Share your thoughts, 1 month free Claude Pro on usSee more

Numerical Reasoning on TableBench (test)

64.48Accuracy

Qwen3-8B

Updated 13d ago

Evaluation Results

Method	Links
Qwen3-8B 2026.05		64.48
Qwen3-4B 2026.05		61.46
DeepSeek-R1-Distill-Llama-8B 2026.05		50.13
Qwen3-1.7B 2026.05		43.83
TableLLM-Qwen2-7B 2026.05		38.54
Qwen2.5-Coder-7B-Instruct 2026.05		37.03
TableGPT2-7B 2026.05		31.23
Qwen2.5-7B-Instruct 2026.05		25.69
Qwen2-7B-Instruct 2026.05		24.69
Meta-Llama-3-8B-Instruct 2026.05		21.91
Llama-3.1-8B-Instruct 2026.05		13.85
Llama-3.2-3B-Instruct 2026.05		4.03
Llama-3.2-3B-Instruct 2026.05		3.78