Pith. sign in

REVIEW 4 major objections 6 minor 60 references

NOCL: Node-Oriented Conceptualization LLM for Graph Tasks without Message Passing

T0 review · 4 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read This paper claims that a large language model can handle node, edge, and graph classification on text-attributed and molecular graphs without any message-passing module, by translating node features into compact semantic embeddings and…

desk verdict NOCL's node-concept compression is genuinely useful, but its link prediction results are likely invalid because the prompt may literally contain the edge being predicted. read the letter →

arxiv 2506.10014 v1 pith:76FD6OEY submitted 2025-05-28 cs.LG

Wei Li , Mengcheng Lan , Jiaxing Xu , Yiping Ke This is my paper
classification cs.LG
keywords graphfoundationmodelslargelanguagenodeconceptembeddingzero-shotlearningmessagepassingneuralnetworksrepresentationdescriptorsLoRAinstructiontuningtext-attributedgraphs
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 introduces NOCL, a framework that lets an instruction-tuned LLM solve graph tasks directly from text prompts instead of propagating messages through a GNN. The central idea is to turn each node's raw features into a natural-language description, compress that description into a short learned embedding with a pretrained language model, and then feed the LLM a graph descriptor that lists nodes and edges. This recasts node, link, and graph classification as ordinary next-token prediction, so one model can serve all three task levels without specialized output heads. If the approach holds, graph learning no longer needs message passing, and a single LLM can transfer zero-shot to unseen graph datasets.

What carries the argument

The load-bearing objects are the node concept embedding and the graph representation descriptor. A node concept is a compact semantic embedding produced by feeding a node's natural-language description through a pretrained language model; a graph descriptor is a token sequence that lists nodes' concept embeddings and then their edges, delimited by special tokens like <|BON|>, <|NC|>, and <|BOE|>. Together they let the LLM see both attributes and adjacency in one prompt, and they convert every downstream task into a question answered by next-token generation.

What would settle it

Construct a synthetic benchmark where the class label is determined entirely by a node's second- or third-hop structure and the 1-hop features are class-balanced noise; if NOCL's accuracy drops to chance while an MPNN with message passing stays high, the sufficiency of the 11-node induced subgraph is refuted.

Watch

Extended reading notes

Core claim

NOCL claims that graph structure can be traded for language: a frozen LLM, lightly adapted with LoRA, can classify nodes, predict links, and classify whole graphs by reading graph representation descriptors built from node concept embeddings. The node concept is a fixed-size vector made by encoding a node's description with a sentence transformer, which cuts input token counts by up to 93.9% while keeping the semantics. With node-level and link-level tasks re-expressed as graph-level questions over induced subgraphs, all tasks become text comprehension, and experiments on Cora, PubMed, ogbn-arxiv, MUTAG, and ogbg-molhiv report supervised accuracy competitive with MPNNs and better zero-shot transfer than existing LLM baselines.

Load-bearing premise

For node and link tasks, NOCL assumes that a node's 1-hop neighborhood capped at ten neighbors contains enough structural signal for correct classification; if multi-hop context is essential, the reported performance would not hold.

Editorial extensions

If this is right

  • A single LoRA-tuned LLM can solve node, link, and graph classification together, with no message-passing module and no task-specific heads.
  • Token compression from node concepts (up to 93.9%) brings training memory low enough for commercial GPUs, enabling datasets that previously caused out-of-memory failures.
  • Zero-shot transfer improves over LLM baselines on Cora and matches supervised MPNN link prediction on PubMed, suggesting labels are not strictly required for competitive graph reasoning.
  • Because outputs come from next-token prediction, the same framework extends beyond fixed labels to free-form graph explanations and questions.
  • The approach extends LLM use from text-attributed graphs to molecular graphs, where node features are converted into natural-language descriptions.

Reading between the lines

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

  • Editorial inference: if the 1-hop-truncation assumption is robust, NOCL-style models could scale to much larger graphs by subsampling neighborhoods, a setting the paper explicitly leaves untested.
  • Editorial inference: the paper's framing suggests that structural regularities used by MPNNs, such as oversmoothing and locality bias, can be bypassed entirely; a testable corollary is that NOCL should maintain accuracy on graphs with long-range dependencies that hurt shallow MPNNs.
  • Editorial inference: replacing hand-written non-TAG templates with an LLM-generated or learned description step would likely improve generalization to new molecule types, since the current templates require expert curation.
  • Editorial inference: one could probe how much structure the LLM actually uses by ablating edge lines from the graph descriptor; if accuracy stays high, the model may be relying mostly on node semantics rather than topology.
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

4 major / 6 minor

Summary. The paper proposes NOCL, an MPNN-free framework that converts node features into natural-language descriptions, encodes them into compact 'node concept' embeddings via a pretrained language model and a linear connector, and feeds graph representation descriptors (node and edge listings) into a LoRA-tuned LLM. The authors claim that this unified, text-based formulation handles node, link, and graph classification on both textual-attributed and non-textual graphs, with large token-length reductions and strong zero-shot generalization, and they report experiments on ogbn-arxiv, Cora, PubMed, MUTAG, and ogbg-molhiv.

Significance. If the reported results hold, NOCL would be a useful step toward graph foundation models that do not rely on message passing, with concrete efficiency gains (up to 93.9% token reduction, substantial memory and time savings) and a unified task format. The paper also provides a reproducibility link and discusses limitations honestly in Section 6. However, several load-bearing aspects of the evaluation—particularly the link prediction protocol and the zero-shot setting—require scrutiny before the claims can be accepted.

major comments (4)
  1. [3.4, Table 3] The link prediction evaluation does not mask the target edge in the graph representation descriptor. Section 3.4 states that the descriptor lists all edges, and for edge-level tasks edges from the first target node's induced graph Gvi precede edges from the second target node's induced graph Gvj. For a positive test pair (vi, vj), the edge (vi, vj) is by definition in Gvi, so the descriptor contains the tuple <|EDGE|> vi vj; for a negative pair it does not. Because the prompt asks 'Should node vi connect node vj?' with the full descriptor visible, the LLM can answer by checking string membership rather than by structural reasoning. The paper never states that this edge is masked (Section 4.1 only describes the 1-hop cap and the linear projection head). Consequently, the supervised Cora link results (0.8842/0.8965) and zero-shot PubMed link results (0.6057/0.6351) do not currently support the claimed link prediction capability. The authors should specify and implement a masking protocol (e.g., removing the queried edge from the descriptor) and re-report all link-level numbers.
  2. [3.3, 4.1, Appendix B] The 'zero-shot' results are partially self-referential because the node concept connector is tuned with unlabeled node descriptions from the same target datasets used for zero-shot evaluation. Section 4.1 states 'We utilize all nodes from these datasets to generate the corresponding node descriptions and node concepts, which are then employed in training the connector,' and Appendix B shows that title/abstract recovery data are generated for all nodes, with only the class-prediction task restricted to training nodes. Thus, the PubMed and Cora zero-shot numbers are obtained with a model whose connector has already seen the raw text of those graphs. This weakens the claim of generalization to unseen domains made in Section 1. The authors should either exclude target-dataset text from connector tuning entirely and re-run the zero-shot experiments, or explicitly relabel the setting as 'transductive zero-label' and adjust the claims accordingly.
  3. [4.1, Tables 3-4] The ROC_AUC numbers for link and graph tasks are not produced by text generation but by a learned linear projection applied to the LLM's last hidden state. Section 4.1 states: 'To adapt LLM outputs for the ROC_AUC evaluation metric, we flatten the last hidden state corresponding to the first output token of the LLM and apply a linear projection to produce the final numeric prediction.' This contradicts the paper's claim in Section 3.4 that NOCL does not rely on specialized heads or task-specific formats. It is also unclear how this projection head is trained relative to the LoRA adapter, and how it transfers in the zero-shot PubMed link prediction. Please report the exact training protocol for the projection head, or use a text-derived scoring rule (e.g., answer-token probability) for all tasks.
  4. [4.1, Table 3, Checklist item 7] No uncertainty quantification is reported, and MUTAG has only 188 graphs with an 80/20 split, giving roughly 38 test graphs. The claimed improvement over the best MPNN on MUTAG (0.7262 vs 0.6310 for the 1B model) could be within sampling noise. The authors acknowledge the absence of error bars in the checklist but still make strong comparative statements (e.g., 'exceeds the best-performing MPNN by a margin of 0.09'). At minimum, report standard deviations over multiple random splits or bootstrap confidence intervals for MUTAG and for the zero-shot results, or temper the claims accordingly.
minor comments (6)
  1. [3.4] The formula 'The overall token length of graph representation descriptors for G is 4 + 2n + 3n' should read '4 + 2n + 3m', since the second term counts edge tuples and there are m edges, not n.
  2. [5] There are typos such as 'The LLaV A seires' and 'pretraind'; these should be corrected.
  3. [Table 8] In Table 8, the 'Ave. Node' and 'Ave Edge' columns are empty for the citation networks and molecular graphs; clarify whether these are per-graph averages and fill in the values for consistency.
  4. [Checklist item 12] The authors answer 'No' to the question about licenses but then provide the names of the Llama 3.2 and Apache 2.0 licenses; please clarify whether full license texts and terms are included with the released code and data.
  5. [Checklist item 16] The declaration of LLM usage is inconsistent: the paper answers 'No' with the justification that the LLM is used only for writing, whereas the core methodology is built around LoRA fine-tuning of Llama-3.2. This should be corrected to accurately reflect the use of LLMs in the research method.
  6. [Table 2] The base model zero-shot results for PubMed show 0.0000 and 0.0002 accuracy, which appear degenerate; a brief explanation of these outputs would help the reader interpret the comparison.

Circularity Check

1 steps flagged · score 6.0 of 10

Link prediction in Table 3 reduces to edge-tuple lookup because the queried edge is included in the graph descriptor by construction; node and graph classification results remain independent.

  1. self definitional [Section 3.4 (Graph representation descriptors; edge-level reformulation), Section 4.1, Table 3]
    "For edge-level tasks, we similarly position one target node vi first and generate all nodes within its induced graph Gvi. We subsequently generate the second target node vj along with its induced graph Gvj. Edges are generated randomly within graph descriptors, except for edge-level tasks, where edges from Gvi precede edges from Gvj."

    For a positive link-prediction test pair (vi, vj), the edge (vi, vj) is an actual edge, so it belongs to the induced graph Gvi; by the descriptor definition in Section 3.4, the prompt therefore contains the tuple <|EDGE|> vi vj. For a negative pair, no such tuple is present. The paper never states that this target edge is masked in Section 4.1 or elsewhere. The LLM can thus answer the query 'Should node 0 connect node 7?' by checking whether the edge tuple appears verbatim in the prompt, reducing link prediction to string membership. The reported Cora (0.8965) and PubMed (0.6351) link ROC_AUC values in Table 3 are consequently forced by the input construction rather than being structural predictions, so the unified link-task claim is not independently measured.

full rationale

Most of NOCL is an empirical LLM fine-tuning pipeline with no derivation-based circularity: node classification and graph classification accuracies are computed on held-out labels, and the node-concept connector is an alignment module rather than a restatement of downstream targets. The one load-bearing reduction is in edge-level tasks. Section 3.4 defines graph descriptors to list every edge of the induced subgraph, including, for a positive query (vi, vj), the tuple <|EDGE|> vi vj via Gvi, and Section 4.1 describes no masking of that target edge. Consequently the LLM can answer the link-prediction question by checking presence of the edge tuple in the prompt; the reported Cora (0.8965) and PubMed (0.6351) link ROC_AUC values in Table 3 are therefore forced by the input representation rather than measuring structural link prediction. This is a task-level circularity, not a collapse of the whole framework: node and graph descriptors do not contain their labels. The zero-shot claims are transductive in that the connector is trained on all nodes' descriptions from the same corpora, but labels are held out; this lowers external validity without being definitional circularity. Score 6 reflects one set of reported predictions reducing by construction, while the central node and graph results remain independently evaluated.

Assumptions & free parameters 1 free parameters · 4 assumptions · 1 invented entities

NOCL's central claims rest on four domain assumptions specific to this pipeline: the sufficiency of a single compressed embedding, the sufficiency of 1-hop subgraphs, the losslessness of the non-TAG description templates, and the adequacy of the connector tuning recipe. The only hand-chosen scalar that shapes the method is the subgraph node cap. No independent external benchmark validates the node concept representation outside the paper's own experiments.

free parameters (1)
  • Subgraph node cap K=10 = 10
    In Section 4.1, for node- and link-level tasks, if a target node has more than 10 neighbors, 10 are randomly sampled to construct the induced subgraph. This cap is a hand-chosen design parameter that directly shapes the graph context available to the LLM and the paper states scaling to larger neighborhoods is open.
assumptions (4)
  • domain assumption A single Sentence-BERT embedding of a node's textual description, after a linear connector, preserves enough semantics for an LLM to perform graph tasks.
    The whole node concept design relies on this; the paper tests it empirically but provides no guarantee or analysis of information loss. See Section 3.3.
  • domain assumption 1-hop induced subgraphs capped at 10 neighbors are sufficient structural context for node classification and link prediction on Cora and PubMed.
    Section 4.1 imposes this subgraph construction; if multi-hop context is needed, the reported results would not generalize. The paper itself lists this as an open question in Section 6.
  • domain assumption The hand-written templates for ogbg-molhiv and MUTAG convert atom features into natural language without discarding task-relevant information.
    Section A.1 defines the templates; the molecular predictions depend on this conversion being lossless enough for HIV inhibition and mutagenicity.
  • ad hoc to paper Connector tuning via description recovery and class-prediction on training nodes is sufficient to align PLM and LLM embedding spaces for downstream graph tasks.
    Section 3.3 introduces this two-stage training recipe; it is a design choice specific to NOCL, not a standard result.
invented entities (1)
  • Node concept embedding (with <|NC|> token)
    purpose: A fixed-size vector that stands in for a node's full description in the LLM input, reducing token count.
    The paper demonstrates its utility only through its own connector tuning and downstream experiments; no external benchmark or falsifiable prediction about the embedding itself is provided outside this pipeline. The <|NC|> placeholder is a new special token introduced in Section 3.4.

how reviews work

0 comments
Cite this review

Pith. "Pith review of NOCL: Node-Oriented Conceptualization LLM for Graph Tasks without Message Passing." pith.science (2026). https://pith.science/paper/76FD6OEY

@misc{pith2026250610014,
  author       = {Pith},
  title        = {Pith review of: NOCL: Node-Oriented Conceptualization LLM for Graph Tasks without Message Passing},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/76FD6OEY}},
  note         = {Machine review of arXiv:2506.10014}
}
read the original abstract

Graphs are essential for modeling complex interactions across domains such as social networks, biology, and recommendation systems. Traditional Graph Neural Networks, particularly Message Passing Neural Networks (MPNNs), rely heavily on supervised learning, limiting their generalization and applicability in label-scarce scenarios. Recent self-supervised approaches still require labeled fine-tuning, limiting their effectiveness in zero-shot scenarios. Meanwhile, Large Language Models (LLMs) excel in natural language tasks but face significant challenges when applied to graphs, including preserving reasoning abilities, managing extensive token lengths from rich node attributes, and being limited to textual-attributed graphs (TAGs) and a single level task. To overcome these limitations, we propose the Node-Oriented Conceptualization LLM (NOCL), a novel framework that leverages two core techniques: 1) node description, which converts heterogeneous node attributes into structured natural language, extending LLM from TAGs to non-TAGs; 2) node concept, which encodes node descriptions into compact semantic embeddings using pretrained language models, significantly reducing token lengths by up to 93.9% compared to directly using node descriptions. Additionally, our NOCL employs graph representation descriptors to unify graph tasks at various levels into a shared, language-based query format, paving a new direction for Graph Foundation Models. Experimental results validate NOCL's competitive supervised performance relative to traditional MPNNs and hybrid LLM-MPNN methods and demonstrate superior generalization in zero-shot settings.

Figures

Figures reproduced from arXiv: 2506.10014 by the authors.

Figure 1
Figure 1. The overall framework of NOCL. The training process of NOCL consists of two stages. [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Node Concept Examples. Connector Tuning Query: <|NC|>This is an embedding of an atom in a molecule. What is the element type of this atom? Response: This atom is boron [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 4
Figure 4. Graph Instruct Tuning Examples. Instruction tuning of downstream tasks. Leveraging the proposed graph descriptors, we construct graph instruction datasets from existing structured datasets. Each <graph, task> pair is formatted into a query-response template: Query: This is a graph: <Graph Descriptors>. <Downstream Task Query> Response: <Corresponding Text Label> [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figures from the paper (6 more)
Figure 5
Figure 5. Figure 5: Three examples of the node description on ogbn-arxiv dataset. [PITH_FULL_IMAGE:figures/full_fig_p022_5.png]
Figure 6
Figure 6. Figure 6: Two examples of the node description on PubMed dataset. [PITH_FULL_IMAGE:figures/full_fig_p023_6.png]
Figure 7
Figure 7. Figure 7: Two examples of the node description on Cora dataset. [PITH_FULL_IMAGE:figures/full_fig_p024_7.png]
Figure 8
Figure 8. Figure 8: Examples of connector tuning for TAGs. To prevent data leakage, we ensure that the class [PITH_FULL_IMAGE:figures/full_fig_p025_8.png]
Figure 9
Figure 9. Figure 9: Examples of connector tuning for non-TAGs [PITH_FULL_IMAGE:figures/full_fig_p025_9.png]
Figure 10
Figure 10. Figure 10: Examples of reformulation node classification task [PITH_FULL_IMAGE:figures/full_fig_p026_10.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

60 extracted references · 28 canonical work pages

  1. [1]

    Jean-Baptiste Alayrac, Jeff Donahue, Pauline Luc, Antoine Miech, Iain Barr, Yana Hasson, Karel Lenc, Arthur Mensch, Katherine Millican, Malcolm Reynolds, et al . 2022. Flamingo: a visual language model for few-shot learning.Advances in neural information processing systems35 (2022), 23716–23736

    work page 2022

  2. [2]

    James Atwood and Don Towsley. 2016. Diffusion-convolutional neural networks.Advances in neural information processing systems29 (2016)

    work page 2016

  3. [3]

    Baichuan. 2023. A large-scale 7b pretraining language model developed by baichuan-inc. https://github.com/baichuan-inc/Baichuan-7B

    work page 2023

  4. [4]

    He Cao, Zijing Liu, Xingyu Lu, Yuan Yao, and Yu Li. 2023. Instructmol: Multi-modal integration for building a versatile and reliable molecular assistant in drug discovery.arXiv preprint arXiv:2311.16208(2023)

    arXiv 2023

  5. [5]

    Zhikai Chen, Haitao Mao, Hongzhi Wen, Haoyu Han, Wei Jin, Haiyang Zhang, Hui Liu, and Jiliang Tang. 2023. Label-free node classification on graphs with large language models (llms). arXiv preprint arXiv:2310.04668(2023)

    arXiv 2023

  6. [6]

    Wei-Lin Chiang, Zhuohan Li, Ziqing Lin, Ying Sheng, Zhanghao Wu, Hao Zhang, Lianmin Zheng, Siyuan Zhuang, Yonghao Zhuang, Joseph E Gonzalez, et al. 2023. Vicuna: An open- source chatbot impressing gpt-4 with 90%* chatgpt quality.See https://vicuna. lmsys. org (accessed 14 April 2023)2, 3 (2023), 6

    work page 2023

  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 neighbor- hood prediction.arXiv preprint arXiv:2111.00064(2021)

    arXiv 2021

  8. [8]

    Asim Kumar Debnath, Rosa L Lopez de Compadre, Gargi Debnath, Alan J Shusterman, and Corwin Hansch. 1991. Structure-activity relationship of mutagenic aromatic and heteroaromatic nitro compounds. correlation with molecular orbital energies and hydrophobicity.Journal of medicinal chemistry34, 2 (1991), 786–797

    work page 1991

Show all 60 references
  1. [9]

    Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. Bert: Pre-training of deep bidirectional transformers for language understanding. InProceedings of the 2019 conference of the North American chapter of the association for computational linguistics: human ...

    2019

  2. [10]

    Yingtong Dou, Zhiwei Liu, Li Sun, Yutong Deng, Hao Peng, and Philip S Yu. 2020. Enhancing graph neural network-based fraud detectors against camouflaged fraudsters. InProceedings of the 29th ACM international conference on information & knowledge management. 315–324

    2020

  3. [11]

    Yin Fang, Xiaozhuan Liang, Ningyu Zhang, Kangwei Liu, Rui Huang, Zhuo Chen, Xiaohui Fan, and Huajun Chen. 2023. Mol-instructions: A large-scale biomolecular instruction dataset for large language models.arXiv preprint arXiv:2306.08018(2023)

    2023 arXiv

  4. [12]

    Jiyang Gao, Chen Sun, Hang Zhao, Yi Shen, Dragomir Anguelov, Congcong Li, and Cordelia Schmid. 2020. Vectornet: Encoding hd maps and agent dynamics from vectorized representation. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 11525–11533

    2020

  5. [13]

    Taicheng Guo, Kehan Guo, Zhengwen Liang, Zhichun Guo, Nitesh V Chawla, Olaf Wiest, Xiangliang Zhang, et al. 2023. What indeed can GPT models do in chemistry? A comprehensive benchmark on eight tasks.arXiv preprint arXiv:2305.183653 (2023)

    2023 arXiv

  6. [14]

    Will Hamilton, Zhitao Ying, and Jure Leskovec. 2017. Inductive representation learning on large graphs.Advances in neural information processing systems30 (2017)

    2017

  7. [15]

    Xiaoxin He, Xavier Bresson, Thomas Laurent, Adam Perold, Yann LeCun, and Bryan Hooi

  8. [16]

    Zhenyu Hou, Xiao Liu, Yukuo Cen, Yuxiao Dong, Hongxia Yang, Chunjie Wang, and Jie Tang

  9. [17]

    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.ICLR1, 2 (2022), 3

    2022

  10. [18]

    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.arXiv preprint arXiv:2005.00687(2020)

    2020 arXiv

  11. [19]

    Bowen Jin, Gang Liu, Chi Han, Meng Jiang, Heng Ji, and Jiawei Han. 2024. Large language models on graphs: A comprehensive survey.IEEE Transactions on Knowledge and Data Engineering(2024)

    2024

  12. [20]

    Wei Jin, Tyler Derr, Haochen Liu, Yiqi Wang, Suhang Wang, Zitao Liu, and Jiliang Tang

  13. [21]

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

    2016 arXiv

  14. [22]

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

    2016 arXiv

  15. [23]

    Guohao Li, Matthias Müller, Bernard Ghanem, and Vladlen Koltun. 2021. Training graph neural networks with 1000 layers. InInternational conference on machine learning. PMLR, 6437–6449

    2021

  16. [24]

    Junnan Li, Dongxu Li, Silvio Savarese, and Steven Hoi. 2023. Blip-2: Bootstrapping language- image pre-training with frozen image encoders and large language models. InInternational conference on machine learning. PMLR, 19730–19742

    2023

  17. [25]

    Haotian Liu, Chunyuan Li, Yuheng Li, and Yong Jae Lee. 2024. Improved baselines with visual instruction tuning. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 26296–26306

    2024

  18. [26]

    Haotian Liu, Chunyuan Li, Qingyang Wu, and Yong Jae Lee. 2023. Visual instruction tuning. Advances in neural information processing systems36 (2023), 34892–34916

    2023

  19. [27]

    Zhiyuan Liu, Sihang Li, Yanchen Luo, Hao Fei, Yixin Cao, Kenji Kawaguchi, Xiang Wang, and Tat-Seng Chua. 2023. Molca: Molecular graph-language modeling with cross-modal projector and uni-modal adapter.arXiv preprint arXiv:2310.12798(2023)

    2023 arXiv

  20. [28]

    Jianxin Ma, Chang Zhou, Peng Cui, Hongxia Yang, and Wenwu Zhu. 2019. Learning disentan- gled representations for recommendation.Advances in neural information processing systems 32 (2019)

    2019

  21. [29]

    Haitao Mao, Zhikai Chen, Wenzhuo Tang, Jianan Zhao, Yao Ma, Tong Zhao, Neil Shah, Mikhail Galkin, and Jiliang Tang. 2024. Position: Graph foundation models are already here. In Forty-first International Conference on Machine Learning

    2024

  22. [30]

    Andrew Kachites McCallum, Kamal Nigam, Jason Rennie, and Kristie Seymore. 2000. Au- tomating the construction of internet portals with machine learning.Information Retrieval3 (2000), 127–163

    2000

  23. [31]

    Meta AI. 2024. LLaMA 3: Open Foundation and Instruction Models. https://ai.meta. com/blog/meta-llama-3/

    2024

  24. [32]

    Shirui Pan, Ruiqi Hu, Guodong Long, Jing Jiang, Lina Yao, and Chengqi Zhang. 2018. Adver- sarially regularized graph autoencoder for graph embedding.arXiv preprint arXiv:1802.04407 (2018). 12

    2018 arXiv

  25. [33]

    Chen Qian, Huayi Tang, Zhirui Yang, Hong Liang, and Yong Liu. 2023. Can large language models empower molecular property prediction?arXiv preprint arXiv:2307.07443(2023)

    2023 arXiv

  26. [34]

    Nils Reimers and Iryna Gurevych. 2019. Sentence-bert: Sentence embeddings using siamese bert-networks.arXiv preprint arXiv:1908.10084(2019)

    2019 arXiv

  27. [35]

    Prithviraj Sen, Galileo Namata, Mustafa Bilgic, Lise Getoor, Brian Galligher, and Tina Eliassi- Rad. 2008. Collective classification in network data.AI magazine29, 3 (2008), 93–93

    2008

  28. [36]

    Yaorui Shi, An Zhang, Enzhi Zhang, Zhiyuan Liu, and Xiang Wang. 2023. Relm: Leveraging language models for enhanced chemical reaction prediction.arXiv preprint arXiv:2310.13590 (2023)

    2023 arXiv

  29. [37]

    Martin Simonovsky and Nikos Komodakis. 2017. Dynamic edge-conditioned filters in convolu- tional neural networks on graphs. InProceedings of the IEEE conference on computer vision and pattern recognition. 3693–3702

    2017

  30. [38]

    Xiangguo Sun, Hong Cheng, Jia Li, Bo Liu, and Jihong Guan. 2023. All in one: Multi-task prompting for graph neural networks. InProceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. 2120–2131

    2023

  31. [39]

    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. 787–795

    2023

  32. [40]

    Jie Tang, Jimeng Sun, Chi Wang, and Zi Yang. 2009. Social influence analysis in large-scale networks. InProceedings of the 15th ACM SIGKDD international conference on Knowledge discovery and data mining. 807–816

    2009

  33. [41]

    Jiabin Tang, Yuhao Yang, Wei Wei, Lei Shi, Lixin Su, Suqi Cheng, Dawei Yin, and Chao Huang

  34. [42]

    Ross Taylor, Marcin Kardas, Guillem Cucurull, Thomas Scialom, Anthony Hartshorn, Elvis Saravia, Andrew Poulton, Viktor Kerkez, and Robert Stojnic. 2022. Galactica: A large language model for science.arXiv preprint arXiv:2211.09085(2022)

    2022 arXiv

  35. [43]

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

    2023 arXiv

  36. [44]

    Petar Veliˇckovi´c, Guillem Cucurull, Arantxa Casanova, Adriana Romero, Pietro Lio, and Yoshua Bengio. 2017. Graph attention networks.arXiv preprint arXiv:1710.10903(2017)

    2017 arXiv

  37. [45]

    Zonghan Wu, Shirui Pan, Guodong Long, Jing Jiang, Xiaojun Chang, and Chengqi Zhang

  38. [46]

    Jiaxing Xu, Kai He, Mengcheng Lan, Qingtian Bian, Wei Li, Tieying Li, Yiping Ke, and Miao Qiao. 2024. Contrasformer: a brain network contrastive transformer for neurodegenerative con- dition identification. InProceedings of the 33rd ACM International Conference on Information ...

    2024

  39. [47]

    Chenxiao Yang, Qitian Wu, and Junchi Yan. 2022. Geometric knowledge distillation: Topology compression for graph neural networks.Advances in Neural Information Processing Systems 35 (2022), 29761–29775

    2022

  40. [48]

    Ruosong Ye, Caiqi Zhang, Runhui Wang, Shuyuan Xu, and Yongfeng Zhang. 2023. Language is all a graph needs.arXiv preprint arXiv:2308.07134(2023). 13

    2023 arXiv

  41. [49]

    Rex Ying, Ruining He, Kaifeng Chen, Pong Eksombatchai, William L Hamilton, and Jure Leskovec. 2018. Graph convolutional neural networks for web-scale recommender systems. InProceedings of the 24th ACM SIGKDD international conference on knowledge discovery & data mining. 974–983

    2018

  42. [50]

    InProceedings of the 26th ACM SIGKDD international conference on knowledge discovery & data mining

    Connecting the dots: Multivariate time series forecasting with graph neural networks. InProceedings of the 26th ACM SIGKDD international conference on knowledge discovery & data mining. 753–763

  43. [51]

    Jianan Zhao, Meng Qu, Chaozhuo Li, Hao Yan, Qian Liu, Rui Li, Xing Xie, and Jian Tang

  44. [52]

    Limitations

    Yanqiao Zhu, Yichen Xu, Feng Yu, Qiang Liu, Shu Wu, and Liang Wang. 2021. Graph contrastive learning with adaptive augmentation. InProceedings of the web conference 2021. 2069–2080. 14 NeurIPS Paper Checklist 1.Claims Question: Do the main claims made in the abstract and intro...

    2021

  45. [55]

    Shichang Zhang, Yozen Liu, Yizhou Sun, and Neil Shah. 2021. Graph-less neural networks: Teaching old mlps new tricks via distillation.arXiv preprint arXiv:2110.08727(2021)

    2021 arXiv

  46. [57]

    Learning on large-scale text-attributed graphs via variational inference.arXiv preprint arXiv:2210.14709(2022)

    2022 arXiv

  47. [59]

    Guidelines: • The answer NA means that the paper does not involve crowdsourcing nor research with human subjects

    Institutional review board (IRB) approvals or equivalent for research with human subjects Question: Does the paper describe potential risks incurred by study participants, whether such risks were disclosed to the subjects, and whether Institutional Review Board (IRB) approvals...

    2025

  48. [60]

    Yes, these two nodes should be connected

    class prediction. Representative examples are provided in Fig. 8. To prevent data leakage, we ensure that the class prediction task is applied only to nodes from the training set. Non-TAGsFor non-TAGs, we simply make LLM to reconstruct the original node description. We provide...

  49. [2020]

    Self-supervised learning on graphs: Deep insights and new direction.arXiv preprint arXiv:2006.10141(2020)

    2020 arXiv

  50. [2022]

    InProceedings of the 28th ACM SIGKDD conference on knowledge discovery and data mining

    Graphmae: Self-supervised masked graph autoencoders. InProceedings of the 28th ACM SIGKDD conference on knowledge discovery and data mining. 594–604

  51. [2023]

    Harnessing explanations: Llm-to-lm interpreter for enhanced text-attributed graph representation learning.arXiv preprint arXiv:2305.19523(2023). 11

    2023 arXiv

  52. [2024]

    InProceedings of the 47th International ACM SIGIR Conference on Research and Development in Information Retrieval

    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

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.