List your company
ResearchFormerly WizWand

Table Question Answering on WikiTQ

91.84AccuracyGPT-5-nano
50.988861.594472.282.8056Jan 16, 2022Oct 10, 2022Jul 5, 2023Mar 28, 2024Dec 21, 2024Sep 14, 2025Jun 9, 2026
Updated 4mo ago

Evaluation results

MethodLinks
GPT-5-nanoWorkflow=Tree-of-Table (Tools), ST Reward=true, SR Reward=RG (TABROUGE)2026.0191.84
GPT-5-nanoWorkflow=Chain-of-Table (Tools), ST Reward=true, SR Reward=RG (TABROUGE)2026.0189.8
CRAFTBackbone=Qwen3.52026.0686.1
GPT-oss-20bWorkflow=Chain-of-Table (Tools), ST Reward=true, SR Reward=RG (TABROUGE)2026.0185.71
GPT-oss-20bWorkflow=Tree-of-Table (Tools), ST Reward=true, SR Reward=RG (TABROUGE)2026.0185.71
Table-CriticBackbone=Qwen3.52026.0685.4
CRAFT + CoTBackbone=Qwen3.52026.0685.2
Formula-R1-PlusBackbone=Qwen 2.5-Coder 7B, Evaluation Protocol=Finetuning-Based2025.0582.54
GPT-4.1-nanoWorkflow=Chain-of-Table (Tools), ST Reward=true, SR Reward=RG (TABROUGE)2026.0181.63
Chain-of-TableBackbone=Qwen3.52026.0680
TableMasterBackbone=GPT-4o-mini, Evaluation Protocol=Prompting-Based2025.0578.13
SS-CoTBackbone=Llama-3.1 70B, Evaluation Protocol=Prompting-Based2025.0576.8
DeepSeek-R1Setting=Others2025.1175.76
TabClawCategory=Ours, Backbone=DeepSeek-v3.22026.0675.2
TIDE DP&AgentBackbone=GPT-3.5, Evaluation Protocol=Prompting-Based2025.0575
TabAFBackbone=Qwen 2.5-Coder 7B, Evaluation Protocol=Finetuning-Based2025.0574.72
Qwen2.5-CoderBase Model=Qwen2.5, Size=7B2025.0574.5
Table-R1 (w/ RE-SFT & TARPO)Base Model=Qwen3, Size=8B2025.0573.87
Norm-DP&AgentBackbone=GPT-3.5, Evaluation Protocol=Prompting-Based2025.0573.65
SheetAgentCategory=Autonomous Agent, Backbone=DeepSeek-v3.22026.0673.2
TableLLM (w/ SFT)Base Model=Qwen3, Size=8B2025.0573.02
BinderCategory=Context Augmentation, Backbone=DeepSeek-v3.22026.0672.5
Chain-of-TableCategory=Context Augmentation, Backbone=DeepSeek-v3.22026.0672.1
TableGPT2Base Model=Qwen2.5, Size=72B2025.0571.45
GPT-4.1-nanoWorkflow=Tree-of-Table (Tools), ST Reward=true, SR Reward=RG (TABROUGE)2026.0171.43
Few-shotBackbone=Qwen3.52026.0671.4
QWQBase Model=Qwen2.5, Size=32B2025.0570.5
Zero-shotBackbone=Qwen3.52026.0670
VD-LB (HQ Correct vs. LQ Wrong)Model=Qwen2.5-7B-VL, Train Set=HiTab2026.0369.9
ReActCategory=Autonomous Agent, Backbone=DeepSeek-v3.22026.0669.8
VD-LB (HQ Correct vs. LQ Wrong)Model=Qwen2.5-7B-VL, Train Set=GQA2026.0369.6
Specialist SOTASetting=Others2025.1169.1
RE-TableLLM (w/ RE-SFT)Base Model=Qwen3, Size=8B2025.0569.06
Qwen2.5-InstructBase Model=Qwen2.5, Size=7B2025.0568.55
GPT-4Setting=Others2025.1168.4
GPT-4o2025.0568.4
VD-LF (HQ vs. LQ)Model=Qwen2.5-7B-VL, Train Set=GQA2026.0368.2
VD-LB (HQ Correct vs. LQ Wrong)Model=Qwen2.5-7B-VL, Train Set=VQA2026.0368.15
GPT-4.1Setting=Others2025.1168.14
ReAcTableBackbone=CodeX, Evaluation Protocol=Prompting-Based2025.0568
Ministral-3-8BWorkflow=Chain-of-Table (Tools), ST Reward=true, SR Reward=RG (TABROUGE)2026.0167.35
CHAIN-OF-TABLELLM Backbone=PaLM 22024.0167.31
Chain-of-TableBackbone=PaLM 2, Evaluation Protocol=Prompting-Based2025.0567.31
VD-LF (HQ vs. LQ)Model=Qwen2.5-7B-VL, Train Set=HiTab2026.0367.15
Formula-R1Backbone=Qwen 2.5-Coder 7B, Evaluation Protocol=Finetuning-Based2025.0567.05
BaselineModel=Qwen2.5-7B-VL, Training=Inference Only2026.0366.4
TableLLMBase Model=CodeLlama, Size=13B2025.0566.3
SpreadsheetLLMCategory=Context Augmentation, Backbone=DeepSeek-v3.22026.0666.1
SFT (HQ Correct Only)Model=Qwen2.5-7B-VL, Train Set=GQA2026.0366.05
SFT (HQ Correct Only)Model=Qwen2.5-7B-VL, Train Set=HiTab2026.0365.9
DaterBackbone=CodeX, Evaluation Protocol=Prompting-Based2025.0565.9
NormTabCategory=Context Augmentation, Backbone=DeepSeek-v3.22026.0665.9
VD-LF (HQ vs. LQ)Model=Qwen2.5-7B-VL, Train Set=VQA2026.0365.4
Table-R1 (w/ RE-SFT & TARPO)Base Model=TeleChat2, Size=7B2025.0565.27
BinderBackbone=CodeX, Evaluation Protocol=Prompting-Based2025.0564.6
TableLLM (w/ SFT)Base Model=Qwen2, Size=7B2025.0564.16
RE-TableLLM (w/ RE-SFT)Base Model=TeleChat2, Size=7B2025.0563.52
Table-R1 (w/ RE-SFT & TARPO)Base Model=Llama3.1, Size=8B2025.0563.51
TableLLM (w/ SFT)Base Model=TeleChat2, Size=7B2025.0563.37
Ministral-3-8BWorkflow=Tree-of-Table (Tools), ST Reward=true, SR Reward=RG (TABROUGE)2026.0163.27
SFT (HQ Correct Only)Model=Qwen2.5-7B-VL, Train Set=VQA2026.0362.85
OmniTabBackbone=BART-Large, Evaluation Protocol=Finetuning-Based2025.0562.8
Table-R1 (w/ RE-SFT & TARPO)Base Model=Qwen2, Size=7B2025.0561.69
DaterLLM Backbone=PaLM 22024.0161.48
TableGPT2Backbone=Qwen 2.5 7B, Evaluation Protocol=Finetuning-Based2025.0561.42
CFMSBackbone=GPT 3.52026.0461.42
TableGPT2Base Model=Qwen2.5, Size=7B2025.0561.42
VanillaCategory=Direct Inference, Backbone=DeepSeek-v3.22026.0661.4
RE-TableLLM (w/ RE-SFT)Base Model=Llama3.1, Size=8B2025.0561.03
End-to-End QALLM Backbone=PaLM 22024.0160.59
Chain-of-ThoughtLLM Backbone=PaLM 22024.0160.43
Few-Shot QALLM Backbone=PaLM 22024.0160.33
CHAIN-OF-TABLELLM Backbone=GPT 3.52024.0159.94
Chain-of-TableBackbone=GPT 3.52026.0459.94
CHAIN-OF-TABLEBase Model=GPT3.52025.0559.94
TAPEX-LargeBackbone=BART-Large, Evaluation Protocol=Finetuning-Based2025.0559.1
QWEN3-8BWorkflow=Chain-of-Table (Tools), ST Reward=true, SR Reward=AS2026.0158.16
RE-TableLLM (w/ RE-SFT)Base Model=Qwen2, Size=7B2025.0557.55
Liu et al. (2021)Extra Pre-training=true2022.0157.5
QWEN3-8BWorkflow=Tree-of-Table (Tools), ST Reward=true, SR Reward=AS2026.0157.14
TAMOSetting=Frozen LLM, Backbone=Llama 3.1 8B2025.1156.93
BinderLLM Backbone=GPT 3.52024.0156.74
BinderBackbone=GPT 3.52026.0456.74
Table-R1 (w/ RE-SFT & TARPO)Base Model=Qwen2.5, Size=3B2025.0556.51
QWEN3-8BWorkflow=Chain-of-Table (Tools), ST Reward=true, SR Reward=RG (TABROUGE)2026.0155.61
TableLLM (w/ SFT)Base Model=Qwen2.5, Size=3B2025.0555.5
QWEN3-8BWorkflow=Tree-of-Table (Tools), ST Reward=true, SR Reward=RG (TABROUGE)2026.0155.1
BinderLLM Backbone=PaLM 22024.0154.88
Prompt tuningSetting=Frozen LLM, Backbone=Llama 3.1 8B2025.1153.71
TableLLMBackbone=Qwen 7B, Evaluation Protocol=Finetuning-Based2025.0553.59
Chain-of-ThoughtLLM Backbone=GPT 3.52024.0153.48
Chain-of-ThoughtBackbone=GPT 3.52026.0453.48
GPT-3.5Setting=Others2025.1153.13
QWEN3-8BWorkflow=Tree-of-Table (Tools), ST Reward=true, SR Reward=AC2026.0153.06
Text-to-SQLLLM Backbone=GPT 3.52024.0152.9
Text-to-SQLBackbone=GPT 3.52026.0452.9
DaterLLM Backbone=GPT 3.52024.0152.81
DaterBackbone=GPT 3.52026.0452.81
Few-Shot QALLM Backbone=GPT 3.52024.0152.56
Few-Shot QABackbone=GPT 3.52026.0452.56