GraphLAMA: Enabling Efficient Adaptation of Graph Language Models with Limited Annotations

Cheng Yang; Chuan Shi; Junping Du; Junze Chen; Shujie Li; Yawen Li; Zhiqiang Zhang

REVIEW 2 major objections 3 minor 48 references

GraphLAMA: Enabling Efficient Adaptation of Graph Language Models with Limited Annotations

T0 review · 2 major / 3 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 GraphLAMA claims a small parameter-adaptation stage can specialize a graph language model to an unseen graph using only a few labels.

desk verdict Useful intermediate-adaptation idea with strong empirical results, but an internal inconsistency about which parameters are tuned undermines the central efficiency claim. read the letter →

arxiv 2506.21559 v1 pith:ZT2UW2UZ submitted 2025-06-11 cs.CL

Junze Chen , Cheng Yang , Shujie Li , Zhiqiang Zhang , Yawen Li , Junping Du , Chuan Shi This is my paper

classification cs.CL

keywords GraphlanguagemodelFew-shotlearningParameter-efficientadaptationNodeclassificationSummarygenerationText-attributedgraphsIn-contextneuralnetwork

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 argues that graph language models currently have two inadequate adaptation modes: in-context learning, which is flexible but slow and cannot change its parameters, and instruction tuning, which is strong but requires many labels. GraphLAMA inserts a third mode between them: a short adaptation stage that updates only about 726k parameters (3MB of storage) on top of a frozen large language model, using a handful of labeled examples for the specific target graph and task. In comparisons on Cora, Wiki-CS, and ogbn-products, the method reports the best few-shot and zero-shot classification and summary generation results among the baselines considered, with a 4.91% absolute accuracy gain over the strongest baseline and roughly 10x faster inference than in-context learning under the 5-shot setting. If the claim holds, per-graph and per-task specialization of graph language models becomes practical in settings where labeled data is scarce.

What carries the argument

The load-bearing object is the tiny tunable subset of the backbone: the task-related mask $\mathbf{m}_{rel}$, the task-related weight matrix $\mathbf{W}_{rel}$, the aggregation weight matrix $\mathbf{W}_{agg}$, and the learnable hop encodings. These sit between a frozen GNN encoder and a frozen projector into LLM token space, letting the task text condition which node features the frozen LLM reads. The task-invariant gate stays frozen after pretraining, while the task-related mask is initialized from the task text by a small sentence encoder, so each target task changes only a few megabytes of parameters.

What would settle it

Run the 5-shot adaptation on a text-attributed graph from a domain far from academic citation networks, such as a biomedical or social network corpus, and compare accuracy against a tuned GNN baseline; the central claim fails if accuracy falls to near the no-pretraining level or below that baseline.

Watch

Extended reading notes

Core claim

The central claim is that the gap between in-context learning and instruction tuning can be filled by a third paradigm: a lightweight adaptation stage in which, with the LLM frozen, only a task-related gate, an aggregation weight matrix, and hop encodings are updated from a few labels. The backbone converts nodes into token embeddings through a GNN, learnable hop encodings, two gating modules (one task-invariant and one task-related), and a projector; task text is encoded by a small sentence model to initialize the task-related mask. After pretraining on citation networks with node matching, node classification, and link prediction, adaptation changes roughly 726k parameters, about 1/104 of a 7B LLM, occupying only 3MB. In the paper's experiments this yields the best few-shot and zero-shot classification and summary generation results among the methods compared, with a 4.91% absolute accuracy improvement over the best baseline and roughly 10x faster inference than in-context learning under the 5-shot setting.

Load-bearing premise

The approach assumes that pretraining on ArXiv and PubMed transfers to unseen graphs with different text and structure well enough that tuning a few parameters on a few labels yields top accuracy.

Editorial extensions

If this is right

Graph language models can be specialized per graph and per task with only a few labeled examples, making them usable where annotation budgets are tiny.
Inference cost stays constant as the number of shots grows because examples are not repeated in the prompt; the paper reports 0.52s per target at 50 shots versus 54s for in-context learning.
The per-task memory footprint is roughly 3MB, so many task-specific adapters can be stored and swapped cheaply.
Because the LLM is never modified, the model retains generative abilities such as explanations and summaries without catastrophic forgetting.

Reading between the lines

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

A plausible extension is that the same gate-and-mask mechanism could adapt graph language models to non-text modalities, such as molecules or tabular networks, whenever a sentence can describe the task; this is an extrapolation rather than a claim of the paper.
A practical corollary of the efficiency numbers is that the method's value grows with the number of target tasks, since each adapter is only 3MB and inference avoids long prompts.
A testable extension would be to scale GraphLAMA's frozen LLM up or across model families; the paper compares scaled baselines but does not test whether the 4.91% gain persists when GraphLAMA itself uses a larger LLM.

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.

Desk Editor's Note

read the letter

Two things to know about GraphLAMA. The core idea is genuinely useful: rather than relying on ICL or full instruction tuning, insert a cheap parameter-adaptation stage that specializes a frozen-LLM graph model to an unseen graph and task with a handful of labels. The design, built around task-invariant/task-related gates and learnable hop encodings, is a sensible way to keep that adaptation lightweight. The experimental campaign is broad – six few-shot classification settings, zero-shot, summary generation, ablations – and the headline gains are large (4.91% absolute over the best baseline, roughly 10x faster inference than ICL). The ablations show each component matters, and the cross-domain results on ogbn-products are the most convincing evidence that the pre-training transfers.

Now the soft spots, in order of severity. The paper has a load-bearing internal inconsistency about which parameters are actually updated in the adaptation stage. Section 3.5 says only mrel, Wrel, and Wagg are tuned, occupying less than 3MB. Table 6 reports 726,658 tuned parameters. That number is, to within rounding, the full non-LLM backbone (GNN + projector + gates). Section 3.2 and 3.3 also say hop encodings are tunable, while Algorithm 2's output line and update line disagree with each other and with Section 3.5. If the full backbone is tuned, the 'lightweight adapter' claim is false; if only the gates are tuned, the reported parameter count is wrong. Either way, the paper's central efficiency claim is not currently supported by the text. This is not a nitpick; the whole framing of the paper rests on having a minimal per-target parameter set.

The alignment loss in Eq. (6) is also described inaccurately. With gamma=0.01, exp(-gamma*cos) ranges only from about 0.99 to 1.01, so the 'node matching' term is essentially constant and the loss reduces to L2 alignment. That does not invalidate the empirical results, but the contrastive framing should be corrected. The zero-shot baseline comparison is acknowledged to be non-standard, which is honest but means the zero-shot claims are weaker than they look. Evaluation is limited to three test datasets, all with word2vec features; the cross-domain claim would be stronger with one more distant domain. The GitHub link may or may not have code at this point; I could not verify.

Overall: the idea is good and the experiments are extensive, but the parameter-count inconsistency is serious enough that I would not trust the efficiency claims until it is resolved. This paper deserves a serious referee – the problem is important and the approach is promising – but the referee should insist on a corrected, reproducible parameter count and a rewritten description of what exactly is tuned.

Referee Report

2 major / 3 minor

Summary. GraphLAMA proposes a parameter-adaptation stage for graph language models: a GNN with hop encodings and two gating modules produces node-token embeddings aligned to an LLM's token space; after pre-training on ArXiv/PubMed with node matching, classification, and link prediction, only a small set of parameters is tuned on few-shot examples of the target graph and task, allowing inference without in-context examples. The paper reports state-of-the-art few/zero-shot classification and summary-generation results on Cora, Wiki-CS, and ogbn-products, with a 4.91% absolute accuracy improvement over the best baseline and roughly 10x faster inference than in-context learning under the 5-shot setting. It also includes ablations for backbone components, pre-training tasks, and efficiency measurements.

Significance. If the claims hold, the paper fills a practical gap between in-context learning and instruction tuning for graph language models, showing that a small adaptation stage can specialize a frozen LLM to an unseen graph and task with few labels while reducing inference cost. The empirical breadth is a clear strength: six few-shot classification settings, zero-shot classification, summary generation, ablations for each backbone component and pre-training task, and an efficiency comparison. The public code link is also a positive. However, the internal inconsistency in the reported tunable parameter counts and the near-constant term in the alignment loss mean the efficiency and pre-training descriptions are not currently reliable; both need to be resolved before the contribution can be assessed fairly.

major comments (2)

[Section 3.5, Table 6, Appendix D] Section 3.5 states that adaptation updates only mrel, Wrel, and Wagg, occupying less than 3MB, but Table 6 reports 726,658 tuned parameters for the adaptation stage. Using the dimensions given in Appendix D (128-d word2vec features, 4-d hop encodings, 132-d mask vectors, and a 132-to-4096 projector), mrel, Wrel, Wagg, and the hop encodings contain about 52k parameters; adding the projector (544,768) and the GNN (on the order of 129k) reproduces the reported 726,658 almost exactly. This is a load-bearing inconsistency: the paper's efficiency claims ("only a few pre-trained parameters", "1/104 of a 7B LLM", "occupying less than 3MB") depend on which parameters are actually updated. The authors must either correct the parameter count or, if the projector and GNN are in fact tuned during adaptation, revise the description of the adaptation stage in Sections 3.2, 3.3, 3.5, and Algorithm 2, and recompute the associated efficiency claims.
[Section 3.4, Eq. (6)] Equation (6) defines the node-matching loss as an average over neighbors of exp(-gamma * cos(X_u, C_u)) + ||X_u - C_u||^2. With gamma fixed at 0.01, the exponential term ranges only between roughly e^{-0.01} and e^{0.01}, i.e., 0.990 to 1.010, over the full cosine-similarity range [-1, 1]. The exponential term is therefore an almost constant offset, and the loss is effectively the squared L2 distance. The text in Section 3.4 describes this as an unsupervised contrastive alignment task; that description is not supported by the equation as written. Because Table 5 attributes a large part of the final performance to the self-supervised pre-training, the actual learning signal should be clarified, for example by using a larger temperature so that the cosine term is meaningful, or by reporting an ablation with the exponential term removed.

minor comments (3)

[Algorithm 2] The output line of Algorithm 2 lists alpha and phi_agg as the well-tuned parameters, but the update step in the same algorithm updates mrel, phi_agg, and phi_rel; the pseudocode should be aligned with the text in Section 3.5, which says only mrel, Wrel, and Wagg are updated.
[Introduction] The claim that the tunable parameters are "only 1/104 of a 7B LLM" is arithmetically inconsistent with the reported 726,658 parameters: 726,658 / 7e9 is approximately 1/9,635, not 1/104. Please correct the intended comparison.
[Section 4.2] The abstract and Section 4.2 report a 4.91% average absolute improvement over the best baseline, but the text does not specify how this average is computed across the six few-shot settings; some rows in Table 2 show GraphLAMA close to or slightly below G2P2. Please define the aggregation used for this headline number.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: GraphLAMA's claims are empirical, evaluated on held-out test nodes, and not forced by definition or self-citation.

full rationale

This is an empirical method paper rather than a derivation, and none of its central claims reduce to its inputs by construction. The reported few-shot and zero-shot accuracies are measured on held-out test nodes (100 target nodes per way), and the few-shot examples used in the adaptation stage are disjoint from the test set, so the accuracy numbers are not statistically forced by the training labels. The adaptation-stage loss is a cross-entropy loss on the same few-shot examples used for tuning; this is standard supervised adaptation, not a case where a fitted parameter is renamed as a prediction. The pre-training stage trains the GNN, gates, projector, and hop encodings on ArXiv and PubMed, then freezes them during adaptation; the test datasets (Cora, Wiki-CS, ogbn-products) are different domains, so the transfer claim is externally falsifiable rather than definitionally guaranteed. The paper does invoke prior work from the same research group (e.g., GraphTranslator [42] and the graph foundation model survey [21]), but these citations are used for context and for a pre-training alignment formulation, not as a load-bearing uniqueness theorem or as justification for the main efficiency or accuracy claims. The internal inconsistency between the stated tunable parameter set (mrel, Wrel, Wagg) and the reported 726,658 tuned parameters in Table 6 is a potentially serious correctness or reporting issue, but it is not circularity: the accuracy and speed numbers are still measured outcomes rather than consequences of the parameter count definition. No equation in the paper is equivalent to its own input, no fitted quantity is relabeled as a prediction, and no central result is imported solely from a self-citation. The honest finding is therefore no significant circularity, with a score of 0.

Assumptions & free parameters 4 free parameters · 4 assumptions · 3 invented entities

GraphLAMA is an empirical method, so the ledger focuses on hand-chosen hyperparameters (gamma, lambda, max neighbors, learning rate) and domain assumptions about transfer, the frozen LLM, and the text encoding. The invented entities are model components, not physical postulates, and each has only internal ablation support.

free parameters (4)

gamma = 0.01
Constant in Eq. (6), hand-chosen. With gamma=0.01, the exponential term varies only between 0.990 and 1.010 over cos in [-1,1], so the loss is effectively L2 alignment; the 'contrastive' label overstates its role.
max hop lambda = 2
Subgraph extraction depth, hand-picked following GraphGPT's settings in Appendix D; determines how much neighborhood is encoded into node tokens.
max neighbors = 100
Number of neighbors sampled per node in Appendix D; chosen by hand, affects the neighborhood context seen by the GNN.
learning rate = 1e-6
Adam learning rate for both pre-training and adaptation in Appendix D; chosen from GraphGPT experience, not swept.

assumptions (4)

domain assumption Vicuna-7B remains capable of interpreting projected node tokens and generating correct labels despite being frozen.
Section 3.3 locks the LLM parameters; the entire method relies on the frozen LLM using node tokens and text content to answer correctly.
domain assumption Pre-training on ArXiv and PubMed transfers to Cora, Wiki-CS and ogbn-products.
Section 3.4 and Section 4.1; the few-shot adaptation is initialized from these pre-trained weights, so transfer is essential.
domain assumption Sentence-BERT text encoding f_alpha provides a useful initialization for the task-related mask mrel.
Section 3.3; adaptation starts from mrel = f_alpha(T_task), and if the text encoding does not reflect the task, adaptation starts from a poor point.
domain assumption Word2vec features (128-d) preserve enough text information for the GNN.
Appendix D; all node text is reduced to word2vec vectors before the GNN, which may discard semantics.

invented entities (3)

Node tokens <node_i>
purpose: Virtual token embeddings in the LLM embedding space representing graph nodes, so the LLM can attend to node content and structure.
Introduced by the proposed backbone; no external evidence beyond the paper's own experiments and ablations.
Hop encodings e_i
purpose: Learnable vectors added to node embeddings to encode the hop distance from the target node.
New component; Table 4 ablation shows it matters, but only within this paper.
Task-related mask mrel and task-invariant mask minv
purpose: Channel masks that select task-specific vs task-invariant node features in the gating modules.
Core of the proposed adapter; validated by the 'w/o Task-Related Mask' and 'w/o Gates Module' ablations in Table 4.

how reviews work

0 comments

Cite this review

Pith. "Pith review of GraphLAMA: Enabling Efficient Adaptation of Graph Language Models with Limited Annotations." pith.science (2026). https://pith.science/paper/ZT2UW2UZ

@misc{pith2026250621559,
  author       = {Pith},
  title        = {Pith review of: GraphLAMA: Enabling Efficient Adaptation of Graph Language Models with Limited Annotations},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZT2UW2UZ}},
  note         = {Machine review of arXiv:2506.21559}
}

read the original abstract

Large language models (LLMs) have demonstrated their strong capabilities in various domains, and have been recently integrated for graph analysis as graph language models (GLMs). With LLMs as the predictor, some GLMs can interpret unseen tasks described by natural language, and learn from a few examples in the prompts without parameter tuning, known as in-context learning (ICL). Another subset of GLMs utilizes abundant training labels to enhance model performance, known as instruction tuning. However, we argue that ICL on graphs has effectiveness issues due to fixed parameters and efficiency issues due to long context. Meanwhile, the large amount of labeled data required for instruction tuning can be difficult to obtain in real-world scenarios. To this end, we aim to introduce an extra parameter adaptation stage that can efficiently tailor GLMs to an unseen graph and task with only a few labeled examples, in exchange for better prediction accuracy and faster inference speed. For implementation, in this paper we propose GraphLAMA method, with its model backbone and learning schemes specialized for efficient tuning and inference. Specifically, for model backbone, we use a graph neural network (GNN) with several well-designed components to transform nodes into the representation space of LLM tokens. Task instructions can then be represented as a mixture of node and language tokens. In the pre-training stage, model parameters except the LLM will be trained with different tasks to capture general knowledge. In the adaptation stage, only a few pre-trained parameters will be updated based on few-shot examples. Extensive experiments on few/zero-shot node classification and summary generation show that our proposed GraphLAMA achieves state-of-the-art performance with 4.91% absolution improvement in accuracy. Compared with ICL, our inference speed can be 10 times faster under 5-shot setting.

Figures

Figures reproduced from arXiv: 2506.21559 by the authors.

**Figure 1.** The overall framework of our proposed GraphLAMA, with explanations presented in Section 3.2. [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗

**Figure 2.** Results on zero-shot classification. Our proposed GraphLAMA always has the highest accuracy. [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗

Figures from the paper (2 more)

**Figure 3.** Figure 3: The results of summary generation task in three test datasets. The missing result in 50-shot setting is because the text [PITH_FULL_IMAGE:figures/full_fig_p007_3.png]

**Figure 4.** Figure 4: Case study of classification with explanations. [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

48 extracted references · 21 canonical work pages

[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]

Runjin Chen, Tong Zhao, Ajay Jaiswal, Neil Shah, and Zhangyang Wang. 2024. Llaga: Large language and graph assistant.In Forty-first International Conference on Machine Learning(2024)

work page 2024
[3]

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
[4]

Wei-Lin Chiang, Zhuohan Li, Zi 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

2023
[5]

Yushun Dong, Ninghao Liu, Brian Jalaian, and Jundong Li. 2022. Edits: Modeling and mitigating data bias for graph neural networks. InProceedings of the ACM web conference 2022. 1259–1269

work page 2022
[6]

Taoran Fang, Yunchao Zhang, Yang Yang, Chunping Wang, and Lei Chen. 2024. Universal prompt tuning for graph neural networks.Advances in Neural Infor- mation Processing Systems36 (2024)

work page 2024
[7]

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

2017
[8]

Xiaoxin He, Xavier Bresson, Thomas Laurent, and Bryan Hooi. 2023. Explana- tions as features: Llm-based features for text-attributed graphs.arXiv preprint arXiv:2305.19523(2023)

arXiv 2023

Show all 48 references

[9]

Chawla, Thomas Laurent, Yann LeCun, Xavier Bresson, and Bryan Hooi

Xiaoxin He, Yijun Tian, Yifei Sun, N. Chawla, Thomas Laurent, Yann LeCun, Xavier Bresson, and Bryan Hooi. 2024. G-Retriever: Retrieval-Augmented Gen- eration for Textual Graph Understanding and Question Answering.ArXiv abs/2402.07630 (2024). https://api.semanticscholar.org/Cor...

2024 arXiv
[10]

Zhenyu Hou, Xiao Liu, Yukuo Cen, Yuxiao Dong, Hongxia Yang, Chunjie Wang, and Jie Tang. 2022. Graphmae: Self-supervised masked graph autoencoders. In Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. 594–604

2022
[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]

Ziniu Hu, Yuxiao Dong, Kuansan Wang, Kai-Wei Chang, and Yizhou Sun. 2020. Gpt-gnn: Generative pre-training of graph neural networks. InProceedings of the 26th ACM SIGKDD international conference on knowledge discovery & data mining. 1857–1867

2020
[13]

Qian Huang, Hongyu Ren, Peng Chen, Gregor Kržmanc, Daniel Zeng, Percy S Liang, and Jure Leskovec. 2024. Prodigy: Enabling in-context learning over graphs.Advances in Neural Information Processing Systems36 (2024)

2024
[14]

Baoyu Jing, Chanyoung Park, and Hanghang Tong. 2021. Hdmi: High-order deep multiplex infomax. InProceedings of the Web Conference 2021. 2414–2424

2021
[15]

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

2016 arXiv
[16]

Yuhan Li, Zhixun Li, Peisong Wang, Jia Li, Xiangguo Sun, Hong Cheng, and Jeffrey Xu Yu. 2023. A survey of graph meets large language model: Progress and future directions.arXiv preprint arXiv:2311.12399(2023)

2023 arXiv
[17]

Xuan Lin, Lichang Dai, Yafang Zhou, Zu-Guo Yu, Wen Zhang, Jian-Yu Shi, Dong- Sheng Cao, Li Zeng, Haowen Chen, Bosheng Song, et al. 2023. Comprehensive evaluation of deep and graph learning on drug–drug interactions prediction. Briefings in Bioinformatics24, 4 (2023), bbad235

2023
[18]

Xuan Lin, Zhe Quan, Zhi-Jie Wang, Yan Guo, Xiangxiang Zeng, and S Yu Philip
[19]

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)

2023 arXiv
[20]

Haokun Liu, Derek Tam, Mohammed Muqeeth, Jay Mohta, Tenghao Huang, Mohit Bansal, and Colin A Raffel. 2022. Few-shot parameter-efficient fine-tuning is better and cheaper than in-context learning.Advances in Neural Information Processing Systems35 (2022), 1950–1965

2022
[21]

Jiawei Liu, Cheng Yang, Zhiyuan Lu, Junze Chen, Yibo Li, Mengmei Zhang, Ting Bai, Yuan Fang, Lichao Sun, Philip S Yu, et al. 2023. Towards graph foundation models: A survey and beyond.arXiv preprint arXiv:2310.11829(2023)

2023 arXiv
[22]

Pengfei Liu, Weizhe Yuan, Jinlan Fu, Zhengbao Jiang, Hiroaki Hayashi, and Graham Neubig. 2023. Pre-train, prompt, and predict: A systematic survey of prompting methods in natural language processing.Comput. Surveys55, 9 (2023), 1–35

2023
[23]

Zemin Liu, Xingtong Yu, Yuan Fang, and Xinming Zhang. 2023. Graphprompt: Unifying pre-training and downstream tasks for graph neural networks. InPro- ceedings of the ACM Web Conference 2023. 417–428

2023
[24]

Andrew Kachites McCallum, Kamal Nigam, Jason Rennie, and Kristie Seymore
[25]

Tomas Mikolov, Kai Chen, Greg Corrado, and Jeffrey Dean. 2013. Efficient estimation of word representations in vector space.arXiv preprint arXiv:1301.3781 (2013)

2013 arXiv
[26]

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

2019 arXiv
[27]

Mingchen Sun, Kaixiong Zhou, Xin He, Ying Wang, and Xin Wang. 2022. Gppt: Graph pre-training and prompt tuning to generalize graph neural networks. In Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. 1717–1727

2022
[28]

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
[29]

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. InProceedings of the sixteenth ACM international conference on web search and data mining. 787–795

2023
[30]

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)

2023 arXiv
[31]

Jiabin Tang, Yuhao Yang, Wei Wei, Lei Shi, Long Xia, Dawei Yin, and Chao Huang. 2024. Higpt: Heterogeneous graph language model.arXiv preprint arXiv:2402.16024(2024)

2024 arXiv
[32]

Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timothée Lacroix, Baptiste Rozière, Naman Goyal, Eric Hambro, Faisal Azhar, et al. 2023. Llama: Open and efficient foundation language models.arXiv preprint arXiv:2302.13971(2023)

2023 arXiv
[33]

Petar Velickovic, Guillem Cucurull, Arantxa Casanova, Adriana Romero, Pietro Lio, Yoshua Bengio, et al. 2017. Graph attention networks.stat1050, 20 (2017), 10–48550

2017
[34]

Heng Wang, Shangbin Feng, Tianxing He, Zhaoxuan Tan, Xiaochuang Han, and Yulia Tsvetkov. 2024. Can language models solve graph problems in natural language?Advances in Neural Information Processing Systems36 (2024)

2024
[35]

Xiaoyang Wang, Yao Ma, Yiqi Wang, Wei Jin, Xin Wang, Jiliang Tang, Caiyan Jia, and Jian Yu. 2020. Traffic flow prediction via spatial temporal graph neural network. InProceedings of the web conference 2020. 1082–1092

2020
[36]

Zhihao Wen and Yuan Fang. 2023. Augmenting low-resource text classification with graph-grounded pre-training and prompting. InProceedings of the 46th International ACM SIGIR Conference on Research and Development in Information Retrieval. 506–516

2023
[37]

Jun Xia, Lirong Wu, Jintao Chen, Bozhen Hu, and Stan Z Li. 2022. Simgrace: A simple framework for graph contrastive learning without data augmentation. In Proceedings of the ACM Web Conference 2022. 1070–1079

2022
[38]

Lianghao Xia, Chao Huang, Chunzhen Huang, Kangyi Lin, Tao Yu, and Ben Kao
[39]

Junhan Yang, Zheng Liu, Shitao Xiao, Chaozhuo Li, Defu Lian, Sanjay Agrawal, Amit Singh, Guangzhong Sun, and Xing Xie. 2021. Graphformers: Gnn-nested transformers for representation learning on textual graph.Advances in Neural Information Processing Systems34 (2021), 28798–28810

2021
[40]

Yuning You, Tianlong Chen, Yongduo Sui, Ting Chen, Zhangyang Wang, and Yang Shen. 2020. Graph contrastive learning with augmentations.Advances in neural information processing systems33 (2020), 5812–5823

2020
[41]

Seongjun Yun, Minbyul Jeong, Raehyun Kim, Jaewoo Kang, and Hyunwoo J Kim
[42]

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.arXiv preprint arXiv:2402.07197(2024)

2024 arXiv
[43]

Haiteng Zhao, Shengchao Liu, Ma Chang, Hannan Xu, Jie Fu, Zhihong Deng, Ling- peng Kong, and Qi Liu. 2024. Gimlet: A unified graph-text model for instruction- based molecule zero-shot learning.Advances in Neural Information Processing Systems36 (2024)

2024
[44]

all-MiniLM-L6-v2

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. GraphLAMA: Enabling Efficient Adaptation of Graph Language Models with Limited Annotations KDD ...

2021
[2000]

Information Retrieval3 (2000), 127–163

Automating the construction of internet portals with machine learning. Information Retrieval3 (2000), 127–163

2000
[2019]

Graph transformer networks.Advances in neural information processing systems32 (2019)

2019
[2022]

Effectively identifying compound-protein interaction using graph neural representation.IEEE/ACM Transactions on Computational Biology and Bioinfor- matics20, 2 (2022), 932–943

2022
[2023]

InProceedings of the ACM Web Conference 2023

Automated self-supervised learning for recommendation. InProceedings of the ACM Web Conference 2023. 992–1002

2023

Pith tools

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

[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]

Runjin Chen, Tong Zhao, Ajay Jaiswal, Neil Shah, and Zhangyang Wang. 2024. Llaga: Large language and graph assistant.In Forty-first International Conference on Machine Learning(2024)

work page 2024

[3] [3]

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

[4] [4]

Wei-Lin Chiang, Zhuohan Li, Zi 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

2023

[5] [5]

Yushun Dong, Ninghao Liu, Brian Jalaian, and Jundong Li. 2022. Edits: Modeling and mitigating data bias for graph neural networks. InProceedings of the ACM web conference 2022. 1259–1269

work page 2022

[6] [6]

Taoran Fang, Yunchao Zhang, Yang Yang, Chunping Wang, and Lei Chen. 2024. Universal prompt tuning for graph neural networks.Advances in Neural Infor- mation Processing Systems36 (2024)

work page 2024

[7] [7]

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

2017

[8] [8]

Xiaoxin He, Xavier Bresson, Thomas Laurent, and Bryan Hooi. 2023. Explana- tions as features: Llm-based features for text-attributed graphs.arXiv preprint arXiv:2305.19523(2023)

arXiv 2023