Question Answering on OfficeQA 246 questions
80.1AccuracySkillSmith
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SkillSmithBackbone=Qwen3.5-397B2026.05 | 80.1 | 197 | |
| EvoSkillBackbone=Qwen3.5-397B2026.05 | 71.5 | 176 | |
| Base AgentBackbone=Qwen3.5-397B2026.05 | 61.8 | 152 | |
| SkillSmithBackbone=Qwen3.5-122B2026.05 | 60.2 | 148 | |
| EvoSkillBackbone=Qwen3.5-122B2026.05 | 53.3 | 131 | |
| Base AgentBackbone=Qwen3.5-122B2026.05 | 43.9 | 108 | |
| SkillSmithBackbone=Qwen3.5-35B2026.05 | 38.6 | 95 | |
| EvoSkillBackbone=Qwen3.5-35B2026.05 | 34.1 | 84 | |
| Base AgentBackbone=Qwen3.5-35B2026.05 | 28.9 | 71 | |
| SkillSmithBackbone=Qwen3.5-27B2026.05 | 28.5 | 70 | |
| EvoSkillBackbone=Qwen3.5-27B2026.05 | 25.6 | 63 | |
| Base AgentBackbone=Qwen3.5-27B2026.05 | 22 | 54 | |
| SkillSmithBackbone=Qwen3.5-9B2026.05 | 14.6 | 36 | |
| EvoSkillBackbone=Qwen3.5-9B2026.05 | 13.4 | 33 | |
| Base AgentBackbone=Qwen3.5-9B2026.05 | 11.8 | 29 |