Multi-table Question Answering on TQABench 8k
38.14AccuracyDelimScaling
Evaluation Results
| Method | Links | |
|---|---|---|
| DelimScalingBackbone=Qwen2.5-7B2026.02 | 38.14 | |
| DelimScalingBackbone=Qwen2.5-3B2026.02 | 37.84 | |
| Qwen2.5-7BBackbone=Qwen2.5-7B2026.02 | 37.5 | |
| Qwen2.5-3BBackbone=Qwen2.5-3B2026.02 | 37.38 |