Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in Coakley et alK. L. Coakley, T. Snelleman, H. Hoos, and O. E. Gundersen, "The embrace of open science: An analysis of a decade of AI research and 56 800 conference papers," Under Review, 2026..
Self-Improved Retrosynthetic Planning
Authors: Junsu Kim, Sungsoo Ahn, Hankook Lee, Jinwoo Shin
ICML 2021 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Our experiments demonstrate that our scheme significantly improves the success rate of solving the retrosynthetic problem from 86.84% to 96.32% while maintaining the performance of DNN for predicting valid reactions. |
| Researcher Affiliation | Academia | 1Korea Advanced Institute of Science and Technology (KAIST) 2Mohamed bin Zayed University of Artificial Intelligence (MBZUAI). |
| Pseudocode | Yes | We provide an illustration and a detailed description of our framework in Figure 2 and Algorithm 1, respectively. Algorithm 1 Self-Improved Retrosynthetic Planning |
| Open Source Code | No | The paper thanks Binghong Chen for providing the dataset and source implementation of RETRO* and provides a link to their GitHub (https://github.com/binghong-ml/retro_star). This is code for a baseline used, not the authors' own implementation of their proposed self-improved framework. |
| Open Datasets | Yes | For the target molecules Dtarget, we choose synthesizable molecules from I and reactions in the United States Patent Office (USPTO) database (Lowe, 2012). For the reaction dataset Dreaction, we use reactions extracted from USPTO, following training/validation/test splits by Chen et al. (2020b). |
| Dataset Splits | Yes | For the reaction dataset Dreaction, we use reactions extracted from USPTO, following training/validation/test splits by Chen et al. (2020b). |
| Hardware Specification | No | The paper does not provide any specific hardware details (e.g., GPU/CPU models, memory, or cloud instance types) used for running the experiments. |
| Software Dependencies | No | The paper mentions software components like RDChiral, Morgan fingerprint, and Adam optimizer but does not specify their version numbers, which are required for reproducibility. |
| Experiment Setup | Yes | The forward reaction model pf is trained with a learning rate of 0.001 for 100 epochs. ... the backward reaction model pb is trained with a learning rate of 0.0001 for 20 epochs. Adam optimizer (Kingma & Ba, 2014) is used with a mini-batch of size 1024 for training all the models. We iterate our overall procedure three times. ... We set both thresholds ϵ, ϵaug as 0.8. |