Question Answering over Tables and Text on HybridQA
91AccuracyToT
Evaluation results
| Method | Links | ||
|---|---|---|---|
| ToTBackbone=Qwen3-30B2026.03 | 91 | 12,355 | |
| RouteGoTModel Pool={Qwen3-4B, 8B, 30B}2026.03 | 91 | 14,743 | |
| GoT*Backbone=Qwen3-30B2026.03 | 88 | 65,134 | |
| CoTBackbone=Qwen3-30B2026.03 | 84 | 10,239 | |
| AGoTBackbone=Qwen3-30B2026.03 | 84 | 112,953 | |
| RandomModel Pool={Qwen3-4B, 8B, 30B}2026.03 | 70 | 21,138 | |
| KNNModel Pool={Qwen3-4B, 8B, 30B}2026.03 | 68 | 20,686 | |
| RTRModel Pool={Qwen3-4B, 8B, 30B}2026.03 | 68 | 20,839 | |
| IOBackbone=Qwen3-30B2026.03 | 67 | 10,462 | |
| EmbedLLMModel Pool={Qwen3-4B, 8B, 30B}2026.03 | 66 | 21,875 | |
| RouteLLMModel Pool={Qwen3-4B, 8B, 30B}2026.03 | 65 | 20,652 |