Pith. sign in

REVIEW 3 major objections 6 minor 1 cited by

GraphICL: Unlocking Graph Learning Potential in LLMs through Structured Prompt Design

T0 review · 3 major / 6 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read Structured prompt templates let frozen LLMs beat trained graph models in low-label and cross-domain settings.

desk verdict Useful prompt-suite artifact undermined by post-hoc test-set selection and target-label leakage in the cross-domain comparison; worth refereeing but the empirical claims need a major rework. read the letter →

arxiv 2501.15755 v1 pith:JYXAJXE5 submitted 2025-01-27 cs.LG

Yuanfu Sun , Zhengnan Ma , Yi Fang , Jing Ma , Qiaoyu Tan This is my paper · ORCID
classification cs.LG
keywords graphin-contextlearningtext-attributedgraphspromptengineeringnodeclassificationlinkpredictionlargelanguagemodelssemi-supervisedreasoning
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper's central claim is that a general-purpose LLM, given the right prompt template, can do graph reasoning tasks on text-attributed graphs as well as or better than specialized graph LLMs and graph neural networks that have been trained for those tasks. The authors build a benchmark of 55 prompt templates assembled from four components: task description, anchor node text, structure-aware neighbor information, and labeled demonstrations. Across nine datasets, the best GraphICL prompts improve accuracy by roughly 20 percent relative over specialized baselines in semi-supervised node classification and also win in out-of-domain tests without any weight updates. The intended message is a warning and a baseline: prior comparisons that pitted trained graph LLMs against naive zero-shot LLM prompts understated what frozen LLMs can do. If the claim holds, prompt design is a serious alternative to instruction tuning for graph learning under label scarcity.

What carries the argument

The central object is the GraphICL prompt template, which combines four components: task description, anchor node text, structure-aware neighbor information, and labeled demonstrations. Structure-aware information simulates message passing at the text level by adding 1-hop or 2-hop neighbor texts, selected randomly, by cosine similarity, or by PageRank. Demonstrations are chosen globally or class-aware using the same three selection strategies, and the paper evaluates all 55 combinations. The mechanism is that a frozen LLM performs in-context inference from these textual signals alone; no gradient updates are applied.

What would settle it

Run the cross-domain comparison again with GraphICL demonstrations restricted to the source or mixed training sets only, with no target-distribution labels in the prompt. If accuracy then falls to or below the specialized graph LLMs, the reported out-of-domain advantage is an artifact of in-prompt label leakage rather than graph reasoning.

Watch

Extended reading notes

Core claim

GraphICL establishes that in-context prompting can release graph reasoning from frozen LLMs. The paper reports that LLaMA3-70B-Instruct with GraphICL reaches 87.37 percent accuracy on Amazon-Computers, 91.39 percent on Amazon-Sports, 93.18 percent on PubMed, 83.58 percent on Cora, 73.68 percent on OGB-Arxiv, and 81.48 percent on OGB-Products in semi-supervised node classification, beating trained GNNs and specialized graph LLMs. In link prediction on Cora it reports 88.08 percent, above the specialized LLaGA-HO's 86.82 percent in the supervised setup. The authors interpret these numbers as evidence that the benefit of specialized graph LLMs has been overstated because their baseline comparisons used simplistic prompts; with structured prompts, general LLMs become the stronger option.

Load-bearing premise

In the cross-domain experiments, the strongest GraphICL prompts draw their few-shot demonstrations from the target dataset's own training labels, while the specialized graph LLMs must work without any target labels; the claimed out-of-domain advantage depends on treating that label access as fair rather than as leakage.

Editorial extensions

If this is right

  • Specialized graph LLMs should be re-evaluated against structured prompts rather than naive zero-shot prompts before their gains are attributed to training.
  • In label-scarce settings, prompt engineering can replace training: GraphICL's semi-supervised results match or beat supervised trained models on several datasets.
  • Cross-domain deployment can be handled by a frozen LLM plus in-prompt labeled demonstrations, avoiding retraining when the target graph changes.
  • Similarity-based neighbor selection and class-aware demonstrations are the most consistently useful prompt choices, while chain-of-thought prompting is unreliable and often neutral or harmful.
  • The link prediction results indicate the template transfers beyond node classification to edge-level reasoning tasks.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If the cross-domain gains come mainly from in-prompt demonstrations sampled from the target dataset's own training labels, the reported out-of-domain superiority is closer to in-context label propagation than to transferable graph reasoning; re-running with only source-domain demonstrations would separate these explanations.
  • The 55-template sweep treats prompt design as a searchable hyperparameter space, so an automated search over neighbor and demonstration selection strategies could extend the benchmark further.
  • Similarity-based neighbor selection is likely exploiting label homophily, which suggests a testable prediction: GraphICL should degrade on heterophilous graphs where neighbors frequently carry different labels.
  • The strong semi-supervised performance hints that frozen LLMs may be effective as data-efficient annotators or priors for graph models, a use the paper leaves implicit.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 6 minor

Summary. The paper introduces GraphICL, a set of 55 prompt templates for using frozen LLMs on text-attributed graph tasks, with components for task description, anchor-node text, structure-aware neighbor information, and few-shot demonstrations. The authors evaluate LLaMA2-13B-Chat, LLaMA3-70B-Instruct, and GPT-4o on node classification and link prediction across 9 datasets, comparing against GNNs, specialized graph LLMs, and zero-shot LLMs, and claim that prompt design alone lets general-purpose LLMs outperform specialized graph LLMs in both in-domain and out-of-domain settings. The paper also reports ablations of neighbor-selection strategy, demonstration-selection strategy, structural depth, and chain-of-thought prompting.

Significance. If the claims held, the paper would be an important result: it would show that a carefully designed set of prompts can close or reverse the gap between frozen general-purpose LLMs and specialized graph LLMs, with implications for how graph LLM research is evaluated. The taxonomy of prompt components, the 55-template design space, and the ablations are potentially useful resources for future benchmarking. However, the central comparison is currently undermined by two evaluation-protocol problems: the reported GraphICL numbers appear to be per-dataset optima over 55 template choices, and the cross-domain experiments allow GraphICL to use labeled demonstrations from the target dataset while the graph-LLM baselines have no such access. These issues mean the headline claims of in-domain and out-of-domain superiority are not established as written. The empirical infrastructure could still be valuable after a corrected evaluation with validation-based prompt selection and symmetric information access.

major comments (3)
  1. [Section 5.2.1, Appendix A.2, Table 8, Figure 3] The reported GraphICL results for each dataset are the first- and second-best of 55 prompt templates, but no validation-based selection procedure is described. The caption of Table 1 states that S1 and S2 denote the first- and second-best GraphICL prompts, and Appendix A.3.4 refers to "the best results from the 55 configurations tested." If the best configuration was chosen using test-set accuracy, then the accuracies in Tables 1, 2, 10, 11, and 12 are post-hoc optima and do not represent a single well-defined prompting method. Comparisons against fixed graph-LLM baselines are inflated by this selection. Please either select prompts on a validation split, fix one template per task in advance, or report the full distribution of accuracies across all 55 configurations, and adjust the conclusions accordingly.
  2. [Section 5.2.2, Table 2, Appendix A.3.4, Table 8] The cross-domain comparison is not out-of-domain for GraphICL. Table 8 shows that the settings used for Table 2 are the same S1/S2 settings as in the in-domain experiments, and Appendix A.2 defines "G" and "C" demonstrations as selected from the training set. Thus, for example, the "1RCP" setting on Computers and Cora uses class-aware PageRank demonstrations drawn from the target dataset's training split. The graph-LLM baselines, by contrast, were trained only on other datasets and have no access to target labels. GraphICL therefore receives in-distribution labeled examples that the baselines do not, so the large relative gains in Table 2 (e.g., +487% on Computers and +2459% on Sports) are consistent with label-information asymmetry rather than with graph reasoning ability. The cross-domain experiment must be rerun with demonstrations restricted to source-domain data, or the baselines must be given the same few-shot target examples, before any out-of-domain superiority claim can be made.
  3. [Abstract and Section 5.2.2, Observation 2] The central claim that GraphICL outperforms specialized graph LLMs in resource-constrained settings and out-of-domain tasks rests on the two evaluation issues above. In particular, Observation 2 and the abstract should not state that GraphICL enables LLaMA to outperform specialized graph LLMs without training until the test-set prompt selection and the target-label demonstration advantage are addressed. The relative improvement percentages over near-zero baselines (e.g., +2459.94% on Sports) are also misleading; please report absolute accuracy differences alongside any relative percentages.
minor comments (6)
  1. [Section 5.1] There is a typo in "Ad dtional details" in the opening of Section 5.1; it should read "Additional details."
  2. [Table 1 and Appendix A.2] The notation S1/S2 and template codes such as "1RCP" are used in the main text but defined only in the appendix; please define or briefly explain the notation in the main text for readability.
  3. [Tables 1, 2, and 10-12] Relative improvements over very small baselines are hard to interpret; please also report absolute accuracy gaps, especially in the cross-domain tables.
  4. [Problem Statement and Section 4] The paper alternates between the terms "GICL" and "GraphICL" without a clear distinction; please make the terminology consistent throughout.
  5. [Appendix A.3] For reproducibility of a benchmark paper, please state whether code and configuration files for all 55 templates, prompt texts, and sampling procedures will be released.
  6. [Table 13] The OGB-Products system prompt in Table 13 says "(Categories omitted due to length)"; the actual category list is needed for reproducibility and should be provided in full or in a supplementary file.

Circularity Check

2 steps flagged · score 6.0 of 10

Best-of-55 test-set prompt selection and target-training-set demonstrations make the headline GraphICL comparisons partially fitted rather than predictive.

  1. fitted input called prediction [Table 1 caption; Appendix A.2; Table 8]
    "For each LLaMA model in GraphICL, S1 and S2 denote the first- and second-best GraphICL prompts. ... In total, there are 55 possible combinations: 7 × 7 + 1× 6 = 55methods."

    Reported GraphICL accuracy is, by the paper's own definition, the maximum over 55 prompt templates evaluated on the same test split. S1/S2 are chosen as first- and second-best after seeing test accuracy, so the headline comparisons are best-case selections rather than predictions of a fixed, pre-specified method. The claim 'GraphICL outperforms' therefore reduces to 'the best of 55 test-set-tuned prompts outperforms', which is a fitted evaluation protocol renamed as a prediction.

  2. other [Section 5.2.2 (Observation 2); Appendix A.2; Table 8]
    "For the GICL settings we used in Table 1 and Table 2, please refer to Table 8. ... 'G' indicates that demonstrations are selected without regard to labels, while 'C' ensures that one demonstration is selected per label from the training set."

    Section 5.2 defines the cross-domain scenario as test datasets having no overlap with the training data, but the Table 2 GraphICL runs reuse the Table 8 settings, whose few-shot demonstrations are drawn from the target dataset's training split (e.g., 1RGR and 1RCP on Cora and Arxiv). Thus, in the 'out-of-domain' comparison, GraphICL receives in-distribution labeled examples while the specialized graph-LLM baselines have no access to target labels. The reported superiority is therefore not a prediction from out-of-domain inputs; it reduces to few-shot target-label prompting versus zero-shot transfer, making the out-of-domain claim an artifact of label leakage rather than graph reasoning.

full rationale

The prompt-design benchmark itself is self-contained: the LLM weights are not trained, and the paper does ablate structure-aware information, neighbor selection, demonstrations, and chain-of-thought. However, the central comparative claims rest on two evaluation choices that make the reported 'predictions' depend on target labels. First, S1/S2 are selected as the first- and second-best of 55 prompt configurations per dataset after evaluating on the test set, so the reported accuracy is a test-label-driven maximum rather than a fixed-method prediction. Second, the cross-domain table reuses these in-domain settings, and Table 8's configurations sample demonstrations from the target dataset's training set, giving GraphICL target-distribution labels that the specialized graph-LLM baselines never see. These issues partially reduce the headline superiority claims to selection and leakage effects, although the underlying LLM generalization signal remains real. Hence partial circularity, scored 6.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The central results rest on assumptions about LLM prompting and graph structure; no new physical or mathematical entities are introduced. The main free parameters are prompt configuration choices selected on test data, and the axioms are standard domain assumptions about homophily, embedding similarity, and LLM in-context behavior.

free parameters (4)
  • Per-dataset best prompt configuration (S1/S2), selected from 55 templates on test set = e.g., 1RCP for Computers with LLaMA3 S1
    The reported GraphICL accuracy is the best or second-best of 55 prompt configurations evaluated on the test set with no validation-based selection, inflating apparent performance.
  • Number of neighbors and demonstrations M = 6 for node classification; 6 neighbors plus 3 demonstrations for link prediction
    Set by context-window constraint (Appendix A.3.3), not by systematic study; results may be sensitive to this choice.
  • Hop count for structure-aware information = 1 or 2
    Best hop count chosen per dataset from the 55-template grid; 2-hop generally helps citation networks.
  • Neighbor and demonstration selection strategy = random, PageRank, or similarity
    Three strategies tested; similarity-based selection reported as best on average, but per-dataset best varies; no statistical comparison.
assumptions (5)
  • domain assumption LLMs can perform in-context learning from textual node descriptions and neighbor lists
    The entire GraphICL method assumes frozen LLMs can use provided text to classify nodes and predict links, invoked throughout Sections 4 and 5.
  • domain assumption Graph homophily: neighboring nodes tend to share labels
    Structure-aware prompting adds neighbor text on the assumption that neighbors inform the anchor node's label (Sections 4.2, 5.3.2).
  • domain assumption GIA text embeddings capture label-relevant similarity
    Similarity-based neighbor and demonstration selection uses GIA embeddings (Appendix A.3.3), assuming cosine similarity in this embedding space aligns with label relevance.
  • domain assumption PageRank importance correlates with demonstration usefulness
    PageRank-based selection treats structurally central nodes as better demonstrations (Section 4.2).
  • domain assumption LLM pretraining does not memorize benchmark labels
    The paper does not test for label memorization or contamination for classic datasets (Cora, PubMed, OGB-Arxiv), which could inflate results independent of prompting.

how reviews work

0 comments
Cite this review

Pith. "Pith review of GraphICL: Unlocking Graph Learning Potential in LLMs through Structured Prompt Design." pith.science (2026). https://pith.science/paper/JYXAJXE5

@misc{pith2026250115755,
  author       = {Pith},
  title        = {Pith review of: GraphICL: Unlocking Graph Learning Potential in LLMs through Structured Prompt Design},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/JYXAJXE5}},
  note         = {Machine review of arXiv:2501.15755}
}
read the original abstract

The growing importance of textual and relational systems has driven interest in enhancing large language models (LLMs) for graph-structured data, particularly Text-Attributed Graphs (TAGs), where samples are represented by textual descriptions interconnected by edges. While research has largely focused on developing specialized graph LLMs through task-specific instruction tuning, a comprehensive benchmark for evaluating LLMs solely through prompt design remains surprisingly absent. Without such a carefully crafted evaluation benchmark, most if not all, tailored graph LLMs are compared against general LLMs using simplistic queries (e.g., zero-shot reasoning with LLaMA), which can potentially camouflage many advantages as well as unexpected predicaments of them. To achieve more general evaluations and unveil the true potential of LLMs for graph tasks, we introduce Graph In-context Learning (GraphICL) Benchmark, a comprehensive benchmark comprising novel prompt templates designed to capture graph structure and handle limited label knowledge. Our systematic evaluation shows that general-purpose LLMs equipped with our GraphICL outperform state-of-the-art specialized graph LLMs and graph neural network models in resource-constrained settings and out-of-domain tasks. These findings highlight the significant potential of prompt engineering to enhance LLM performance on graph learning tasks without training and offer a strong baseline for advancing research in graph LLMs.

Figures

Figures reproduced from arXiv: 2501.15755 by the authors.

Figure 1
Figure 1. The overall framework of our GraphICL. We implement various graph in-context learning templates by [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 3
Figure 3. The comparison heat map between the GraphICL method based on LLaMA3-70B-Instruct and the best [PITH_FULL_IMAGE:figures/full_fig_p016_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. MLaGA: Multimodal Large Language and Graph Assistant

    cs.AI 2025-06 conditional novelty 6.0 of 10

    MLaGA extends LLM-based graph reasoning from text-only graphs to multimodal graphs with image and text node attributes via a structure-aware aligner and multimodal instruction tuning.

Reference graph

Works this paper leans on

56 extracted references · 22 canonical work pages · cited by 1 Pith paper

  1. [1]

    online" 'onlinestring :=

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint eprinttype howpublished institution journal key month note number organization pages publisher school series title type volume year doi pubmed url lastchecked label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block STRING...

  2. [2]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...

  3. [3]

    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 systems, 33:1877--1901

    2020

  4. [4]

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

    arXiv 2024

  5. [5]

    Runjin Chen, Tong Zhao, AJAY KUMAR JAISWAL, Neil Shah, and Zhangyang Wang. 2024 b . Llaga: Large language and graph assistant. In Forty-first International Conference on Machine Learning

    work page 2024

  6. [6]

    Zhikai Chen, Haitao Mao, Hang Li, Wei Jin, Hongzhi Wen, Xiaochi Wei, Shuaiqiang Wang, Dawei Yin, Wenqi Fan, Hui Liu, et al. 2024 c . Exploring the potential of large language models (llms) in learning on graphs. ACM SIGKDD Explorations Newsletter, 25(2):42--61

    work page 2024

  7. [7]

    Eli Chien, Wei-Cheng Chang, Cho-Jui Hsieh, Hsiang-Fu Yu, Jiong Zhang, Olgica Milenkovic, and Inderjit S Dhillon. 2021. Node feature extraction by self-supervised multi-scale neighborhood prediction. arXiv preprint arXiv:2111.00064

    arXiv 2021

  8. [8]

    Yuanning Cui, Zequn Sun, and Wei Hu. 2024. A prompt-based knowledge graph foundation model for universal in-context reasoning. arXiv preprint arXiv:2410.12288

    work page Pith review arXiv 2024

Show all 56 references
  1. [9]

    Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2018. Bert: Pre-training of deep bidirectional transformers for language understanding. arXiv preprint arXiv:1810.04805

    2018 arXiv

  2. [10]

    Qingxiu Dong, Lei Li, Damai Dai, Ce Zheng, Zhiyong Wu, Baobao Chang, Xu Sun, Jingjing Xu, and Zhifang Sui. 2022. A survey on in-context learning. arXiv preprint arXiv:2301.00234

    2022 arXiv

  3. [11]

    Wenqi Fan, Yao Ma, Qing Li, Yuan He, Eric Zhao, Jiliang Tang, and Dawei Yin. 2019. Graph neural networks for social recommendation. In The world wide web conference, pages 417--426

    2019

  4. [12]

    Yi Fang, Dongzhe Fan, Sirui Ding, Ninghao Liu, and Qiaoyu Tan. 2024 a . Uniglm: Training one unified language model for text-attributed graphs. arXiv preprint arXiv:2406.12052

    2024 arXiv

  5. [13]

    Yi Fang, Dongzhe Fan, Daochen Zha, and Qiaoyu Tan. 2024 b . Gaugllm: Improving graph contrastive learning for text-attributed graphs with large language models. In Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, pages 747--758

    2024

  6. [14]

    Mikhail Galkin, Xinyu Yuan, Hesham Mostafa, Jian Tang, and Zhaocheng Zhu. 2023. Towards foundation models for knowledge graph reasoning. arXiv preprint arXiv:2310.04562

    2023 arXiv

  7. [15]

    Victor Garcia and Joan Bruna. 2017. Few-shot learning with graph neural networks. arXiv preprint arXiv:1711.04043

    2017 arXiv

  8. [16]

    Louie Giray. 2023. Prompt engineering with chatgpt: a guide for academic writers. Annals of biomedical engineering, 51(12):2629--2633

    2023

  9. [17]

    Jiayan Guo, Lun Du, Hengyu Liu, Mengyu Zhou, Xinyi He, and Shi Han. 2023. Gpt4graph: Can large language models understand graph structured data? an empirical evaluation and benchmarking. arXiv preprint arXiv:2305.15066

    2023 arXiv

  10. [18]

    Will Hamilton, Zhitao Ying, and Jure Leskovec. 2017. Inductive representation learning on large graphs. Advances in neural information processing systems, 30

    2017

  11. [19]

    ZS Harris. 1954. Distributional structure

    1954

  12. [20]

    Mohammad Al Hasan and Mohammed J Zaki. 2011. A survey of link prediction in social networks. Social network data analytics, pages 243--275

    2011

  13. [21]

    Xiaoxin He, Xavier Bresson, Thomas Laurent, Bryan Hooi, et al. 2023 a . Explanations as features: Llm-based features for text-attributed graphs. arXiv preprint arXiv:2305.19523, 2(4):8

    2023 arXiv

  14. [22]

    Xiaoxin He, Xavier Bresson, Thomas Laurent, Adam Perold, Yann LeCun, and Bryan Hooi. 2023 b . Harnessing explanations: Llm-to-lm interpreter for enhanced text-attributed graph representation learning. In The Twelfth International Conference on Learning Representations

    2023

  15. [23]

    Yufei He and Bryan Hooi. 2024. Unigraph: Learning a cross-domain graph foundation model from natural language. arXiv preprint arXiv:2402.13630

    2024 arXiv

  16. [24]

    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, 33:22118--22133

    2020

  17. [25]

    Zhengyu Hu, Yichuan Li, Zhengyu Chen, Jingang Wang, Han Liu, Kyumin Lee, and Kaize Ding. 2024. Let's ask gnn: Empowering large language model for graph in-context learning. arXiv preprint arXiv:2410.07074

    2024 arXiv

  18. [26]

    Jin Huang, Xingjian Zhang, Qiaozhu Mei, and Jiaqi Ma. 2023. Can llms effectively leverage graph structural information: when and why. arXiv preprint arXiv:2309.16595

    2023 arXiv

  19. [27]

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

    2016 arXiv

  20. [28]

    Thomas N Kipf and Max Welling. 2016 b . Variational graph auto-encoders. arXiv preprint arXiv:1611.07308

    2016 arXiv

  21. [29]

    Gonzalez, Hao Zhang, and Ion Stoica

    Woosuk Kwon, Zhuohan Li, Siyuan Zhuang, Ying Sheng, Lianmin Zheng, Cody Hao Yu, Joseph E. Gonzalez, Hao Zhang, and Ion Stoica. 2023. https://arxiv.org/abs/2309.06180 Efficient memory management for large language model serving with pagedattention . Preprint, arXiv:2309.06180

    2023 arXiv

  22. [30]

    Guohao Li, Matthias M \"u ller, Bernard Ghanem, and Vladlen Koltun. 2021. Training graph neural networks with 1000 layers. In International conference on machine learning, pages 6437--6449. PMLR

    2021

  23. [31]

    Rui Li, Jiwei Li, Jiawei Han, and Guoyin Wang. 2024 a . Similarity-based neighbor selection for graph llms. arXiv preprint arXiv:2402.03720

    2024 arXiv

  24. [32]

    Yun Li, Yi Yang, Jiaqi Zhu, Hui Chen, and Hongan Wang. 2024 b . Llm-empowered few-shot node classification on incomplete graphs with real node degrees. In Proceedings of the 33rd ACM International Conference on Information and Knowledge Management, pages 1306--1315

    2024

  25. [33]

    Yuyan Liu, Sirui Ding, Sheng Zhou, Wenqi Fan, and Qiaoyu Tan. 2024 a . Moleculargpt: Open large language model (llm) for few-shot molecular property prediction. arXiv preprint arXiv:2406.12950

    2024 arXiv

  26. [34]

    Zheyuan Liu, Xiaoxin He, Yijun Tian, and Nitesh V Chawla. 2024 b . Can we soft prompt llms for graph learning tasks? In Companion Proceedings of the ACM on Web Conference 2024, pages 481--484

    2024

  27. [35]

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

    2000

  28. [36]

    Lawrence Page. 1999. The pagerank citation ranking: Bringing order to the web. Technical report, Technical Report

    1999

  29. [37]

    N Reimers. 2019. Sentence-bert: Sentence embeddings using siamese bert-networks. arXiv preprint arXiv:1908.10084

    2019 arXiv

  30. [38]

    Patrick Reiser, Marlen Neubert, Andr \'e Eberhard, Luca Torresi, Chen Zhou, Chen Shao, Houssam Metni, Clint van Hoesel, Henrik Schopmans, Timo Sommer, et al. 2022. Graph neural networks for materials science and chemistry. Communications Materials, 3(1):93

    2022

  31. [39]

    Prithviraj Sen, Galileo Namata, Mustafa Bilgic, Lise Getoor, Brian Galligher, and Tina Eliassi-Rad. 2008. Collective classification in network data. AI magazine, 29(3):93--93

    2008

  32. [40]

    Oleksandr Shchur, Maximilian Mumme, Aleksandar Bojchevski, and Stephan G \"u nnemann. 2018. Pitfalls of graph neural network evaluation. arXiv preprint arXiv:1811.05868

    2018 arXiv

  33. [41]

    Yucheng Shi, Yushun Dong, Qiaoyu Tan, Jundong Li, and Ninghao Liu. 2023. Gigamae: Generalizable graph masked autoencoder via collaborative latent space reconstruction. In Proceedings of the 32nd ACM International Conference on Information and Knowledge Management, pages 2259--2269

    2023

  34. [42]

    Yucheng Shi, Qiaoyu Tan, Xuansheng Wu, Shaochen Zhong, Kaixiong Zhou, and Ninghao Liu. 2024. Retrieval-enhanced knowledge editing for multi-hop question answering in language models. arXiv preprint arXiv:2403.19631

    2024 arXiv

  35. [43]

    Naoki Shibata, Yuya Kajikawa, and Ichiro Sakata. 2012. Link prediction in citation networks. Journal of the American society for information science and technology, 63(1):78--85

    2012

  36. [44]

    Chuxiong Sun, Hongming Gu, and Jie Hu. 2021. Scalable and adaptive graph neural networks with self-label-enhanced training. arXiv preprint arXiv:2104.09376

    2021 arXiv

  37. [45]

    Qiaoyu Tan, Ninghao Liu, and Xia Hu. 2019. Deep representation learning for social network analysis. Frontiers in big Data, 2:2

    2019

  38. [46]

    Qiaoyu Tan, Ninghao Liu, Xiao Huang, Soo-Hyun Choi, Li Li, Rui Chen, and Xia Hu. 2023. S2gae: Self-supervised graph autoencoders are generalizable learners with graph masking. In Proceedings of the sixteenth ACM international conference on web search and data mining, pages 787--795

    2023

  39. [47]

    Jiabin Tang, Yuhao Yang, Wei Wei, Lei Shi, Lixin Su, Suqi Cheng, Dawei Yin, and Chao Huang. 2023. Graphgpt: Graph instruction tuning for large language models. arXiv preprint arXiv:2310.13023

    2023 arXiv

  40. [48]

    Zheng Wang, Hongming Ding, Li Pan, Jianhua Li, Zhiguo Gong, and S Yu Philip. 2024. From cluster assumption to graph convolution: Graph-based semi-supervised learning revisited. IEEE Transactions on Neural Networks and Learning Systems

    2024

  41. [49]

    Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Fei Xia, Ed Chi, Quoc V Le, Denny Zhou, et al. 2022. Chain-of-thought prompting elicits reasoning in large language models. Advances in neural information processing systems, 35:24824--24837

    2022

  42. [50]

    Shiwen Wu, Fei Sun, Wentao Zhang, Xu Xie, and Bin Cui. 2022. Graph neural networks in recommender systems: a survey. ACM Computing Surveys, 55(5):1--37

    2022

  43. [51]

    Zonghan Wu, Shirui Pan, Fengwen Chen, Guodong Long, Chengqi Zhang, and S Yu Philip. 2020. A comprehensive survey on graph neural networks. IEEE transactions on neural networks and learning systems, 32(1):4--24

    2020

  44. [52]

    Yuhao Xu, Xinqi Liu, Keyu Duan, Yi Fang, Yu-Neng Chuang, Daochen Zha, and Qiaoyu Tan. 2024. Graphfm: A comprehensive benchmark for graph foundation model. arXiv preprint arXiv:2406.08310

    2024 arXiv

  45. [53]

    Ruosong Ye, Caiqi Zhang, Runhui Wang, Shuyuan Xu, and Yongfeng Zhang. 2023. Natural language is all a graph needs. arXiv preprint arXiv:2308.07134

    2023 arXiv

  46. [54]

    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. In Proceedings of the ACM on Web Conference 2024, pages 1003--1014

    2024

  47. [55]

    Jianan Zhao, Le Zhuo, Yikang Shen, Meng Qu, Kai Liu, Michael Bronstein, Zhaocheng Zhu, and Jian Tang. 2023. Graphtext: Graph reasoning in text space. arXiv preprint arXiv:2310.01089

    2023 arXiv

  48. [56]

    Jie Zhou, Ganqu Cui, Shengding Hu, Zhengyan Zhang, Cheng Yang, Zhiyuan Liu, Lifeng Wang, Changcheng Li, and Maosong Sun. 2020. Graph neural networks: A review of methods and applications. AI open, 1:57--81

    2020

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.