Scientific Idea Generation on ICLR 2024
4.22Absolute NoveltyOur Method - BoN(10)
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Our Method - BoN(10)Backbone=Qwen 2.5 14B, Quantization=4-bit, Best-of-N=102026.04 | 4.22 | 3.88 | 4.4 | 5 | 1.25 | 5 | |
| Our MethodModel=unsloth-Qwen2.5-7B, Strategy=BoN (10)2026.04 | 4.18 | 4.1 | 4.53 | 5 | 1.98 | 5 | |
| SFTBackbone=Qwen 2.5 14B, Mode=Supervised fine-tuning2026.04 | 4.1 | 3.8 | 4.25 | 2.42 | 2.66 | 1 | |
| GPT Researcher2026.04 | 4.08 | 3.75 | 4.3 | 4.01 | 1.99 | 3.21 | |
| Research Agent2026.04 | 4.08 | 3.95 | 4.67 | 3.63 | 2.1 | 4.89 | |
| LDC2026.04 | 4.08 | 3.98 | 4.38 | 4.38 | 1 | 4.55 | |
| Base ModelModel=unsloth-Qwen2.5-7B2026.04 | 4.03 | 4.12 | 4.4 | 2.15 | 5 | 3.6 | |
| AI Scientist V22026.04 | 4 | 4.05 | 4.25 | 3.25 | 3.88 | 3.79 | |
| Base Model 14B 4-bitBackbone=Qwen 2.5 14B, Quantization=4-bit2026.04 | 3.92 | 4.08 | 4.42 | 1.05 | 5 | 2.43 | |
| AI Scientist V22026.04 | 3.9 | 4.22 | 4.22 | 1 | 4.44 | 1.92 | |
| GPT Researcher2026.04 | 3.9 | 3.9 | 4.45 | 4.4 | 1 | 3.33 | |
| Research Agent2026.04 | 3.75 | 3.62 | 4.1 | 1 | 4.16 | 1 |