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..
MIPT: Multilevel Informed Prompt Tuning for Robust Molecular Property Prediction
Authors: Yeyun Chen, Jiangming Shi
ICML 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experimental results show that MIPT surpasses all baselines, aligning graph space and task space while achieving significant improvements in molecule-related tasks, demonstrating its scalability and versatility for molecular tasks. |
| Researcher Affiliation | Academia | 1Institute of Artificial Intelligence, Xiamen University, Xiamen Fujian, China 2Shanghai Innovation Institute, Shanghai, China. Correspondence to: Jiangming Shi <EMAIL>. |
| Pseudocode | Yes | Pseudocode is presented in Algorithm 1. |
| Open Source Code | No | The paper does not provide concrete access to source code. It does not contain an explicit code release statement, a specific repository link, or mention of code in supplementary materials. |
| Open Datasets | Yes | We employ eight common datasets from Molecule Net (Wu et al., 2018) as our benchmark datasets: BBBP, Tox21, Tox Cast, SIDER, Clin Tox, MUV, HIV and BACE. |
| Dataset Splits | No | Random splits and scaffold splits for these datasets are adopted. |
| Hardware Specification | Yes | All experiments were conducted on a high-performance computing server equipped with an NVIDIA 3090 GPU (24 GB memory). |
| Software Dependencies | Yes | The implementation was based on Python 3.9, Py Torch 1.12, and the torc geometric library. |
| Experiment Setup | Yes | For the GNN architecture, we utilized Graph Isomorphism Network (GIN), configured with a hidden dimension of 300, 3 graph convolutional layers, Re LU activation, and batch normalization. The optimizer was Adam with a learning rate of 0.001, dropout rate is 0.5, the mask probability is 0.2. ... The experiments were conducted on Molecule Net datasets, running for 100 epochs with a batch size of 32. |