Agentic Compositional Generalization on SkillBench
8.69Skill Depth (k=2)GPT-5.2
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| GPT-5.2Model Category=Advanced Agentic Models2026.01 | 8.69 | 8.72 | 8.84 | 8.67 | 8.66 | 8.17 | 8.62 | |
| DeepSeek-v3.2Model Category=Advanced Agentic Models2026.01 | 8.42 | 8.43 | 8.08 | 8.16 | 8.19 | 8 | 8.21 | |
| Gemini-3-ProModel Category=Advanced Agentic Models2026.01 | 7.87 | 7.89 | 7.89 | 7.7 | 7.59 | 6.81 | 7.62 | |
| Qwen2.5-32BModel Category=Instruct Models2026.01 | 5.5 | 5.16 | 4.89 | 5.25 | 5.29 | 5.16 | 5.21 | |
| Qwen2.5-14BModel Category=Instruct Models2026.01 | 5.39 | 5.22 | 5 | 5.04 | 5.13 | 4.94 | 5.12 | |
| Qwen2.5-7BModel Category=Instruct Models2026.01 | 5.24 | 5.05 | 4.81 | 4.91 | 5.24 | 4.97 | 5.04 | |
| SearchR1Model Category=Tool-Integrated Agentic Models2026.01 | 4.73 | 4.36 | 4.44 | 4.53 | 4.82 | 4.32 | 4.54 | |
| ReSearchModel Category=Tool-Integrated Agentic Models2026.01 | 4.72 | 4.46 | 4.06 | 4.37 | 4.41 | 4.35 | 4.4 | |
| AutoTIRModel Category=Tool-Integrated Agentic Models2026.01 | 4.12 | 4.21 | 4.1 | 4.03 | 4.07 | 4.02 | 4.09 |