Table Question Answering on TableBench (test)
0.4887Overall AccuracyLlama-3.1-405B-Instruct + TCoT
Evaluation results
| Method | Links | |
|---|---|---|
| Llama-3.1-405B-Instruct + TCoT# Params.=405B2026.05 | 0.4887 | |
| Qwen2.5-72B-Instruct + TCoT# Params.=72B2026.05 | 0.4879 | |
| Qwen3-8B + TGN# Params.=8B2026.05 | 0.4846 | |
| Llama-4-Scout-17B-16E-Instruct + TCoT# Params.=17B2026.05 | 0.4653 | |
| Qwen2.5-Coder-32B-Instruct + TCoT# Params.=32B2026.05 | 0.4551 | |
| QWQ-32B + DP# Params.=32B2026.05 | 0.4387 | |
| Llama3.1-70B-Instruct + TCoT# Params.=70B2026.05 | 0.4105 | |
| Llama3-70B-Chat + TCoT# Params.=70B2026.05 | 0.3868 | |
| DeepSeek-R1-Distill-Llama-8B + PIP# Params.=8B2026.05 | 0.3748 | |
| GPT-3.5-Turbo + PoT2026.05 | 0.3715 |