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; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in [1]
Discrete Diffusion Schrödinger Bridge Matching for Graph Transformation
Authors: Jun Hyeong Kim, Seonghwan Kim, Seokhyun Moon, Hyeongwoo Kim, Jeheon Woo, Woo Youn Kim
ICLR 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experimental results demonstrate that DDSBM effectively optimizes molecules property-of-interest with minimal graph transformation, successfully retaining other features. (...) To validate our framework, we evaluated the performance of DDSBM on molecular optimization tasks, with criteria of demonstrating optimal structural modifications to achieve desired property. |
| Researcher Affiliation | Academia | Jun Hyeong Kim1 , Seonghwan Kim1 , Seokhyun Moon1 , Hyeongwoo Kim1 , Jeheon Woo1 , Woo Youn Kim1 (...) Correspondence to EMAIL |
| Pseudocode | Yes | In this manner, the iterative Markov projection following the reciprocal projection is performed alternately in a forward and backward (time-reversal) fashion (see Algorithm 1). (...) Appendix C ALGORITHM Algorithm 1 Iterative Discrete Markovian Fitting |
| Open Source Code | Yes | Source code is available at here. |
| Open Datasets | Yes | We conducted experiments on two different molecule datasets: ZINC250K (Kusner et al., 2017) and Polymer (St. John et al., 2019). |
| Dataset Splits | Yes | The molecular pairs are divided into training and test datasets in a ratio of 8:2. (...) Our final training and test dataset contain 23,936 and 5,984 molecule pairs, respectively. |
| Hardware Specification | Yes | Both DBM and DDSBM reported in this work were trained using four RTX A4000 GPUs. (...) Both models were trained until maximum epochs by default with a single RTX A4000 GPU. |
| Software Dependencies | No | The paper mentions 'Pygmtools (Wang et al., 2024)' and 'RDKit package' but does not specify any version numbers for these software dependencies or other software. |
| Experiment Setup | Yes | The detailed hyperparameters for training are shown in Table 4. (...) Table 9: Training hyperparameters of DBM and DDSBM for unconditional graph generation |