List your company
ResearchFormerly WizWand

Table Question Answering on WikiSQL

92.07AccuracySpecialist SOTA
18.614837.684956.75575.8251May 18, 2025Jul 24, 2025Sep 30, 2025Dec 6, 2025Feb 12, 2026Apr 20, 2026Jun 27, 2026
Updated 4mo ago

Evaluation results

MethodLinks
Specialist SOTASetting=Others2025.1192.07——
GRABSetting=Tuned LLM (LoRA)2026.0690.23——
zero-shotSetting=Tuned LLM (LoRA)2026.0690.17——
GRABSetting=Frozen LLM2026.0688.3——
TableLLamaSetting=Others2026.0687.37——
TAMOSetting=Frozen LLM2026.0686.29——
TAMO+Setting=Tuned LLM (SFT)2025.1185.9——
Prompt tuningSetting=Frozen LLM2026.0685.71——
TAMOSetting=Frozen LLM, Backbone=Llama 3.1 8B2025.1185.44——
TAMO+Setting=Tuned LLM (LoRA)2025.1184.43——
SFTSetting=Tuned LLM (SFT)2025.1179.86——
Prompt tuningSetting=Frozen LLM, Backbone=Llama 3.1 8B2025.1177.06——
TAMOSetting=Frozen LLM2025.1176.45——
GPT-5.4-miniSetting=Others2026.0672.99——
Table-R1 (w/ RE-SFT & TARPO)Base Model=Qwen3, Size=8B2025.0572.5——
DeepSeek-R1Setting=Others2025.1171.91——
GPT-4.1Setting=Others2025.1171.21——
TableLLM (w/ SFT)Base Model=Qwen3, Size=8B2025.0571.14——
RE-TableLLM (w/ RE-SFT)Base Model=Qwen3, Size=8B2025.0569.57——
Table-R1 (w/ RE-SFT & TARPO)Base Model=TeleChat2, Size=7B2025.0566.12——
TableLLM (w/ SFT)Base Model=TeleChat2, Size=7B2025.0565.58——
RE-TableLLM (w/ RE-SFT)Base Model=TeleChat2, Size=7B2025.0564.93——
Table-R1 (w/ RE-SFT & TARPO)Base Model=Qwen2, Size=7B2025.0564.8——
RE-TableLLM (w/ RE-SFT)Base Model=Qwen2, Size=7B2025.0564.42——
Table-R1 (w/ RE-SFT & TARPO)Base Model=Llama3.1, Size=8B2025.0564.3——
Table-R1 (w/ RE-SFT & TARPO)Base Model=Qwen2.5, Size=3B2025.0563.98——
RE-TableLLM (w/ RE-SFT)Base Model=Llama3.1, Size=8B2025.0562.92——
RE-TableLLM (w/ RE-SFT)Base Model=Qwen2.5, Size=3B2025.0561.55——
TableLLM (w/ SFT)Base Model=Qwen2.5, Size=3B2025.0561.46——
Prompt tuningSetting=Frozen LLM2025.1161.24——
Table-R1 (w/ RE-SFT & TARPO)Base Model=DS-Coder, Size=7B2025.0560.45——
Table-R1 (w/ RE-SFT & TARPO)Base Model=CodeQwen, Size=7B2025.0559.89——
RE-TableLLM (w/ RE-SFT)Base Model=DS-Coder, Size=7B2025.0558.86——
RE-TableLLM (w/ RE-SFT)Base Model=CodeQwen, Size=7B2025.0557.35——
TableGPT2Base Model=Qwen2.5, Size=72B2025.0557.32——
LoRASetting=Tuned LLM (LoRA)2025.1157.1——
zero-shotSetting=Inference-only2026.0653.86——
TableGPT2Base Model=Qwen2.5, Size=7B2025.0553.74——
QWQBase Model=Qwen2.5, Size=32B2025.0547.67——
GPT-4Setting=Others2025.1147.6——
GPT-4o2025.0547.6——
Qwen2.5-InstructBase Model=Qwen2.5, Size=7B2025.0547.42——
Qwen2.5-CoderBase Model=Qwen2.5, Size=7B2025.0546.9——
TableLLM (w/ SFT)Base Model=Qwen2, Size=7B2025.0546.5——
TableLlamaSetting=Tuned LLM (SFT)2025.1146.26——
CHAIN-OF-TABLEBase Model=GPT3.52025.0543.72——
GPT-3.5Setting=Others2025.1141.91——
TableLLMBase Model=CodeLlama, Size=13B2025.0541.1——
TableLLM (w/ SFT)Base Model=Llama3.1, Size=8B2025.0539——
DS-liteBase Model=DS-Coder, Size=16B2025.0538.3——
TableLLM (w/ SFT)Base Model=CodeQwen, Size=7B2025.0537.2——
TableLLM (w/ SFT)Base Model=DS-Coder, Size=7B2025.0536.14——
Llama 3.1 8BSetting=Inference Only2025.1131.6——
Yi-CoderBase Model=Yi, Size=9B2025.0525.34——
Zero-shotSetting=Inference Only2025.1121.44——
BARTMode=Text Baseline2023.05—85.8—
CRAFTQA2026.06——86.1
Dublinfixed_resResolution=fixed2023.05—75.3—
Dublinvariable_resResolution=variable2023.05—75.3—
PoT2026.06——44.42
Readi2026.06——64.7
StructGPT2026.06——57.9
TAPEXPipeline=specialized2023.05—89.2—
TrustUQA2026.06——85.7