Image Only)2025.02		10.09	35.69
Llama3-Instruct2025.02		6.97	31.98
Monkey2025.02		6.41	14.16
TabPedia2025.02		5.58	20.84
Vary-toy2025.02		3.42	5.77
LLaVA v1.52025.02		2.03	9.82
mPLUG-Owl2025.02		0.25	3.16
MiniGPT-42025.02		0.2	0.28
mPLUG-Owl22025.02		0.13	7.76
InternLM-XComposer2025.02		0.12	0.77
Llama3.1-Instruct2025.02		0.08	35.42
Qwen-VL2025.02		0.06	0.83