Algorithm Synthesis on Novel Hard Benchmark 15 problems version 2
7Correct Count (out of 15)Reflexion
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| ReflexionSource=version 22026.06 | 7 | 86 | |
| AlgoSkill-MCTSSource=version 22026.06 | 7 | 100 | |
| MapCoderSource=prior-only2026.06 | 6 | 50 | |
| CoTSource=version 22026.06 | 4 | 100 | |
| Self-RefineSource=version 22026.06 | 4 | — | |
| AlgoSkill-GreedySource=version 22026.06 | 4 | 100 | |
| Direct LLMSource=version 22026.06 | 3 | 100 |