Retrosynthetic Planning on ref-cnv-400
100Tier-0 ValidityAiZynthFinder MCTS
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| AiZynthFinder MCTSTraining dataset=v2026-05-12-reaction, MCTS iterations=100, Maximum search depth=102026.06 | 100 | 76.5 | 16.8 | 24.5 | 20.2 | 54.8 | 44.7 | 1.3 | |
| Ariadne 24LTraining dataset=v2026-05-12-reaction, Model layers=24, Prompting type=target-only, Beam size=502026.06 | 100 | 33 | 8 | 9.8 | 7 | 18 | 54.2 | 1 | |
| AiZynthFinder MCTSTraining dataset=v2026-05-12-route, MCTS iterations=100, Maximum search depth=102026.06 | 100 | 85.2 | 23.8 | 36.2 | 31.5 | 63.2 | 57.3 | 1.1 | |
| Ariadne 24LTraining dataset=v2026-05-12-route, Model layers=24, Prompting type=target-only, Beam size=502026.06 | 100 | 68.2 | 30.5 | 41.5 | 36.8 | 49.8 | 83.4 | 1.1 |