arxiv: 2605.12998 · v2 · submitted 2026-05-13 · 💻 cs.LG

Recognition: no theorem link

DRIFT: A Benchmark for Task-Free Continual Graph Learning with Continuous Distribution Shifts

Guiquan Sun , Xikun Zhang , Jingchao Ni , Dongjin Song

Authors on Pith no claims yet

Pith reviewed 2026-05-15 05:42 UTC · model grok-4.3

classification 💻 cs.LG

keywords continual learninggraph neural networksdistribution shifttask-free learningcatastrophic forgettingbenchmark

0 comments

The pith

Many existing continual graph learning methods implicitly depend on task boundaries and degrade under continuous distribution shifts in task-free streams.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper argues that continual graph learning has been evaluated under unrealistic assumptions of discrete tasks with known boundaries. It introduces a task-free formulation where graph data streams are modeled as time-varying mixtures of latent distributions with smooth transitions. The DRIFT benchmark implements this using Gaussian parameterizations to simulate various drift speeds. Evaluations show that standard methods suffer substantial performance loss in this setting. This matters because real environments lack task labels and exhibit gradual changes rather than abrupt switches.

Core claim

By reformulating continual graph learning as learning from a continuous stream without task identities, where the data is a mixture of latent task distributions evolving over time, the DRIFT benchmark demonstrates that representative methods experience significant degradation compared to task-based settings, indicating their reliance on boundary information.

What carries the argument

The unified task-free formulation that treats the data stream as a time-varying mixture of latent task distributions parameterized by Gaussians for transition dynamics.

If this is right

Existing CGL approaches must be adapted to operate without access to task boundaries.
New algorithms are needed that can handle smooth distributional drifts in graph data.
Benchmarks for continual learning should incorporate continuous shift scenarios to better reflect real-world conditions.
Performance evaluation in graph streams should focus on long-term adaptation rather than per-task accuracy.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

This could extend to other continual learning domains where data evolves gradually without explicit task divisions.
Developers of streaming graph systems may need to incorporate drift detection mechanisms that do not rely on task cues.
Testing on real-world evolving networks like social media or citation graphs with natural drifts would validate the benchmark's relevance.

Load-bearing premise

The assumption that Gaussian mixtures can represent the full range of real-world continuous shifts in graph distributions.

What would settle it

A continual graph learning method that maintains high performance across both discrete task-based and continuous task-free settings on the same underlying data would challenge the claim that existing methods rely on boundaries.

Figures

Figures reproduced from arXiv: 2605.12998 by Dongjin Song, Guiquan Sun, Jingchao Ni, Xikun Zhang.

**Figure 2.** Figure 2: The dynamics of test accuracy of all implemented baselines on four datasets. [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗

**Figure 3.** Figure 3: Effect of the transition scale σ on CoraFull-CL. However, improved adaptation is accompanied by increased forgetting. ER degrades from −37.6% forgetting at σ=3 to −48.0% at σ=20, while A-GEM drops from −38.4% to −53.9%. Although smoother transitions improve online adaptation, they simultaneously weaken the effective training signal associated with each latent distribution, making old knowledge harder to … view at source ↗

**Figure 4.** Figure 4: Effect of with- vs. without-replacement sampling [PITH_FULL_IMAGE:figures/full_fig_p009_4.png] view at source ↗

**Figure 5.** Figure 5: t-SNE visualization of learned node embeddings on Reddit-CL. [PITH_FULL_IMAGE:figures/full_fig_p020_5.png] view at source ↗

read the original abstract

Continual graph learning (CGL) aims to learn from dynamically evolving graphs while mitigating catastrophic forgetting. Existing CGL approaches typically adopt a task-based formulation, where the data stream is partitioned into a sequence of discrete tasks with pre-defined boundaries. However, such assumptions rarely hold in real-world environments, where data distributions evolve continuously and task identity is often unavailable. To better reflect realistic non-stationary environments, we revisit continual graph learning from a task-free perspective. We propose a unified formulation that models the data stream as a time-varying mixture of latent task distributions, enabling continuous modeling of distribution drift. Based on this formulation, we construct \emph{DRIFT}, a benchmark that spans a spectrum of transition dynamics ranging from hard task switches to smooth distributional drift through a Gaussian parameterization. We evaluate representative continual learning methods under this task-free setting and observe substantial performance degradation compared to traditional task-based protocols. Our findings indicate that many existing approaches implicitly rely on task boundary information and struggle under realistic task-free graph streams. This work highlights the importance of studying continual graph learning under realistic non-stationary conditions and provides a benchmark for future research in this direction. Our code is available at https://github.com/UConn-DSIS/DRIFT.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

DRIFT gives a clean benchmark for task-free continual graph learning and shows clear degradation without boundaries, but the Gaussian drift model still needs real-data checks.

read the letter

The main point is that the authors move continual graph learning away from the usual task-based setup with known boundaries and instead model streams as time-varying mixtures of latent tasks, with transitions parameterized by Gaussians to produce everything from abrupt switches to smooth drift. They build DRIFT around this and run representative methods on it, finding noticeable drops in performance once boundary information is removed. That observation is useful and the benchmark construction itself is a concrete step forward for the subfield.

Referee Report

1 major / 1 minor

Summary. The paper introduces a task-free formulation for continual graph learning (CGL) that models evolving graph streams as time-varying mixtures of latent task distributions with transitions drawn from a Gaussian process. It constructs the DRIFT benchmark spanning hard task switches to smooth distributional drifts, evaluates representative CGL methods in this setting, and reports substantial performance degradation relative to task-based protocols, concluding that existing approaches implicitly rely on task boundary information.

Significance. If the generated streams faithfully capture the spectrum of real-world continuous distribution shifts, the benchmark would be a valuable contribution for developing and evaluating task-free CGL methods. The open-source code is a clear strength that supports reproducibility.

major comments (1)

[Benchmark Construction] Benchmark construction: The Gaussian parameterization of transition dynamics between latent task distributions is used both to generate DRIFT (hard switches to smooth drift) and to support the claim of realism, yet no section compares the induced statistics (e.g., rate of change in node/edge features, community overlap, or degree distributions) to observed real-world non-stationary graph streams such as evolving citation or social networks. This assumption is load-bearing for attributing measured degradation specifically to the absence of task boundaries rather than to the chosen dynamics.

minor comments (1)

[Abstract] Abstract: The description of observed degradation would be strengthened by briefly naming the primary metrics, base datasets, and any statistical significance tests employed.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. We address the major comment point by point below and will incorporate revisions to strengthen the benchmark validation.

read point-by-point responses

Referee: [Benchmark Construction] Benchmark construction: The Gaussian parameterization of transition dynamics between latent task distributions is used both to generate DRIFT (hard switches to smooth drift) and to support the claim of realism, yet no section compares the induced statistics (e.g., rate of change in node/edge features, community overlap, or degree distributions) to observed real-world non-stationary graph streams such as evolving citation or social networks. This assumption is load-bearing for attributing measured degradation specifically to the absence of task boundaries rather than to the chosen dynamics.

Authors: We agree that direct empirical comparisons of the generated streams' statistics to real-world non-stationary graphs would strengthen the realism claim and better support attribution of the observed degradation to the task-free continuous-shift setting. The Gaussian mixture parameterization with Gaussian process transitions was selected primarily to enable controlled variation across a spectrum of drift regimes (abrupt to smooth) while remaining computationally tractable and reproducible. In the revised manuscript we will add a new subsection under benchmark construction that reports quantitative statistics for representative DRIFT configurations—including node/edge feature change rates, community overlap (e.g., via normalized mutual information or modularity), and degree-distribution shifts—and compares them to publicly available temporal graph datasets such as citation networks (DBLP, arXiv) and social networks with timestamped edges. This addition will clarify the relationship to observed real-world dynamics without altering the core contribution of the task-free formulation or the benchmark itself. revision: yes

Circularity Check

0 steps flagged

No circularity: benchmark is an independent empirical construction

full rationale

The paper defines a task-free formulation for continual graph learning as a time-varying mixture of latent distributions and generates the DRIFT benchmark via an explicit Gaussian parameterization of transitions. This is a modeling choice for creating evaluation streams, not a derivation that reduces any claimed result to its own inputs by construction. No equations, predictions, or uniqueness claims are shown to collapse into fitted parameters or self-citations; the performance degradation findings are direct empirical observations on the generated streams rather than forced outputs. The work is self-contained as benchmark construction with independent evaluation protocols.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

The paper is an empirical benchmark contribution with no mathematical derivations, so it introduces no free parameters, axioms, or invented entities beyond standard graph and distribution assumptions already present in the field.

pith-pipeline@v0.9.0 · 5526 in / 916 out tokens · 37196 ms · 2026-05-15T05:42:58.834938+00:00 · methodology

Review history (2 revisions) →

discussion (0)

Reference graph

Works this paper leans on

51 extracted references · 51 canonical work pages · 3 internal anchors

[1]

Inductive representation learning on large graphs

Will Hamilton, Zhitao Ying, and Jure Leskovec. Inductive representation learning on large graphs. InAdvances in Neural Information Processing Systems, volume 30, 2017

work page 2017
[2]

Open graph benchmark: Datasets for machine learning on graphs

Weihua Hu, Matthias Fey, Marinka Zitnik, Yuxiao Dong, Hongyu Ren, Bowen Liu, Michele Catasta, and Jure Leskovec. Open graph benchmark: Datasets for machine learning on graphs. Advances in neural information processing systems, 33:22118–22133, 2020

work page 2020
[3]

Knowledge graph embedding: A survey of approaches and applications.IEEE transactions on knowledge and data engineering, 29:2724– 2743, 2017

Quan Wang, Zhendong Mao, Bin Wang, and Li Guo. Knowledge graph embedding: A survey of approaches and applications.IEEE transactions on knowledge and data engineering, 29:2724– 2743, 2017

work page 2017
[4]

Overcoming catastrophic forgetting in neural networks.Proceedings of the national academy of sciences, 114(13):3521–3526, 2017

James Kirkpatrick, Razvan Pascanu, Neil Rabinowitz, Joel Veness, Guillaume Desjardins, Andrei A Rusu, Kieran Milan, John Quan, Tiago Ramalho, Agnieszka Grabska-Barwinska, et al. Overcoming catastrophic forgetting in neural networks.Proceedings of the national academy of sciences, 114(13):3521–3526, 2017

work page 2017
[5]

On Tiny Episodic Memories in Continual Learning

Arslan Chaudhry, Marcus Rohrbach, Mohamed Elhoseiny, Thalaiyasingam Ajanthan, Puneet K Dokania, Philip HS Torr, and Marc’Aurelio Ranzato. On tiny episodic memories in continual learning.arXiv preprint arXiv:1902.10486, 2019

work page internal anchor Pith review Pith/arXiv arXiv 1902
[6]

Cglb: Benchmark tasks for continual graph learning.Advances in Neural Information Processing Systems, 35:13006–13021, 2022

Xikun Zhang, Dongjin Song, and Dacheng Tao. Cglb: Benchmark tasks for continual graph learning.Advances in Neural Information Processing Systems, 35:13006–13021, 2022

work page 2022
[7]

Aravind Sankar, Yanhong Wu, Liang Gou, Wei Zhang, and Hao Yang

Emanuele Rossi, Ben Chamberlain, Fabrizio Frasca, Davide Eynard, Federico Monti, and Michael Bronstein. Temporal graph networks for deep learning on dynamic graphs.arXiv preprint arXiv:2006.10637, 2020

work page arXiv 2006
[8]

Online continual learning on class incremental blurry task configuration with anytime inference

Hyunseo Koh, Dahyun Kim, Jung-Woo Ha, and Jonghyun Choi. Online continual learning on class incremental blurry task configuration with anytime inference. InInternational Conference on Learning Representations, 2022

work page 2022
[9]

Online class incremental learning on stochastic blurry task boundary via mask and visual prompt tuning

Jun-Yeong Moon, Keon-Hee Park, Jung Uk Kim, and Gyeong-Moon Park. Online class incremental learning on stochastic blurry task boundary via mask and visual prompt tuning. In Proceedings of the IEEE/CVF international conference on computer vision, pages 11731–11741, 2023

work page 2023
[10]

Efficient Lifelong Learning with A-GEM

Arslan Chaudhry, Marc’Aurelio Ranzato, Marcus Rohrbach, and Mohamed Elhoseiny. Efficient lifelong learning with a-gem.arXiv preprint arXiv:1812.00420, 2018

work page internal anchor Pith review Pith/arXiv arXiv 2018
[11]

Gradient based sample selection for online continual learning.Advances in neural information processing systems, 32, 2019

Rahaf Aljundi, Min Lin, Baptiste Goujaud, and Yoshua Bengio. Gradient based sample selection for online continual learning.Advances in neural information processing systems, 32, 2019

work page 2019
[12]

Dark experience for general continual learning: a strong, simple baseline.Advances in neural information processing systems, 33:15920–15930, 2020

Pietro Buzzega, Matteo Boschini, Angelo Porrello, Davide Abati, and Simone Calderara. Dark experience for general continual learning: a strong, simple baseline.Advances in neural information processing systems, 33:15920–15930, 2020

work page 2020
[13]

Task-free continual learning

Rahaf Aljundi, Klaas Kelchtermans, and Tinne Tuytelaars. Task-free continual learning. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 11254–11263, 2019. 10

work page 2019
[14]

Learning without forgetting.IEEE transactions on pattern analysis and machine intelligence, 40(12):2935–2947, 2017

Zhizhong Li and Derek Hoiem. Learning without forgetting.IEEE transactions on pattern analysis and machine intelligence, 40(12):2935–2947, 2017

work page 2017
[15]

Continual learning on dynamic graphs via parameter isolation

Peiyan Zhang, Yuchen Yan, Chaozhuo Li, Senzhang Wang, Xing Xie, Guojie Song, and Sunghun Kim. Continual learning on dynamic graphs via parameter isolation. InProceedings of the 46th international ACM SIGIR conference on research and development in information retrieval, pages 601–611, 2023

work page 2023
[16]

Online continual learning in image classification: An empirical survey.Neurocomputing, 469:28–51, 2022

Zheda Mai, Ruiwen Li, Jihwan Jeong, David Quispe, Hyunwoo Kim, and Scott Sanner. Online continual learning in image classification: An empirical survey.Neurocomputing, 469:28–51, 2022

work page 2022
[17]

A topology-aware graph coarsening framework for continual graph learning.Advances in Neural Information Processing Systems, 37:132491– 132523, 2024

Xiaoxue Han, Zhuo Feng, and Yue Ning. A topology-aware graph coarsening framework for continual graph learning.Advances in Neural Information Processing Systems, 37:132491– 132523, 2024

work page 2024
[18]

Topology-aware embedding memory for continual learning on expanding networks

Xikun Zhang, Dongjin Song, Yixin Chen, and Dacheng Tao. Topology-aware embedding memory for continual learning on expanding networks. InProceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, pages 4326–4337, 2024

work page 2024
[19]

Hierarchical prototype networks for continual graph representation learning.IEEE Transactions on Pattern Analysis and Machine Intelligence, 45(4):4622–4636, 2022

Xikun Zhang, Dongjin Song, and Dacheng Tao. Hierarchical prototype networks for continual graph representation learning.IEEE Transactions on Pattern Analysis and Machine Intelligence, 45(4):4622–4636, 2022

work page 2022
[20]

Hero: Heterogeneous continual graph learning via meta-knowledge distillation.arXiv preprint arXiv:2505.17458, 2025

Guiquan Sun, Xikun Zhang, Jingchao Ni, and Dongjin Song. Hero: Heterogeneous continual graph learning via meta-knowledge distillation.arXiv preprint arXiv:2505.17458, 2025

work page arXiv 2025
[21]

Cat: Balanced continual graph learning with graph condensation

Yilun Liu, Ruihong Qiu, and Zi Huang. Cat: Balanced continual graph learning with graph condensation. In2023 IEEE International Conference on Data Mining (ICDM), pages 1157–

work page
[22]

Lifelong graph learning

Chen Wang, Yuheng Qiu, Dasong Gao, and Sebastian Scherer. Lifelong graph learning. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 13719–13728, 2022

work page 2022
[23]

Replay-and-forget-free graph class- incremental learning: A task profiling and prompting approach.Advances in Neural Information Processing Systems, 37:87978–88002, 2024

Chaoxi Niu, Guansong Pang, Ling Chen, and Bing Liu. Replay-and-forget-free graph class- incremental learning: A task profiling and prompting approach.Advances in Neural Information Processing Systems, 37:87978–88002, 2024

work page 2024
[24]

Class-domain incremental learning on graphs via disentangled knowledge distillation

Qin Tian, Chen Zhao, Xintao Wu, Dong Li, Minglai Shao, Xujiang Zhao, and Wenjun Wang. Class-domain incremental learning on graphs via disentangled knowledge distillation. In Proceedings of the ACM Web Conference 2026, pages 452–462, 2026

work page 2026
[25]

What matters in graph class incremental learning? an information preservation perspective.Advances in Neural Information Processing Systems, 37:26195–26223, 2024

Jialu Li, Yu Wang, Pengfei Zhu, Wanyu Lin, and Qinghua Hu. What matters in graph class incremental learning? an information preservation perspective.Advances in Neural Information Processing Systems, 37:26195–26223, 2024

work page 2024
[26]

Overcoming catastrophic forgetting in graph neural networks with experience replay

Fan Zhou and Chengtai Cao. Overcoming catastrophic forgetting in graph neural networks with experience replay. InProceedings of the AAAI conference on artificial intelligence, volume 35, pages 4714–4722, 2021

work page 2021
[27]

Streaming graph neural networks via continual learning

Junshan Wang, Guojie Song, Yi Wu, and Liang Wang. Streaming graph neural networks via continual learning. InProceedings of the 29th ACM international conference on information & knowledge management, pages 1515–1524, 2020

work page 2020
[28]

Overcoming catastrophic forgetting in graph neural networks

Huihui Liu, Yiding Yang, and Xinchao Wang. Overcoming catastrophic forgetting in graph neural networks. InProceedings of the AAAI conference on artificial intelligence, volume 35, pages 8653–8661, 2021

work page 2021
[29]

Sparsified subgraph memory for continual graph representation learning

Xikun Zhang, Dongjin Song, and Dacheng Tao. Sparsified subgraph memory for continual graph representation learning. In2022 IEEE International Conference on Data Mining (ICDM), pages 1335–1340. IEEE, 2022. 11

work page 2022
[30]

Ricci curvature-based graph sparsification for continual graph representation learning.IEEE Transactions on Neural Networks and Learning Systems, 35(12):17398–17410, 2023

Xikun Zhang, Dongjin Song, and Dacheng Tao. Ricci curvature-based graph sparsification for continual graph representation learning.IEEE Transactions on Neural Networks and Learning Systems, 35(12):17398–17410, 2023

work page 2023
[31]

Towards continuous reuse of graph models via holistic memory diversification

Ziyue Qiao, Junren Xiao, Qingqiang Sun, Meng Xiao, Xiao Luo, and Hui Xiong. Towards continuous reuse of graph models via holistic memory diversification. InThe Thirteenth International Conference on Learning Representations, 2025

work page 2025
[32]

Continual learning on graphs: Challenges, solutions, and opportunities.arXiv preprint arXiv:2402.11565, 2024

Xikun Zhang, Dongjin Song, and Dacheng Tao. Continual learning on graphs: Challenges, solutions, and opportunities.arXiv preprint arXiv:2402.11565, 2024

work page arXiv 2024
[33]

Online continual graph learning.arXiv preprint arXiv:2508.03283, 2025

Giovanni Donghi, Luca Pasa, Daniele Zambon, Cesare Alippi, and Nicolò Navarin. Online continual graph learning.arXiv preprint arXiv:2508.03283, 2025

work page arXiv 2025
[34]

Inductive representation learning on temporal graphs.arXiv preprint arXiv:2002.07962, 2020

Da Xu, Chuanwei Ruan, Evren Korpeoglu, Sushant Kumar, and Kannan Achan. Inductive representation learning on temporal graphs.arXiv preprint arXiv:2002.07962, 2020

work page arXiv 2002
[35]

Dysat: Deep neural representation learning on dynamic graphs via self-attention networks

Aravind Sankar, Yanhong Wu, Liang Gou, Wei Zhang, and Hao Yang. Dysat: Deep neural representation learning on dynamic graphs via self-attention networks. InProceedings of the 13th international conference on web search and data mining, pages 519–527, 2020

work page 2020
[36]

Simulating task-free continual learning streams from existing datasets

Aristotelis Chrysakis and Marie-Francine Moens. Simulating task-free continual learning streams from existing datasets. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 2516–2524, 2023

work page 2023
[37]

Andrew McCallum, Kamal Nigam, Jason D. M. Rennie, and Kristie Seymore. Automating the construction of internet portals with machine learning.Information Retrieval, 3:127–163, 2000

work page 2000
[38]

A critical look at the evaluation of GNNs under heterophily: Are we re- ally making progress? InThe Eleventh International Conference on Learning Representations, 2023

Oleg Platonov, Denis Kuznedelev, Michael Diskin, Artem Babenko, and Liudmila Prokhorenkova. A critical look at the evaluation of GNNs under heterophily: Are we re- ally making progress? InThe Eleventh International Conference on Learning Representations, 2023

work page 2023
[39]

Be- yond homophily in graph neural networks: Current limitations and effective designs.Advances in neural information processing systems, 33:7793–7804, 2020

Jiong Zhu, Yujun Yan, Lingxiao Zhao, Mark Heimann, Leman Akoglu, and Danai Koutra. Be- yond homophily in graph neural networks: Current limitations and effective designs.Advances in neural information processing systems, 33:7793–7804, 2020

work page 2020
[40]

Semi-Supervised Classification with Graph Convolutional Networks

Thomas N Kipf and Max Welling. Semi-supervised classification with graph convolutional networks.arXiv preprint arXiv:1609.02907, 2016

work page internal anchor Pith review Pith/arXiv arXiv 2016
[41]

Gradient episodic memory for continual learning

David Lopez-Paz and Marc’Aurelio Ranzato. Gradient episodic memory for continual learning. Advances in neural information processing systems, 30, 2017

work page 2017
[42]

Memory aware synapses: Learning what (not) to forget

Rahaf Aljundi, Francesca Babiloni, Mohamed Elhoseiny, Marcus Rohrbach, and Tinne Tuyte- laars. Memory aware synapses: Learning what (not) to forget. InProceedings of the European conference on computer vision (ECCV), pages 139–154, 2018. 12 A Details of DRIFT Benchmark A.1 Details of Benchmark Baselines A brief introduction of the implemented Continual Le...

work page 2018
[43]

Therefore, this can be viewed as the lower bound on the continual learning performance

Bare modeldenotes the backbone GNN without the continual learning technique. Therefore, this can be viewed as the lower bound on the continual learning performance

work page
[44]

We use Reservoir Sampling to select nodes

A-GEM [10]is an efficient version of GEM [ 41], which ensures that the average loss for historical tasks does not increase by projecting the gradient of incoming data onto the orthogonal space of the gradient of historical data. We use Reservoir Sampling to select nodes

work page
[45]

New incoming batches for training are then augmented with nodes sampled uniformly from the buffer

Experience Replay (ER) [ 5]selects nodes from the incoming batch to be stored in the memory buffer by Reservoir Sampling, which is a simple yet effective method for CL. New incoming batches for training are then augmented with nodes sampled uniformly from the buffer

work page
[46]

Specifically, it maintains samples whose gradients are less aligned with those already stored in the memory buffer, thereby promoting gradient diversity and reducing redundancy

Gradient-based Sample Selection (GSS) [ 11]selects representative samples from the incoming data stream by measuring the diversity of their gradients. Specifically, it maintains samples whose gradients are less aligned with those already stored in the memory buffer, thereby promoting gradient diversity and reducing redundancy. New batches for training are...

work page
[47]

Memory Aware Synapses (MAS)* [13]is a task-free version of MAS [ 42], which adds a detector guiding the model when to update the important weights in a streaming fashion

work page
[48]

It constructs sparsified subgraphs by selecting important nodes based on their contribution to the graph topology to reduce redundancy

Sparsified Subgraph Memory (SSM) [ 29]stores representative subgraphs instead of individual nodes to preserve both structural and feature information. It constructs sparsified subgraphs by selecting important nodes based on their contribution to the graph topology to reduce redundancy. Reservoir sampling is used as the sampling strategy

work page
[49]

Subgraph Episodic Memory (SEM) [30]extends subgraph-based memory by introducing a curvature-guided sparsification mechanism. It constructs Subgraph Episodic Memory (SEM) to store computation subgraphs, and further prunes edges based on Ricci curvature to preserve the most informative topological relationships for message passing. This approach reduces red...

work page
[50]

Roman Empire

Diversified Memory Selection and Generation (DMSG) [31]maintains a diversified mem- ory buffer by jointly considering intra- and inter-class diversity when selecting samples. To adequately reuse the knowledge preserved in the buffer, it utilizes a variational layer to gen- erate the distribution of buffer node embeddings and sample synthesized ones for re...

work page
[51]

Optimization details.All experiments use Adam with learning rate 5×10 −3

We do not use dropout or batch normalization. Optimization details.All experiments use Adam with learning rate 5×10 −3. The mini-batch size is fixed at B=10. Each incoming batch is processed for one epoch before the next batch arrives. No learning-rate scheduling, gradient clipping, or warm-up is applied. Method-specific settings.Whenever possible, we fol...

work page