Hierarchical Task Decomposition on MagicEval-Plan Dependency 3
16Step CountQwen3-30B-A3B-Instruct
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Qwen3-30B-A3B-InstructScale=Large-Scale Models2026.02 | 16 | 91.5 | 91.8 | |
| GPT-5.2Scale=Ultra-Scale Models2026.02 | 28.5 | 89.9 | 96.5 | |
| Llama3.3-70B-InstructScale=Large-Scale Models2026.02 | 38.5 | 90.7 | 86.4 | |
| GLM-4.7-FlashScale=Large-Scale Models2026.02 | 46.5 | 87.1 | 79 | |
| Olmo-3.1-32B-InstructScale=Large-Scale Models2026.02 | 47.5 | 90.5 | 89.5 | |
| DeepSeek-V3.1-nothinkScale=Ultra-Scale Models2026.02 | 64 | 96.9 | 88.3 | |
| Qwen3-235B-A22B-InstructScale=Ultra-Scale Models2026.02 | 65.5 | 90.5 | 87 | |
| ERNIE-4.5-21B-A3B-PTScale=Large-Scale Models2026.02 | 71 | 91.6 | 86.6 | |
| Qwen3-32B-nothinkScale=Large-Scale Models2026.02 | 72.5 | 91.6 | 88.6 | |
| Qwen3-MAXScale=Ultra-Scale Models2026.02 | 78 | 91.8 | 88.8 | |
| Kimi-K2-InstructScale=Ultra-Scale Models2026.02 | 79.5 | 91.7 | 88.1 | |
| GLM-4.7Scale=Ultra-Scale Models2026.02 | 82.5 | 96.8 | 89.1 | |
| MagicAgent-30B-A3BScale=MagicAgent Series2026.02 | 97 | 95.6 | 92.3 | |
| MagicAgent-32BScale=MagicAgent Series2026.02 | 98.2 | 95.7 | 91.2 |