Explorative Reasoning on Crosswords Word-level (test)
13.5AccuracyCoT
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CoTBackbone=Qwen3-30B2026.03 | 13.5 | 13,201 | |
| RouteGoTBackbone={Qwen3-4B, 8B, 30B}2026.03 | 11 | 3,804 | |
| GoT*Backbone=Qwen3-30B2026.03 | 6.5 | 11,674 | |
| IOBackbone=Qwen3-30B2026.03 | 6 | 467 | |
| EmbedLLMBackbone={Qwen3-4B, 8B, 30B}2026.03 | 5 | 6,017 | |
| AGoTBackbone=Qwen3-30B2026.03 | 3.5 | 20,893 | |
| RouteLLMBackbone={Qwen3-4B, 8B, 30B}2026.03 | 2.5 | 5,831 | |
| ToTBackbone=Qwen3-30B2026.03 | 2.1 | 9,563 | |
| RTRBackbone={Qwen3-4B, 8B, 30B}2026.03 | 1.5 | 4,065 | |
| RandomBackbone={Qwen3-4B, 8B, 30B}2026.03 | 1 | 6,635 | |
| KNNBackbone={Qwen3-4B, 8B, 30B}2026.03 | 0.5 | 5,813 |