Competitive Programming Reasoning on NOI-119
76Pass RateOracle
Evaluation results
| Method | Links | |
|---|---|---|
| OracleModel=Gemini 3.1 Pro, Protocol=gen-0 pass@20, n=202026.05 | 76 | |
| OpenDeepThink (Generation 3)Protocol=Bradley–Terry top-1, Generation=3, Initial samples (n)=20, Dense round (M)=10, Budget=~285 calls2026.05 | 75 | |
| OpenDeepThink (Generation 1)Protocol=Bradley–Terry top-1, Generation=1, Initial samples (n)=20, Comparisons (K)=4, Budget=~285 calls2026.05 | 70 | |
| OpenDeepThink (Generation 2)Protocol=Bradley–Terry top-1, Generation=2, Initial samples (n)=20, Comparisons (K)=4, Budget=~285 calls2026.05 | 70 | |
| OpenDeepThink (Generation 0)Protocol=Bradley–Terry top-1, Generation=0, Initial samples (n)=20, Comparisons (K)=4, Budget=~285 calls2026.05 | 61 | |
| Pass@1Model=Gemini 3.1 Pro, Protocol=Single unranked gen-0 sample2026.05 | 49 |