Logical Reasoning on GeoShape BBH
90AccuracyMemAPO
Evaluation Results
| Method | Links | |
|---|---|---|
| MemAPOBackbone=GPT-4o-mini2026.03 | 90 | |
| MemAPOBackbone=Qwen3-8B2026.03 | 77 | |
| TextGradBackbone=GPT-4o-mini2026.03 | 76 | |
| ProTeGiBackbone=GPT-4o-mini2026.03 | 71 | |
| CoTBackbone=Qwen3-8B2026.03 | 64 | |
| Step-BackBackbone=GPT-4o-mini2026.03 | 61 | |
| Step-BackBackbone=Qwen3-8B2026.03 | 58 | |
| ProTeGiBackbone=Qwen3-8B2026.03 | 56 | |
| SPOBackbone=GPT-4o-mini2026.03 | 54 | |
| RaRBackbone=Qwen3-8B2026.03 | 53 | |
| CoTBackbone=GPT-4o-mini2026.03 | 47 | |
| RaRBackbone=GPT-4o-mini2026.03 | 43 | |
| PromptBreederBackbone=GPT-4o-mini2026.03 | 42 | |
| TextGradBackbone=Qwen3-8B2026.03 | 39 | |
| OPROBackbone=Qwen3-8B2026.03 | 36 | |
| IOBackbone=GPT-4o-mini2026.03 | 35 | |
| OPROBackbone=GPT-4o-mini2026.03 | 35 | |
| SPOBackbone=Qwen3-8B2026.03 | 28 | |
| IOBackbone=Qwen3-8B2026.03 | 23 | |
| PromptBreederBackbone=Qwen3-8B2026.03 | 23 |