aiqbench analysis
This paper introduces AdNGCL, a framework that optimizes negative sample scheduling in graph contrastive learning by dynamically adjusting exposure to easy, intermediate, and hard negatives based on training behavior. The approach improves model performance across ten benchmark datasets while maintaining the standard symmetric InfoNCE objective, offering a refined methodology for self-supervised graph learning tasks.
AdNGCL: Adaptive Negative Scheduling for Graph Contrastive Learning
arXiv:2605.03076v2 Announce Type: replace Abstract: Negative samples play a crucial role in graph contrastive learning (GCL), yet existing methods mainly focus on which negatives should be selected, mined, filtered, or progressively introduced. Less attention has been given to separating the negative candidates available to an anchor from the subset that actively participates in optimization, and to adapting this exposure according to training behavior. To address this gap, we propose the Adaptive Negative Scheduling for Graph Contrastive Learning (AdNGCL) framework, which maintains a bounded per-anchor candidate reservoir and periodically reassesses candidate hardness as representations evolve. Its core component, Hardness-Aware Negative Scheduling (HANS), uses a category-specific training-response signal to determine whether exposure to easy, intermediate, or hard negatives should increase at each scheduling opportunity. Furthermore, reliability-aware candidate refresh handles potentially positive-like candidates, while within-category diversity refresh changes active identities without altering HANS-controlled exposure. AdNGCL retains a standard symmetric InfoNCE objective. Experiments on ten benchmark datasets against state-of-the-art self-supervised graph learning methods show that AdNGCL achieves the highest mean Micro-F1 among the evaluated methods on eight datasets, with improvements of up to 1.30 percentage points over the second-best method, ranks second on Actor, and remains competitive on Coauthor-CS. Ablation, sensitivity, scheduler and reliability diagnostics, and computational analyses further examine the behavior and contribution of the proposed mechanisms.
Abstract:Negative samples play a crucial role in graph contrastive learning (GCL), yet existing methods mainly focus on which negatives should be selected, mined, filtered, or progressively introduced. Less attention has been given to separating the negative candidates available to an anchor from the subset that actively participates in optimization, and to adapting this exposure according to training behavior. To address this gap, we propose the Adaptive Negative Scheduling for Graph Contrastive Learning (AdNGCL) framework, which maintains a bounded per-anchor candidate reservoir and periodically reassesses candidate hardness as representations evolve. Its core component, Hardness-Aware Negative Scheduling (HANS), uses a category-specific training-response signal to determine whether exposure to easy, intermediate, or hard negatives should increase at each scheduling opportunity. Furthermore, reliability-aware candidate refresh handles potentially positive-like candidates, while within-category diversity refresh changes active identities without altering HANS-controlled exposure. AdNGCL retains a standard symmetric InfoNCE objective. Experiments on ten benchmark datasets against state-of-the-art self-supervised graph learning methods show that AdNGCL achieves the highest mean Micro-F1 among the evaluated methods on eight datasets, with improvements of up to 1.30 percentage points over the second-best method, ranks second on Actor, and remains competitive on Coauthor-CS. Ablation, sensitivity, scheduler and reliability diagnostics, and computational analyses further examine the behavior and contribution of the proposed mechanisms.
Substantially revised methodology and experimental evaluation. Updated HANS scheduling and candidate-management mechanisms, independently evaluated baselines under a unified evaluation protocol, and expanded ablation, sensitivity, and computational analyses
Machine Learning (cs.LG)
I.2.6; I.5.1; H.2.8
arXiv:2605.03076 [cs.LG]
(or arXiv:2605.03076v2 [cs.LG] for this version)
Full text under the publisher's open licence · 250 words
Related coverage
How controllers from industrial machinery can coordinate multitask machine learning
Amazon Science · Jul 30, 2026
HYDRA: Proactive Android Malware Drift Adaptation via Hierarchical Graph Contrastive Learning
arXiv · Sep 23, 2026
Scalable and Efficient Joint Spiking Embedding Predictive Architecture for Large-Scale Dynamic Graphs
arXiv · Jul 22, 2026
Feedback Control for Multi-Objective Graph Self-Supervision
arXiv · Sep 9, 2026