Research Solution Evaluation on ICLR problems 2026 (test)
56Feasibility Win%Mistral-24B
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Mistral-24BComparison Target=Human2025.10 | 56 | 0.48 | 70.7 | 0.01 | 58.7 | 0.3 | |
| Combined LLMComparison Target=Human2025.10 | 52.6 | 0.68 | 73.5 | 0 | 65.2 | 0.006 | |
| GPT-OSS-120BComparison Target=Mistral-24B2025.10 | 50 | 1 | 45.5 | 0.73 | 67.9 | 0.09 | |
| GPT-OSS-120BComparison Target=Human2025.10 | 48.9 | 1 | 76.2 | 0.0009 | 72.1 | 0.005 |