Inverse Folding on CAMEO benchmark 2022
46.67AARMCTD-ME
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| MCTD-MEConfiguration=Multi-Expert, Stage=Final, Selection=Informed expert selection2025.09 | 46.67 | 0.0237 | 41.32 | 0.0239 | 42.04 | 0.0503 | |
| DPLM2-650MConfiguration=Single-Expert, Stage=Final2025.09 | 45.35 | 0.011 | 39.74 | 0.0081 | 36.21 | -0.008 | |
| MCTD-UCTConfiguration=Multi-Expert, Stage=Final, Selection=Standard UCB2025.09 | 44.65 | 0.004 | 41.29 | 0.0236 | 38.95 | 0.0098 | |
| DPLM2-150MConfiguration=Single-Expert, Stage=Baseline2025.09 | 44.25 | — | 38.93 | — | 37.01 | — | |
| ProteinMPNNConfiguration=Single-Expert, Stage=Final2025.09 | 43.26 | -0.0098 | 37.86 | -0.0107 | 36.18 | -0.007 | |
| SamplingConfiguration=Multi-Expert, Stage=Final, Tree Search=false2025.09 | 42.64 | -0.0161 | 41.13 | 0.022 | 41.94 | 0.0593 | |
| Random (MCTD-0)Configuration=Random expert routing, Stage=Final2025.09 | 42.5 | -0.0175 | 37.26 | -0.0167 | 37.63 | 0.0062 |