Table Question Answering on TQA-Bench
91.67Accuracyzero-shot
Evaluation results
| Method | Links | ||
|---|---|---|---|
| zero-shotSetting=Tuned LLM (LoRA)2026.06 | 91.67 | — | |
| GRABSetting=Frozen LLM2026.06 | 91.29 | — | |
| GRABSetting=Tuned LLM (LoRA)2026.06 | 90.05 | — | |
| Prompt tuningSetting=Frozen LLM2026.06 | 84.14 | — | |
| MultiTabQASetting=Others2026.06 | 81.19 | — | |
| STRScale=8k, Representation=STR2026.05 | 77.5 | 6,090 | |
| HTMLScale=8k, Representation=HTML2026.05 | 77.21 | 12,010 | |
| STRScale=16k, Representation=STR2026.05 | 72.14 | 11,717 | |
| HTMLScale=16k, Representation=HTML2026.05 | 70.57 | 22,987 | |
| HTMLScale=32k, Representation=HTML2026.05 | 64.64 | 44,562 | |
| STRScale=32k, Representation=STR2026.05 | 64.47 | 23,092 | |
| HTMLScale=64k, Representation=HTML2026.05 | 61.41 | 98,593 | |
| STRScale=64k, Representation=STR2026.05 | 60.71 | 47,466 | |
| GPT-5.4-miniSetting=Others2026.06 | 54.24 | — | |
| zero-shotSetting=Inference-only2026.06 | 38.19 | — |