| RetroComposer (reaction class as prior) | 65.9 | 91.5 | 85.8 | 89.5 | RetroComposer: Composing Templates for Template-Based Retrosynthesis Prediction | - |
| LocalRetro (reaction class unknown) | 53.4 | 92.4 | 77.5 | 85.9 | Deep Retrosynthetic Reaction Prediction using Local Reactivity and Global Attention | |
| Augmented Transformer (reaction class unknown) | 53.5 | 85.7 | - | 81.0 | State-of-the-Art Augmented NLP Transformer models for direct and single-step retrosynthesis | - |
| Expert System (reaction class as prior) | 35.2 | 65.1 | 52.3 | 59.1 | Retrosynthetic reaction prediction using neural sequence-to-sequence models | - |
| RetroSim (reaction class as prior) | 52.9 | 88.1 | 73.8 | 81.2 | Computer-Assisted Retrosynthesis Based on Molecular Similarity | |
| GraphRetro (reaction class as prior) | 63.9 | 88.1 | 81.5 | 85.2 | Learning Graph Models for Template-Free Retrosynthesis | - |
| GET-LT1 (reaction class unknown) | 44.9 | 65.9 | 58.8 | 62.4 | Molecular Graph Enhanced Transformer for Retrosynthesis Prediction | - |
| MARS (reaction class as prior) | 66.2 | 92.9 | 85.8 | 90.2 | MARS: A Motif-based Autoregressive Model for Retrosynthesis Prediction | - |
| Chemformer-Large (reaction class unknown) | 54.3 | 63.0 | - | 62.3 | Chemformer: a pre-trained transformer for computational chemistry | |
| Graph2SMILES-D-GCN (reaction class unknown) | 52.9 | 72.9 | 66.5 | 70.0 | Permutation invariant graph-to-sequence model for template-free retrosynthesis and reaction prediction | - |
| SemiRetro (reaction class unknown) | 54.9 | 84.1 | 75.3 | 80.4 | SemiRetro: Semi-template framework boosts deep retrosynthesis prediction | - |
| MEGAN (reaction class unknown) | 48.1 | 86.1 | 70.7 | 78.4 | Molecule Edit Graph Attention Network: Modeling Chemical Reactions as Sequences of Graph Edits | - |
| GET-LT1 (reaction class as prior) | 57.4 | 77.4 | 71.3 | 74.8 | Molecular Graph Enhanced Transformer for Retrosynthesis Prediction | - |
| GET-LT2 (reaction class as prior) | 56.2 | 74.7 | 69.4 | 72.5 | Molecular Graph Enhanced Transformer for Retrosynthesis Prediction | - |
| Dual-TF (reaction class as prior) | 65.7 | 85.9 | 81.9 | 84.7 | Energy-based View of Retrosynthesis | - |
| Graph2SMILES-D-GAT (reaction class unknown) | 51.2 | 73.9 | 66.3 | 70.4 | Permutation invariant graph-to-sequence model for template-free retrosynthesis and reaction prediction | - |
| LocalRetro (reaction class as prior) | 63.9 | 96.3 | 86.1 | 92.4 | Deep Retrosynthetic Reaction Prediction using Local Reactivity and Global Attention | |
| GLN | 52.5 | 83.7 | 69 | 75.6 | Retrosynthesis Prediction with Conditional Graph Logic Network | - |
| SemiRetro (reaction class as prior) | 65.8 | 92.8 | 85.7 | 89.8 | SemiRetro: Semi-template framework boosts deep retrosynthesis prediction | - |
| MHNreact (reaction class unknown) | 51.8 | 88.1 | 74.6 | 81.2 | Improving Few- and Zero-Shot Reaction Template Prediction Using Modern Hopfield Networks | |