Open-ended Question Answering on HybridQA (test)
91AccuracyToT
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| ToTBackbone=Qwen3-30B2026.03 | 91 | 3,538 | |
| RouteGoTBackbone={Qwen3-4B, 8B, 30B}2026.03 | 91 | 700 | |
| GoT*Backbone=Qwen3-30B2026.03 | 88 | 1,635 | |
| CoTBackbone=Qwen3-30B2026.03 | 84 | 476 | |
| AGoTBackbone=Qwen3-30B2026.03 | 84 | 2,097 | |
| RandomBackbone={Qwen3-4B, 8B, 30B}2026.03 | 70 | 930 | |
| KNNBackbone={Qwen3-4B, 8B, 30B}2026.03 | 68 | 510 | |
| RTRBackbone={Qwen3-4B, 8B, 30B}2026.03 | 68 | 527 | |
| IOBackbone=Qwen3-30B2026.03 | 67 | 21 | |
| EmbedLLMBackbone={Qwen3-4B, 8B, 30B}2026.03 | 66 | 1,376 | |
| RouteLLMBackbone={Qwen3-4B, 8B, 30B}2026.03 | 65 | 540 |