Research map / Graph learning / Graph algorithms and clustering
Retrosynthesis Prediction with Conditional Graph Logic Network
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use GLN's conditional graph-logic model instead of the strongest reported baseline | USPTO-50k, reaction class unknown · USPTO-50k, reaction class unknown | top-1 exact match accuracy: 44.4 -> 52.5 | Helped |
| use GLN's conditional graph-logic model instead of the strongest reported baseline | USPTO-50k, reaction class given · USPTO-50k, reaction class given | top-1 exact match accuracy: 55.3 -> 64.2 | Helped |
| use GLN instead of the strongest reported baseline | USPTO-full · USPTO-full | top-1 accuracy: 35.8 -> 39.3 | Helped |
| replace structure2vec with GGNN (ablation) | GLN · USPTO-50k | top-1 accuracy: 52.6 -> 51.6 | Hurt |
| replace structure2vec with MPNN (ablation) | GLN · USPTO-50k | top-1 accuracy: 52.6 -> 50.4 | Hurt |
| replace structure2vec with GIN (ablation) | GLN · USPTO-50k | top-1 accuracy: 52.6 -> 51.8 | Hurt |
| replace structure2vec with ECFP (ablation) | GLN · USPTO-50k | top-1 accuracy: 52.6 -> 51.9 | Hurt |
| use zero message-passing layers instead of three (ablation) | GLN structure2vec encoder · USPTO-50k | top-1 accuracy: 52.6 -> 40.7 | Hurt |
| use one message-passing layer instead of three (ablation) | GLN structure2vec encoder · USPTO-50k | top-1 accuracy: 52.6 -> 47.0 | Hurt |
| use two message-passing layers instead of three (ablation) | GLN structure2vec encoder · USPTO-50k | top-1 accuracy: 52.6 -> 51.3 | Hurt |
| replace inner-product compatibility scoring with MLP scoring (ablation) | GLN · USPTO-50k | top-1 accuracy: 52.6 -> 52.7 | No clear effect |
| replace inner-product compatibility scoring with bilinear scoring (ablation) | GLN · USPTO-50k | top-1 accuracy: 52.6 -> 53.5 | Helped |
Closest papers by meaning
- Learning Graph Models for Retrosynthesis Prediction (NeurIPS 2021 · similarity 0.94)
- Advancing Retrosynthesis with Retrieval-Augmented Graph Generation (AAAI 2025 · similarity 0.90)
- GDiffRetro: Retrosynthesis Prediction with Dual Graph Enhanced Molecular Representation and Diffusion Generation (AAAI 2025 · similarity 0.89)
- LLM-Augmented Chemical Synthesis and Design Decision Programs (ICML 2025 · similarity 0.89)
- RETRO SYNFLOW: Discrete Flow-Matching for Accurate and Diverse Single-Step Retrosynthesis (NeurIPS 2025 · similarity 0.88)
- RetroBridge: Modeling Retrosynthesis with Markov Bridges (ICLR 2024 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.