Graduate-level Question Answering on GPQA (Accuracy & Input Length)
65.7AccuracyRouteGoT
Evaluation results
| Method | Links | ||
|---|---|---|---|
| RouteGoTModel Pool={Qwen3-4B, 8B, 30B}2026.03 | 65.7 | 3,696 | |
| AGoTBackbone=Qwen3-30B, Category=Reasoning Baseline2026.03 | 64.6 | 35,564 | |
| CoTBackbone=Qwen3-30B, Category=Reasoning Baseline2026.03 | 63.1 | 247 | |
| KNNModel Pool={Qwen3-4B, 8B, 30B}, Category=Router Baseline2026.03 | 61.1 | 4,039 | |
| RandomModel Pool={Qwen3-4B, 8B, 30B}, Category=Router Baseline2026.03 | 60.1 | 7,823 | |
| GoT*Backbone=Qwen3-30B, Category=Reasoning Baseline2026.03 | 59.6 | 16,826 | |
| EmbedLLMModel Pool={Qwen3-4B, 8B, 30B}, Category=Router Baseline2026.03 | 59.6 | 15,387 | |
| RouteLLMModel Pool={Qwen3-4B, 8B, 30B}, Category=Router Baseline2026.03 | 57.6 | 4,358 | |
| RTRModel Pool={Qwen3-4B, 8B, 30B}, Category=Router Baseline2026.03 | 56.6 | 3,998 | |
| ToTBackbone=Qwen3-30B, Category=Reasoning Baseline2026.03 | 44.9 | 4,928 | |
| IOBackbone=Qwen3-30B, Category=Reasoning Baseline2026.03 | 41.4 | 250 |