Pith. sign in

REVIEW 3 major objections 44 references

Fine-tuning a language model only on the first task, then updating class anchors with uncertainty weights and graph topology, yields state-of-the-art continual node classification on streaming text-attributed graphs.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · grok-4.5

2026-07-14 13:50 UTC pith:46W4BEFK

load-bearing objection Solid incremental engineering on SimGCL: entropy-weighted EMA anchors + light structural features give consistent SOTA gains, but the first-task-only LoRA premise is untested. the 3 major comments →

arxiv 2607.10159 v1 pith:46W4BEFK submitted 2026-07-11 cs.AI

UNIT: Unleash Large Language Models Potential for Graph Continual Learning

Tairan Huang , Yili Wang , Beibei Hu , Yiting Shi , Qiutong Li , Changlong He , Jianliang Gao This is my paper
classification cs.AI
keywords graph continual learninglarge language modelstext-attributed graphsclass-incremental learningsemantic anchorsstructural confluenceknowledge transfercatastrophic forgetting
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

Graph data in the wild arrives as a stream of new classes and edges, so models must keep learning without erasing what they already know. Prior graph methods either ignore rich text semantics or forget early knowledge when later tasks appear. UNIT fine-tunes a large language model once with structured prompts on the first task, freezes it, and thereafter builds class prototypes (anchors) that mix text embeddings weighted by prediction uncertainty with simple topological features such as degree and clustering. These anchors softly accumulate history via exponential decay and directly become the classifier weights. On five standard benchmarks the approach raises both average accuracy and final-task accuracy above fifteen strong baselines, including the previous language-model method SimGCL. A reader who works with evolving social, citation or product graphs therefore gains a practical way to keep a single frozen model useful for years of new classes without storing old data or retraining from scratch.

Core claim

The authors show that a single LoRA fine-tune of an LLM on the first task, followed by continual construction of uncertainty-weighted semantic anchors that are fused with lightweight structural anchors, produces classifier weights that transfer knowledge across class-incremental graph tasks more effectively than regularization, replay or static-prototype alternatives, reaching state-of-the-art accuracy on Cora, Citeseer, WikiCS, Photo and Products.

What carries the argument

Uncertain-aware anchor generation (entropy-weighted soft average of LLM embeddings, decayed by factor λ) fused with structural confluence modeling (degree, average-neighbor-degree and clustering coefficient projected into the same space, decayed by γ) to form the final class weight w_c = β p_c + (1-β) a_c.

Load-bearing premise

Fine-tuning the language model only once, on the language and topology of the very first task, is enough to keep its embeddings reliable for every later task that arrives.

What would settle it

Construct a multi-task stream whose later tasks deliberately shift vocabulary domain or average degree by a large margin, re-run UNIT without re-tuning the LLM, and measure whether ACC_avg and ACC_N drop below a baseline that is allowed to re-fine-tune or retrain on each new task.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Streaming text-attributed graphs can be classified continually without storing historical node data or replaying subgraphs.
  • A single frozen LLM backbone can serve as a permanent feature extractor for an open-ended sequence of new classes once it has seen the first task.
  • Classifier weights that blend uncertainty-calibrated text prototypes with local topology remain stable under class-incremental arrival, reducing catastrophic forgetting relative to pure GNN or pure LLM baselines.
  • Few-shot variants of the same pipeline still outperform prior methods, indicating the anchors remain informative even when labeled data per task is scarce.

Where Pith is reading between the lines

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

  • If later tasks drift far outside the first-task language distribution, the fixed embeddings may become mis-calibrated and the entropy weights uninformative, suggesting a cheap periodic re-adaptation trigger could be added without abandoning the one-shot philosophy.
  • The same uncertainty-plus-structure anchor recipe could be tried on non-text multimodal streams (image-attributed or video-attributed graphs) once a suitable frozen encoder replaces the LLM.
  • Because the structural side uses only cheap local statistics, the method may remain practical on very large product or social graphs where full-neighborhood attention is prohibitive.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

3 major / 0 minor

Summary. The paper proposes UNIT, a graph continual learning framework that fine-tunes an LLM with LoRA only on the first class-incremental task, then freezes it and maintains class-level semantic anchors (entropy-weighted, exponentially decayed) and structural anchors (degree / average neighbor degree / clustering coefficient) that jointly set classifier weights. On five text-attributed graph benchmarks under the LLM4GCL class-incremental protocol, UNIT reports higher ACC_avg and ACC_N than fifteen baselines, including the prior GLM method SimGCL, with supporting ablations, few-shot results, backbone scaling, and hyper-parameter sweeps.

Significance. If the results hold under fair comparison, UNIT is a useful practical advance for class-incremental node classification on text-attributed graphs: a single first-task LoRA pass plus lightweight prototype maintenance is cheaper than per-task GNN retraining or full LLM re-tuning, and the gains over SimGCL are consistent across five datasets and few-shot settings. The ablation (Fig. 4), backbone table (Table 3), and λ/γ sensitivity (Fig. 5) give a reasonably complete empirical package for an applied ML contribution. The work does not claim formal guarantees; its value is empirical and systems-oriented.

major comments (3)
  1. Section 3.2 and Eq. (2): The central design choice—instruction-tuning LoRA only on T_s1, then freezing f_θ for all later tasks—is load-bearing for every subsequent module (Eqs. 3–7, 12–13). The manuscript never measures embedding or probability calibration drift across tasks, nor does it ablate re-tuning or multi-task LoRA. On the reported splits language may stay in-domain, but the SOTA attribution to “uncertain-aware” and “structural confluence” modules is conditional on an untested stationarity assumption. A minimal fix is either (i) a drift/calibration plot of h_i and α_i over task index, or (ii) an ablation that re-tunes or freezes after different first tasks, plus an explicit limitation statement.
  2. Section 3.4, Eq. (8): Structural confluence is claimed to fix “structural information degradation” and semantic–structural imbalance relative to SimGCL’s linearized prompts. The actual structural descriptor is only (degree, mean neighbor degree, clustering coefficient) projected by g(·). That is far weaker than the high-order topology the introduction criticizes as lost under linearization. The ablation “w/o Str” (Fig. 4) shows a drop, but does not show that these three scalars recover the missing geometric prior. Either strengthen the structural encoder (e.g., short random-walk / GNN features) or tone down the claim that confluence “explicitly integrates graph topology” in a way that resolves the stated structural degradation.
  3. Tables 1–2 and experimental protocol: Means ± std are reported, but the number of random seeds / runs is not stated in the main text (only “presumably multiple runs”). Without N and a clear statement that all methods share the same task splits, LLM backbone (LLaMA3-8B), and evaluation protocol as LLM4GCL, the several-point margins over SimGCL cannot be assessed for significance. Please report N, seed list, and any shared codebase settings.

Circularity Check

0 steps flagged

No circularity: empirical prototype-based continual learning method whose anchors and classifier weights are constructed from model outputs and then evaluated on held-out sequential tasks.

full rationale

UNIT is an engineering/empirical paper. Semantic anchors p_c^(T_si) (Eqs. 5–7) and structural anchors a_c (Eqs. 9–11) are defined as (decayed, uncertainty-weighted) averages of the frozen LLM embeddings and hand-crafted topological descriptors of the labeled nodes of the current task; the classifier weights are then the convex combination w_c = β p_c + (1-β) a_c (Eq. 12). Classification accuracy on subsequent tasks is measured against external baselines (Table 1). Nothing is predicted that is algebraically forced by the definition of the anchors, no parameter is fitted to a target quantity and then re-reported as a prediction, and no uniqueness theorem or ansatz is imported via self-citation to force the design. Hyper-parameters λ, γ are fixed once and subjected to a sensitivity sweep (Fig. 5); the single LoRA fine-tune on T_s1 is an explicit modeling choice whose validity is tested by the multi-task accuracy numbers, not assumed by construction. The derivation chain is therefore self-contained and non-circular.

Axiom & Free-Parameter Ledger

3 free parameters · 4 axioms · 2 invented entities

The central performance claim rests on three free decay/mix parameters, standard continual-learning and TAG assumptions, and two invented algorithmic constructs (the uncertainty-aware anchor and the structural confluence module) that have no independent existence outside this paper.

free parameters (3)
  • λ (semantic decay) = 0.7
    Exponential decay factor for historical semantic anchors (Eq. 6–7); set to 0.7 by hand; controls how much early-task knowledge is retained.
  • γ (structural decay) = 0.6
    Exponential decay factor for historical structural anchors (Eq. 10–11); set to 0.6 by hand.
  • β (semantic/structural mix)
    Convex combination weight that forms final classifier weights (Eq. 12); range [0,1] but never ablated or given a default value in the text.
axioms (4)
  • domain assumption Node-level class-incremental learning on text-attributed graphs with disjoint class sets across tasks is the correct formalization of graph continual learning.
    Stated in §3.1 and taken from the LLM4GCL benchmark; all experiments and claims rest on this protocol.
  • ad hoc to paper A single LoRA fine-tune on the first task’s instruction prompts is sufficient to adapt a pre-trained LLM for all future tasks.
    Core design choice of §3.2; if later tasks have large distribution shift the frozen embeddings become unreliable.
  • domain assumption Prediction entropy of the LLM is a reliable proxy for sample uncertainty that can be used as a confidence weight.
    Used to define α_i in Eq. 4–5; standard in uncertainty literature but not re-validated here.
  • ad hoc to paper Degree, average neighbor degree and clustering coefficient are sufficient structural descriptors for class-level confluence.
    Eq. 8; the paper criticizes linearization for losing high-order topology yet replaces it with these three scalar features.
invented entities (2)
  • uncertain-aware semantic anchor no independent evidence
    purpose: Class prototype that softly merges new embeddings weighted by 1−normalized entropy with an exponentially decayed historical prototype, used as classifier weight.
    Defined in §3.3 (Eqs. 5–7); no independent existence or external measurement outside the UNIT pipeline.
  • structural confluence modeling / structural anchor no independent evidence
    purpose: Parallel class prototype built from local topology statistics and mixed with the semantic anchor to form final classifier weights.
    Defined in §3.4 (Eqs. 8–12); introduced solely to address the paper’s own “semantic-structural separation” critique.

pith-pipeline@v1.1.0-grok45 · 42928 in / 3136 out tokens · 33815 ms · 2026-07-14T13:50:16.663509+00:00 · methodology

0 comments
read the original abstract

In real-world multimodal web scenarios, graph-structured data often arrives in a streaming manner, making graph continual learning a crucial paradigm for continuously modeling such evolving structures. However, existing graph continual learning methods still face two fundamental challenges. 1) semantic-structural separation, where the graph-based methods excel at modeling topological relationships but neglect deep semantics. 2) imbalanced knowledge transfer, where existing models fail to effectively leverage general knowledge gained from early tasks to benefit subsequent new tasks. To address above issues, we propose a novel framework, \textbf{UN}leash Large Language Models PotentIal for Graph ConTinual Learning (UNIT). By fine-tuning large language model only on the first task, we bridge the distributional gap between the pre-trained LLM corpus and the target task dataset to enhance the adaptability of LLMs for graph-structured tasks. Meanwhile, we propose an uncertain-aware anchor generation mechanism to effectively preserve representative knowledge across tasks, avoiding the neglect of universal knowledge learned from previous tasks. Additionally, we introduce structural confluence modeling to explicitly integrates graph topology information into semantic information, enhancing the collaborative capabilities between semantic understanding and structural modeling. Extensive experiments demonstrate that our proposed method achieves state-of-the-art performance in the graph continual learning task.

Figures

Figures reproduced from arXiv: 2607.10159 by Beibei Hu, Changlong He, Jianliang Gao, Qiutong Li, Tairan Huang, Yili Wang, Yiting Shi.

Figure 1
Figure 1. Figure 1: Workflow comparison. The existing graph contin [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: The overall framework of Unleash Large Language Models Potential for Graph Continual Learning (UNIT). The [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: The performance comparison between UNIT and state-of-the-art methods in three baseline categories with ACC [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 5
Figure 5. Figure 5: Parameter analysis of 𝜆 and 𝛾 on Cora dataset with ACCavg and ACCN metrics. Note that the values of 𝜆 and 𝛾 range from [0.1, . . ., 0.9]. and structural confluence modeling module play critical roles in UNIT and their combination greatly benefits the graph continual learning. In general, each component within UNIT contributes to unleashing LLMs potential for graph continual learning. 4.5 Parameter Analysis… view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

44 extracted references · 7 linked inside Pith

  1. [1]

    Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, et al. 2020. Language models are few-shot learners.Advances in Neural Information Processing Systems33 (2020), 1877–1901

    2020

  2. [2]

    Qiang Chen, Zhongze Wu, Xiu Su, Xi Lin, Zhe Qu, Shan You, Shuo Yang, and Chang Xu. 2025. Stable Fair Graph Representation Learning with Lipschitz Constraint. InProceedings of the International Conference on Machine Learning

    2025

  3. [3]

    Runjin Chen, Tong Zhao, Ajay Jaiswal, Neil Shah, and Zhangyang Wang. 2024. Llaga: Large language and graph assistant.arXiv preprint arXiv:2402.08170(2024)

    Pith/arXiv arXiv 2024

  4. [4]

    Zhikai Chen, Haitao Mao, Hang Li, Wei Jin, Hongzhi Wen, Xiaochi Wei, Shuaiqiang Wang, Dawei Yin, Wenqi Fan, Hui Liu, et al . 2024. Exploring the potential of large language models (llms) in learning on graphs.ACM SIGKDD Explorations Newsletter25, 2 (2024), 42–61

    2024

  5. [5]

    Ziyang Cheng, Zhixun Li, Yuhan Li, Yixin Song, Kangyi Zhao, Dawei Cheng, Jia Li, and Jeffrey Xu Yu. 2025. Can LLMs Alleviate Catastrophic Forgetting in Graph Continual Learning? A Systematic Study.arXiv preprint arXiv:2505.18697(2025)

    arXiv 2025

  6. [6]

    Seungyoon Choi, Wonjoong Kim, Sungwon Kim, Yeonjun In, Sein Kim, and Chanyoung Park. 2024. DSLR: diversity enhancement and structure learning for rehearsal-based graph continual learning. InProceedings of the ACM Web Conference 2024. 733–744

    2024

  7. [7]

    Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. Bert: Pre-training of deep bidirectional transformers for language understanding. In Proceedings of the Conference of the North American Chapter of the Association for Computational Linguistics: human language technologies, volume 1 (long and short papers). 4171–4186

    2019

  8. [8]

    Xiaolin Han, Yikun Zhang, Chenhao Ma, Lingyun Song, Reynold Cheng, and Xuequn Shang. 2025. TempASD: Temporal Anomalous Subgraph Discovery in Large-Scale Dynamic Financial Networks. InProceedings of the ACM SIGKDD Conference on Knowledge Discovery and Data Mining V. 2. 826–837

    2025

  9. [9]

    Xiaoxin He, Xavier Bresson, Thomas Laurent, Adam Perold, Yann LeCun, and Bryan Hooi. 2023. Harnessing explanations: Llm-to-lm interpreter for enhanced text-attributed graph representation learning.arXiv preprint arXiv:2305.19523 (2023)

    Pith/arXiv arXiv 2023

  10. [10]

    Edward J Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, Weizhu Chen, et al. 2022. Lora: Low-rank adaptation of large language models.. InInternational Conference on Learning Representations. 3

    2022

  11. [11]

    Weihua Hu, Matthias Fey, Marinka Zitnik, Yuxiao Dong, Hongyu Ren, Bowen Liu, Michele Catasta, and Jure Leskovec. 2020. Open graph benchmark: Datasets for machine learning on graphs.Advances in Neural Information Processing Systems 33 (2020), 22118–22133

    2020

  12. [12]

    Huang, Wang Tairan, Li Yili, He Qiutong, Gao Changlong, and Jianliang. 2025. Can LLMs Find Fraudsters? Multi-level LLM Enhanced Graph Fraud Detection. InProceedings of the ACM International Conference on Multimedia

    2025

  13. [13]

    Ching-Yi Hung, Cheng-Hao Tu, Cheng-En Wu, Chien-Hung Chen, Yi-Ming Chan, and Chu-Song Chen. 2019. Compacting, picking and growing for unforgetting continual learning.Advances in Neural Information Processing Systems32 (2019)

    2019

  14. [14]

    Pengyue Jia, Jingtong Gao, Yejing Wang, Yuhao Wang, Xiaopeng Li, Qidong Liu, Yichao Wang, Bo Chen, Huifeng Guo, and Ruiming Tang. 2025. Joint Modeling in Deep Recommender Systems. InCompanion Proceedings of the ACM on Web Conference. 17–20

    2025

  15. [15]

    Thomas N Kipf and Max Welling. 2017. Semi-supervised classification with graph convolutional networks. InInternational Conference on Learning Representations. 1031–1049

    2017

  16. [16]

    James Kirkpatrick, Razvan Pascanu, Neil Rabinowitz, Joel Veness, Guillaume Desjardins, Andrei A Rusu, Kieran Milan, John Quan, Tiago Ramalho, Agnieszka Grabska-Barwinska, et al. 2017. Overcoming catastrophic forgetting in neural networks.Proceedings of the National Academy of Sciences114, 13 (2017), 3521– 3526

    2017

  17. [17]

    Yuhan Li, Peisong Wang, Xiao Zhu, Aochuan Chen, Haiyun Jiang, Deng Cai, Victor W Chan, and Jia Li. 2024. Glbench: A comprehensive benchmark for graph with large language models.Advances in Neural Information Processing Systems 37 (2024), 42349–42368

    2024

  18. [18]

    Zhizhong Li and Derek Hoiem. 2017. Learning without forgetting.IEEE Transac- tions on Pattern Analysis and Machine Intelligence40, 12 (2017), 2935–2947

    2017

  19. [19]

    Yan-Shuo Liang and Wu-Jun Li. 2024. Inflora: Interference-free low-rank adapta- tion for continual learning. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 23638–23647

    2024

  20. [20]

    Hao Liu, Jiarui Feng, Lecheng Kong, Ningyue Liang, Dacheng Tao, Yixin Chen, and Muhan Zhang. 2023. One for all: Towards training one graph model for all classification tasks.arXiv preprint arXiv:2310.00149(2023)

    Pith/arXiv arXiv 2023

  21. [21]

    Yinhan Liu, Myle Ott, Naman Goyal, Jingfei Du, Mandar Joshi, Danqi Chen, Omer Levy, Mike Lewis, Luke Zettlemoyer, and Veselin Stoyanov. 2019. Roberta: A robustly optimized bert pretraining approach.arXiv preprint arXiv:1907.11692 (2019)

    Pith/arXiv arXiv 2019

  22. [22]

    Zheyuan Liu, Xiaoxin He, Yijun Tian, and Nitesh V Chawla. 2024. Can we soft prompt llms for graph learning tasks?. InCompanion Proceedings of the ACM Web Conference. 481–484

    2024

  23. [23]

    2021.Deep learning on graphs

    Yao Ma and Jiliang Tang. 2021.Deep learning on graphs. Cambridge University Press

    2021

  24. [24]

    Chaoxi Niu, Guansong Pang, Ling Chen, and Bing Liu. 2024. Replay-and-forget- free graph class-incremental learning: A task profiling and prompting approach. Advances in Neural Information Processing Systems37 (2024), 87978–88002

    2024

  25. [25]

    Sylvestre-Alvise Rebuffi, Hakan Bilen, and Andrea Vedaldi. 2017. Learning mul- tiple visual domains with residual adapters.Advances in Neural Information Processing Systems30 (2017)

    2017

  26. [26]

    Yanchao Tan, Hang Lv, Pengxiang Zhan, Shiping Wang, and Carl Yang. 2025. Graph-oriented Instruction Tuning of Large Language Models for Generic Graph Mining.IEEE Transactions on Pattern Analysis and Machine Intelligence(2025)

    2025

  27. [27]

    Jiabin Tang, Yuhao Yang, Wei Wei, Lei Shi, Lixin Su, Suqi Cheng, Dawei Yin, and Chao Huang. 2024. Graphgpt: Graph instruction tuning for large language models. InProceedings of the 47th International ACM SIGIR Conference on Research and Development in Information Retrieval. 491–500

    2024

  28. [28]

    Hugo Touvron, Louis Martin, Kevin Stone, Peter Albert, Amjad Almahairi, Yas- mine Babaei, Nikolay Bashlykov, Soumya Batra, Prajjwal Bhargava, Shruti Bhos- ale, et al. 2023. Llama 2: Open foundation and fine-tuned chat models.arXiv preprint arXiv:2307.09288(2023)

    Pith/arXiv arXiv 2023

  29. [29]

    Duo Wang, Yuan Zuo, Fengzhi Li, and Junjie Wu. 2024. Llms as zero-shot graph learners: Alignment of gnn representations with llm token embeddings.Advances in Neural Information Processing Systems37 (2024), 5950–5973

    2024

  30. [30]

    Qi-Wei Wang, Da-Wei Zhou, Yi-Kai Zhang, De-Chuan Zhan, and Han-Jia Ye

  31. [31]

    Advances in Neural Information Processing Systems36 (2023), 15060–15076

    Few-shot class-incremental learning via training-free prototype calibration. Advances in Neural Information Processing Systems36 (2023), 15060–15076

    2023

  32. [32]

    Xixi Wu, Yifei Shen, Fangzhou Ge, Caihua Shan, Yizhu Jiao, Xiangguo Sun, and Hong Cheng. 2025. A comprehensive analysis on llm-based node classification algorithms.arXiv preprint arXiv:2502.00829(2025)

    Pith/arXiv arXiv 2025

  33. [33]

    Hao Yan, Chaozhuo Li, Ruosong Long, Chao Yan, Jianan Zhao, Wenwen Zhuang, Jun Yin, Peiyan Zhang, Weihao Han, Hao Sun, et al. 2023. A comprehensive study on text-attributed graphs: Benchmarking and rethinking.Advances in Neural Information Processing Systems36 (2023), 17238–17264

    2023

  34. [34]

    Hao Yan, Chaozhuo Li, Jun Yin, Zhigang Yu, Weihao Han, Mingzheng Li, Zhengxin Zeng, Hao Sun, and Senzhang Wang. 2025. When Graph Meets Multi- modal: Benchmarking and Meditating on Multimodal Attributed Graph Learning. InProceedings of the ACM SIGKDD Conference on Knowledge Discovery and Data Mining V. 2. 5842–5853

    2025

  35. [35]

    Mengmei Zhang, Mingwei Sun, Peng Wang, Shen Fan, Yanhu Mo, Xiaoxiao Xu, Hong Liu, Cheng Yang, and Chuan Shi. 2024. Graphtranslator: Aligning graph model to large language model for open-ended tasks. InProceedings of the ACM Web Conference 2024. 1003–1014

    2024

  36. [36]

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

    2023

  37. [37]

    Taiyan Zhang, Renchi Yang, Yurui Lai, Mingyu Yan, Xiaochun Ye, and Dongrui Fan. 2025. Leveraging large language models for effective label-free node classifi- cation in text-attributed graphs. InProceedings of the International ACM SIGIR Conference on Research and Development in Information Retrieval. 698–708

    2025

  38. [38]

    Xikun Zhang, Dongjin Song, and Dacheng Tao. 2022. Hierarchical prototype networks for continual graph representation learning.IEEE Transactions on Pattern Analysis and Machine Intelligence45, 4 (2022), 4622–4636

    2022

  39. [39]

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

    2022

  40. [40]

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

    2023

  41. [41]

    Da-Wei Zhou, Zi-Wen Cai, Han-Jia Ye, De-Chuan Zhan, and Ziwei Liu. 2025. Revisiting class-incremental learning with pre-trained models: Generalizability and adaptivity are all you need.International Journal of Computer Vision133, 3 (2025), 1012–1032

    2025

  42. [42]

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

    2021

  43. [43]

    Hao Zhu, Yifei Zhang, Junhao Dong, and Piotr Koniusz. 2025. BiLoRA: Almost- Orthogonal Parameter Spaces for Continual Learning. InProceedings of the Com- puter Vision and Pattern Recognition Conference. 25613–25622

    2025

  44. [44]

    Yun Zhu, Yaoke Wang, Haizhou Shi, and Siliang Tang. 2024. Efficient tun- ing and inference for large language models on textual graphs.arXiv preprint arXiv:2401.15569(2024)

    Pith/arXiv arXiv 2024