Research Idea Generation on Ten benchmark topics (100 generated research ideas)
4.27Average WinsEvoSci
Evaluation results
| Method | Links | ||
|---|---|---|---|
| EvoSciAgent Model=GPT-4o2026.05 | 4.27 | 54 | |
| EvoSciAgent Model=Qwen3-max2026.05 | 4.25 | 50 | |
| EvoSciAgent Model=DeepSeek-v32026.05 | 4.19 | 47 | |
| CoI-AgentAgent Model=DeepSeek-v32026.05 | 4.08 | 37 | |
| VirSciAgent Model=GPT-4o2026.05 | 4.07 | 52 | |
| CoI-AgentAgent Model=Qwen3-max2026.05 | 4 | 39 | |
| VirSciAgent Model=Qwen3-max2026.05 | 3.94 | 34 | |
| VirSciAgent Model=DeepSeek-v32026.05 | 3.9 | 35 | |
| AI ScientistAgent Model=GPT-4o2026.05 | 3.88 | 13 | |
| CoI-AgentAgent Model=GPT-4o2026.05 | 3.58 | 36 | |
| AI ScientistAgent Model=Qwen3-max2026.05 | 2.92 | 7 | |
| AI ScientistAgent Model=DeepSeek-v32026.05 | 2.83 | 8 | |
| SciPIPAgent Model=GPT-4o2026.05 | 2.7 | 7 | |
| SciPIPAgent Model=DeepSeek-v32026.05 | 2.5 | 3 | |
| SciPIPAgent Model=Qwen3-max2026.05 | 2.39 | 1 |