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..
Let Brain Rhythm Shape Machine Intelligence for Connecting Dots on Graphs
Authors: Jiaqi Ding, Tingting Dan, Zhixuan Zhou, Guorong Wu
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Empirical evaluations demonstrate that this synchronization-driven mechanism not only mitigates over-smoothing in deep GNNs but also enhances the model s capacity for reasoning and solving complex graph-based problems. In this section, we present extensive experiments to validate the effectiveness and interpretability of our proposed models across a variety of brain-related and general graph learning tasks. |
| Researcher Affiliation | Academia | 1Department of Computer Science 2Department of Psychiatry The University of North Carolina at Chapel Hill EMAIL EMAIL |
| Pseudocode | Yes | Algorithm 1: Iterative Solver for BRICK Dynamics (Eq. 3) |
| Open Source Code | Yes | The code is released at Anonymous Git Hub. |
| Open Datasets | Yes | To evaluate the effectiveness of BRICK and BIG-NOS, we use the publicly available neuroimaging and graph datasets, respectively. Human brain datasets 1. HCP-Aging (HCP-A) [5]. ... 2. HCP-Young Adults (HCP-YA) [54]. ... Graph-based datasets 1. Node classification. ... Texas ..., Wisconsin ..., Actor ..., Squirrel ..., Chameleon ..., Cornell ..., Citeseer ..., Pubmed ..., Cora .... ogbn-arxiv ... 2. Graph classification. ... ENZYMES and PROTEINS from TUDataset [44] |
| Dataset Splits | Yes | We perform 5-fold cross-validation for both the 4-class (HCP-A), 7-class (HCP-YA) and 8-class (HCP-WM) classification tasks. For homophilic graph data (Cora, Citeseer, Pubmed) in node classification tasks, we adopt the semi-supervised 20-per-class training split from [34] and average over 5 random seeds. For heterophilic graph data and TUDataset, we follow [47] and [21], using 10-fold cross-validation and reporting test-set averages. |
| Hardware Specification | No | The paper does not provide specific hardware details for the experiments performed. |
| Software Dependencies | No | The paper does not explicitly mention specific software dependencies with version numbers. |
| Experiment Setup | Yes | As human data as an example, for all methods, the hidden dimension is set to 256. The network depth and batch size are set to 2 and 64 for the comparison methods, while for our BRICK, they are set to 16 and 256, respectively. |