Iterative Self-Optimization on OPT-BENCH-ML 20 steps 1.0
8Win CountQwen2.5-3B-Instruct
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Qwen2.5-3B-InstructSteps=20, Model Category=Open-Source, Model Type=General2026.05 | 8 | 38 | 58 | 21 | 14 | |
| gemini-2.0-flashSteps=20, Model Category=Proprietary, Model Type=General2026.05 | 7 | 6 | 10 | 45 | 40 | |
| claude-3-5-sonnet-20241022Steps=20, Model Category=Proprietary, Model Type=General2026.05 | 7 | 18 | 28 | 52 | 47 | |
| claude-3-7-sonnet-20250219Steps=20, Model Category=Proprietary, Model Type=General2026.05 | 7 | 6 | 12 | 66 | 58 | |
| grok-3Steps=20, Model Category=Proprietary, Model Type=General2026.05 | 6 | 12 | 11 | 62 | 58 | |
| Qwen3-235B-InstructSteps=20, Model Category=Proprietary, Model Type=General2026.05 | 6 | 6 | 10 | 57 | 55 | |
| Internlm3-8b-InstructSteps=20, Model Category=Open-Source, Model Type=General2026.05 | 6 | 18 | 20 | 24 | 20 | |
| Qwen2.5-7B-InstructSteps=20, Model Category=Open-Source, Model Type=General2026.05 | 6 | 24 | 28 | 35 | 33 | |
| Qwen2.5-14B-InstructSteps=20, Model Category=Open-Source, Model Type=General2026.05 | 6 | 18 | 18 | 47 | 43 | |
| Qwen2.5-32B-InstructSteps=20, Model Category=Open-Source, Model Type=General2026.05 | 6 | 20 | 20 | 52 | 46 | |
| Qwen3-30B-A3BSteps=20, Model Category=Open-Source, Model Type=General2026.05 | 6 | 0 | 0 | 69 | 68 | |
| gpt-o3-miniSteps=20, Model Category=Proprietary, Model Type=Reasoning2026.05 | 5 | 0 | 0 | 73 | 69 | |
| Deepseek-V3.1-ThinkingSteps=20, Model Category=Proprietary, Model Type=Reasoning2026.05 | 5 | 0 | 0 | 79 | 76 | |
| Qwen3-235B-ThinkingSteps=20, Model Category=Proprietary, Model Type=Reasoning2026.05 | 5 | 0 | 0 | 76 | 76 | |
| Qwen3-8BSteps=20, Model Category=Open-Source, Model Type=General2026.05 | 5 | 12 | 20 | 63 | 60 | |
| Qwen3-32BSteps=20, Model Category=Open-Source, Model Type=General2026.05 | 5 | 6 | 10 | 60 | 60 | |
| gpt-4o-2024-08-06Steps=20, Model Category=Proprietary, Model Type=General2026.05 | 4 | 22 | 18 | 47 | 54 | |
| gpt-4.1-2025-04-14Steps=20, Model Category=Proprietary, Model Type=General2026.05 | 4 | 4 | 2 | 53 | 55 | |
| Qwen2.5-72B-InstructSteps=20, Model Category=Open-Source, Model Type=General2026.05 | 4 | 22 | 32 | 47 | 44 |