Research map / Label-efficient and robust learning / Recognition and long-tail
AI for Interpretable Chemistry: Predicting Radical Mechanistic Pathways via Contrastive Learning
Mohammadamin Tavakoli, Pierre Baldi, Ann Marie Carlton, Yinting Chiu, Alexander Shmakov, David Van Vranken
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 GNN instead of atom descriptor for reactive sites | reactive site identification · RMechDB combined test | top-5 accuracy: 89.3 -> 92.1 | Improved |
| use DRFP instead of pre-defined feature vectors for plausibility ranking | plausibility ranker · RMechDB combined test | top-1 accuracy: 73.1 -> 78.6 | Improved |
| use DRFP instead of pretrained rxnfp | plausibility ranker · RMechDB combined test | top-3 accuracy: 86.2 -> 90.2 | Improved |
| use AP reactionFP instead of Morgan2 / TT fingerprints | reactionFP ranker · RMechDB combined test | top-1 accuracy: 74.3 -> 74.6 | No clear effect |
| replace atom descriptor with RxnHypergraph in contrastive model (ablation) | contrastive model · RMechDB Core | top-1 accuracy: 62.9 -> 64.3 | Improved |
| replace atom descriptor with RxnHypergraph in contrastive model (ablation) | contrastive model · RMechDB Atmospheric | top-1 accuracy: 61.0 -> 62.1 | Improved |
| contrastive model instead of two-step method | reaction predictor · RMechDB Core | top-1 accuracy: 62.4 -> 64.3 | Improved |
| contrastive atom descriptor instead of RxnHypergraph for inference speed | contrastive model · RMechDB | inference time: 1.45 -> 0.08 | Improved |
| contrastive atom descriptor instead of two-step for inference speed | reaction predictor · RMechDB | inference time: 1.38 -> 0.08 | Improved |
| fine-tune molecular transformer on RMechDB | text-based model · RMechDB Core | top-1 accuracy: 58.2 -> 57.7 | Hurt |
| fine-tune molecular transformer on RMechDB | text-based model · RMechDB Atmospheric | top-1 accuracy: 58.0 -> 57.1 | Hurt |
| graph-based contrastive model instead of text-based pretrained transformer | reaction predictor · RMechDB Core | top-5 accuracy: 84.2 -> 95.1 | Improved |
Closest papers by meaning
- Reaction Prediction via Interaction Modeling of Symmetric Difference Shingle Sets (NeurIPS 2025 · similarity 0.85)
- ReactGPT: Understanding of Chemical Reactions via In-Context Tuning (AAAI 2025 · similarity 0.84)
- Reaction Graph: Towards Reaction-Level Modeling for Chemical Reactions with 3D Structures (ICML 2025 · similarity 0.84)
- LLM-Augmented Chemical Synthesis and Design Decision Programs (ICML 2025 · similarity 0.83)
- TinyChemVL: Advancing Chemical Vision-Language Models via Efficient Visual Token Reduction and Complex Reaction Tasks (AAAI 2026 · similarity 0.83)
- Uncertainty-Aware Yield Prediction with Multimodal Molecular Features (AAAI 2024 · similarity 0.83)
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.
