Bias Mitigation in Graph Diffusion Models
Authors: Meng Yu, Kun Zhan
ICLR 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Our approach, which requires no network modifications, is validated across multiple models, datasets, and tasks, achieving state-of-the-art results. [...] In this section, we select three generic graph datasets and two molecular datasets to evaluate the performance of our approach. [...] We conduct extensive ablation experiments in 5.4 to demonstrate this point. |
| Researcher Affiliation | Academia | Meng Yu, Kun Zhan School of Information Science & Engineering, Lanzhou University Corresponding Author. Email: EMAIL |
| Pseudocode | Yes | Algorithm 1 The S++ sampling algorithm. Input: An inference model sθ,t( ), a pretrained pseudo model sψ,t( ), and the cut-off time tc . |
| Open Source Code | No | The paper does not explicitly state that the code is open-source or provide a link to a code repository. It only discusses the methodology and experimental results without mentioning code availability. |
| Open Datasets | Yes | We select three generic graph datasets to test our approach: (1) Community-small: 100 artificially generated graphs with community structure; (2) Enzymes: 600 protein maps representing the enzyme structure in BRENDA (Schomburg et al., 2004); (3) Grid: 100 standard 2D grid diagrams. [...] We select two widely recognized molecular datasets to evaluate our approach: QM9 (Ramakrishnan et al., 2014) and ZINC250k (Irwin et al., 2012). |
| Dataset Splits | No | The paper mentions comparing generated graphs with 'test plots' or 'benchmark molecular dataset' but does not specify the exact percentages, sample counts, or explicit methodology for splitting the datasets into training, validation, and test sets for their experiments. |
| Hardware Specification | No | The paper does not provide specific details about the hardware used to run the experiments, such as GPU models, CPU types, or memory specifications. |
| Software Dependencies | No | The paper does not provide specific software names with version numbers for libraries or tools used in the implementation, beyond general mentions of diffusion models. |
| Experiment Setup | Yes | We provide detailed parameters for experiments described in 5, as shown in Tables 10 and 11. Specifically, we distinguish between the relevant parameters for sampling methods with correctors and those without correctors. As shown in Tables 10, both Langevin sampling and the S++ correction are implemented in the reverse-starting alignment stage. [...] Table 10: Experimental parameters for sampling methods without a corrector (OC). Model Hyper. Comm. Enzymes Grid QM9 ZINC250k M 400 420 350 400 400 λ1 0.2 0.0008 0.06 1.19 2.5 GDSS-OC-S++ ω1 0.998 1.0 1.0 1.09 1.0 λ2 0 0 0 0 0 ω2 1.0 1.0 1.0 1.0 1.0 |