Pith. sign in

REVIEW 2 major objections 5 minor 55 references

Finetuned graph language model retrievers transfer better out of domain than GNN-based graph RAG, setting SOTA on multi-hop medical and computer-science benchmarks while staying competitive in-domain.

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

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

T0 review · grok-4.5

2026-07-31 08:42 UTC pith:WAVR2PVT

load-bearing objection Solid comparative systems paper: GLM retriever beats GNN baselines on multi-hop OOD transfer, with a real but acknowledged visibility caveat and concurrent work already moving the SOTA ceiling. the 2 major comments →

arxiv 2607.28397 v1 pith:WAVR2PVT submitted 2026-07-30 cs.AI cs.CLcs.IR

GLM-RAG: Graph Language Models for Graph-Based Retrieval-Augmented Generation

Maya Arseven , Anette Frank , Beni Egressy , Johann Higl , Moritz Plenz This is my paper
classification cs.AI cs.CLcs.IR
keywords graph RAGgraph language modelsmulti-hop retrievalknowledge graphstransferabilityGNN retrieversretrieval-augmented generation
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

This paper argues that retrieval over knowledge graphs for multi-hop question answering needs both graph structure and real semantic understanding of node and relation text, not just topology plus shallow embeddings. It replaces the usual GNN retriever with a graph language model that reads tokenized entities, relations, and the query together under structure-aware attention on a local subgraph. Across Wikipedia multi-hop sets the GLM stays comparable to strong GNN baselines; on held-out single-hop domains plain vector search wins; on multi-hop out-of-domain suites the GLM pulls clearly ahead and reaches state-of-the-art on G-Bench Medical and CS. The practical point is that a transferable graph-RAG foundation model is more useful than another in-domain specialist, and that semantic depth, not just wider message passing, is what carries the transfer.

Core claim

Finetuned GLM-based retrievers generalize better out of domain than GNN-based and vector-search retrievers on multi-hop graph RAG, achieving state-of-the-art on two multi-hop benchmarks while remaining comparable on in-domain multi-hop QA; vanilla vector RAG is sufficient for single-hop datasets, and GNN retrievers win on raw graph coverage and training efficiency.

What carries the argument

The GLM-based retriever: a pretrained language model turned into a graph transformer that tokenizes nodes, relations and the query, applies structure-aware relative attention on a 2-hop subgraph (budgeted to a few hundred triplets), and scores entities by fusing their contextual embeddings with the query.

Load-bearing premise

The out-of-domain gains are attributed to deeper semantic integration only if restricting the GLM to a small local subgraph is a fair, not decisive, handicap relative to full-graph GNN message passing.

What would settle it

Match visibility and capacity exactly: give the GNN the same 2-hop/600-triplet budget (or give the GLM full-graph access) and re-run MultihopRAG plus G-Bench Medical/CS; if the GLM’s large Recall@2 lead disappears, the semantic-integration claim fails.

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

If this is right

  • Multi-hop graph RAG aimed at unseen domains should prefer GLM-style retrievers over pure GNN ones.
  • Single-hop retrieval can stay with ordinary dense vector search; graph machinery is unnecessary there.
  • Larger base LMs and larger subgraph budgets should continue to lift GLM transfer performance.
  • GNN retrievers initialized with node embeddings remain the efficient high-coverage alternative when the domain is known.

Where Pith is reading between the lines

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

  • Pairing the GLM retriever with richer KG-indexing schemes should compound rather than cancel the transfer gains.
  • The roughly 30× latency gap versus GNNs will need distillation or sparse attention before interactive deployment.
  • A hybrid router that sends deep-semantic queries to a GLM and broad-coverage queries to a GNN is a natural next system design.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 5 minor

Summary. The paper proposes GLM-RAG, replacing the GNN-based entity retriever in the GFM-RAG pipeline with a Graph Language Model (GLM) that tokenizes local KG subgraphs and scores entities via structure-aware attention over pretrained LM parameters. After finetuning on HotpotQA, 2Wiki, and MuSiQue, the authors compare GLM-RAG against vanilla dense RAG, GFM-RAG (with/without pretraining), and a node-embedding-augmented GFM-RAG+ across in-domain multi-hop QA, seven single-hop OOD sets, and multi-hop OOD benchmarks (MultihopRAG, G-Bench Novel/Medical/CS). They report that GLM-RAG is comparable in-domain, that vanilla RAG dominates single-hop settings, and that GLM-RAG yields substantially stronger zero-shot multi-hop transfer (e.g., MultihopRAG R@2 60.0 vs ~34–39 for GFM variants; SOTA evidence recall/accuracy on G-Bench Medical and CS), with supporting capacity-matched scaling, subgraph-budget ablations, semantic-vs-hop analyses, and Holm–Bonferroni bootstrap tests.

Significance. If the OOD multi-hop gains are genuinely attributable to deeper token-level semantic integration rather than visibility or sampling differences, the work supplies a clear, controlled demonstration that GLM-style retrievers are preferable foundation components for graph RAG under domain shift, while also clarifying when graph structure is unnecessary (single-hop). The experimental design is a strength: fixed KG construction and generator, multiple GFM variants including node-initialized GFM-RAG+, capacity-matched scaling (Fig. 6 / Table 19), loss and subgraph ablations, semantic similarity and hop-distance diagnostics (Tables 4–5), and significance testing (Table 22). These make the comparative claims actionable for the graph-RAG community even if absolute SOTA numbers are later surpassed by orthogonal indexing improvements (G-Reasoner).

major comments (2)
  1. [§5.5 Table 3; Table 2; App. B; Limitations] The central attribution—that superior OOD multi-hop transfer comes from GLM semantic integration (§3, §4.2.2, Fig. 2)—is confounded by unequal graph visibility. GLM is hard-capped at a hop-prioritized 2-hop neighborhood of ≤600 triplets (App. B; max_hops=2, max_triplets=600), while GFM variants run full-graph message passing. §5.5 Table 3 only restricts GFM-RAG* on the three in-domain Wiki sets and shows monotonic recall drop; it does not re-run MultihopRAG or G-Bench under matched budgets, nor report gold-document reachability under the same budget on those OOD graphs (Fig. 17 is in-domain only). Without a matched-visibility OOD experiment or explicit out-of-budget gold rates on the decisive sets, the large R@2 lift (Table 2: 60.0 vs 34–39) cannot be securely credited to semantics rather than coverage, sampling bias, or noise filtering. This is load-bearing for the SOTA/transfer claim.
  2. [Abstract; §5.2–5.3; Table 2; Table 17; Limitations] In-domain results (Figs. 3–4, Tables 16–17, 20) show GLM-RAG comparable to or slightly behind GFM-RAG+ on HotpotQA/2Wiki retrieval and only modestly ahead on MuSiQue; concurrent G-Reasoner (Luo et al., 2026) substantially leads on the same Wiki sets. The abstract and contribution list correctly emphasize OOD/SOTA on Medical/CS, but the manuscript should more sharply separate “comparable in-domain under restricted visibility” from “better foundation model,” and either integrate or more carefully bound claims relative to the newer indexing pipeline that the authors themselves call orthogonal and complementary.
minor comments (5)
  1. [Figure 1] Figure 1 caption and surrounding text usefully contrast retrieval traces, but the figure itself is hard to parse at column width; consider a simplified schematic or callouts for the decisive neighbor choice (GDR vs Olympics path).
  2. [§5.1 Metrics] Clarify early that document-level Recall@2 (not entity-level) is the primary retrieval metric and why top-k document mapping can hide entity ranking differences; this is stated in §5.1 but easy to miss when reading Tables 2/17.
  3. [§4.2.2; App. B] App. B notes differential LRs and ListCE masking for unscored entities; a one-sentence pointer in §4.2.2 would help readers understand why the loss differs from GFM-RAG without diving into the appendix.
  4. [Table 2; Table 9; §5.4.3] Table 9 and surrounding text correctly flag that MultihopRAG EM/ACC are weakly diagnostic because of strong no-context baselines; consider moving that caveat into the main Table 2 discussion so readers do not over-read QA columns.
  5. [Abstract; §5.4.3] Minor consistency: abstract says “achieving SOTA on two multi-hop benchmarks” while body ranks second on Novel; align wording.

Circularity Check

0 steps flagged

No circularity: empirical systems comparison on external benchmarks; method reuse is not a self-sealing derivation.

full rationale

GLM-RAG is an empirical ML systems paper. Its load-bearing claims (OOD Recall@2/EM/G-Bench gains for a finetuned GLM retriever vs GFM-RAG variants and vanilla RAG; in-domain parity; single-hop sufficiency of vector RAG; scaling trends) are measured on held-out and out-of-domain datasets with standard external metrics (document Recall@k, EM, F1, G-Bench evidence recall/accuracy). Training uses ordinary BCE+ListCE losses and hyperparameters; nothing is fitted to a target quantity and then re-reported as a prediction of that same quantity. Building on GFM-RAG’s pipeline and on the authors’ prior GLM architecture (Plenz & Frank 2024) is engineering reuse of a published encoder, not a uniqueness theorem or ansatz that forces the reported scores by construction. Ablations (subgraph budget, capacity-matched GNNs, linearized T5) and analyses (semantic similarity, hop distance) are comparative diagnostics, not circular reductions. Visibility mismatch between GLM’s ≤600-triplet subgraph and full-graph GNN message passing is a fairness/attribution concern for the OOD claim, not circularity. No step reduces a claimed result to its inputs by definition.

Axiom & Free-Parameter Ledger

5 free parameters · 5 axioms · 2 invented entities

Load-bearing premises are standard ML/IR assumptions plus engineering choices inherited from GFM-RAG and GLM, not new physical entities. Free parameters are training/architecture knobs (loss mix, LR, max_triplets, hops, T5 size). No invented particles or forces; the ‘GLM retriever’ is a composed system from prior components.

free parameters (5)
  • ListCE/BCE loss mix (α≈0.44 BCE / 0.56 ListCE) = 0.44 / 0.56
    Chosen after hyperparameter search; affects ranking objective on the visible subgraph only.
  • max_triplets and max_hops subgraph budget = 600 triplets, 2 hops
    Hard cap on GLM visibility (default 600 triplets, 2 hops); directly limits what can be ranked and is ablated but still a design choice that defines the method’s operating point.
  • Differential learning rates (head 5e-4, T5 backbone 1e-4) = 5e-4 / 1e-4
    Hand-set finetuning schedule for the GLM retriever.
  • T5 encoder size (small/base/large) = t5-large (main)
    Model capacity knob; main results use t5-large (~336M encoder params).
  • GFM-RAG+ query–node fusion design = all entities get query
    Choice to initialize all nodes as query_emb * node_emb rather than seed-only; selected because it scored higher on recall (Table 6).
axioms (5)
  • domain assumption Entity-level OpenIE/NER KGs plus seed entities from the query are an adequate interface for multi-hop document retrieval.
    Inherited from GFM-RAG pipeline (§4.1); document ranking is recovered by mapping entities back to source docs.
  • domain assumption GLM relative-position / structure-aware attention (Plenz & Frank 2024) preserves enough LM semantics while encoding graph topology.
    Core inductive bias of the proposed retriever (§3, §4.2.2); validated indirectly via ablations vs linearized T5.
  • domain assumption Recall@2 / document-level retrieval is the right primary proxy for downstream QA utility.
    Stated in §5.1 Metrics; QA sometimes decouples when the generator already knows the answer.
  • standard math Standard transformer + GNN training math (AdamW, BCE, listwise ranking, message passing).
    Used throughout training sections without modification beyond subgraph-masked ListCE.
  • ad hoc to paper gpt-4o-mini as fixed generator and LLM-as-judge does not invert method rankings.
    Single generator for all systems (§5.1); transfer QA tables show strong no-context baselines, so generator choice is load-bearing for QA claims.
invented entities (2)
  • GLM-RAG retriever (GLM entity scorer over tokenized local KG subgraph) no independent evidence
    purpose: Replace GNN message-passing ranker with end-to-end token-level graph-text ranking.
    Composition of existing GLM + GFM-RAG stack; not a new physical entity, but the paper’s named system contribution.
  • GFM-RAG+ (GNN with full node sentence-embedding init and query fusion) no independent evidence
    purpose: Stronger semantic GNN baseline to isolate GLM benefits beyond ‘adding text features’.
    Baseline invented for this paper’s comparison; useful control, not a claimed discovery.

pith-pipeline@v1.2.0-daily-grok45 · 34747 in / 3656 out tokens · 72028 ms · 2026-07-31T08:42:50.000897+00:00 · methodology

0 comments
read the original abstract

Retrieval-augmented generation (RAG) over knowledge graphs requires retrievers that can effectively capture both graph structure and semantic information. Recent approaches have explored graph neural network (GNN)-based retrievers to model graph topology in multi-hop reasoning tasks. In parallel, graph language models (GLMs) have emerged as a promising paradigm that integrates graph reasoning and the semantic capabilities of language models. In this work, we introduce a GLM-based retriever and investigate the comparative strengths of GLM-based, GNN-based, and traditional vector-search-based retrievers in single- and multi-hop RAG settings, and with a particular focus on transferability to unseen domains. Our findings suggest that finetuned GLM retrievers generalize better out of domain, achieving SOTA on two multi-hop benchmarks. On in-domain multi-hop QA datasets they remain comparable to prior work, with promising scaling as parameters and subgraph coverage increase. GNN-based retrievers achieve higher graph coverage with an efficient training setup, whereas the vector-search baseline excels at single-hop datasets.

Figures

Figures reproduced from arXiv: 2607.28397 by Anette Frank, Beni Egressy, Johann Higl, Maya Arseven, Moritz Plenz.

Figure 1
Figure 1. Figure 1: GLM-RAG and GFM-RAG knowledge graph retrieval traces for the query “What is the three letter abbreviation for the country, which maintains border troops, and claims Bernd Baumgart as a citizen?” with “GDR” being the golden answer. Example chosen from the test set of MuSiQue. knowledge-intensive tasks, they lag behind in tasks requiring multi-hop reasoning across multiple doc￾uments (Tang and Yang, 2024). G… view at source ↗
Figure 2
Figure 2. Figure 2: Differences between how GNN- and GLM-based retrievers reason over the KG for the query “ [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: shows retrieval performance on the test sets of the three Wikipedia-based datasets. All graph-enhanced methods outperform the RAG baseline, highlighting the contribution of graph RAG methods in multi-hop tasks. GFM-RAG and GFM-RAG* (i.e., with and without pretraining) show near-identical performance on most datasets, calling into question the need for costly pretraining. Hence, we also only finetune our GL… view at source ↗
Figure 4
Figure 4. Figure 4: QA performance (exact match) on in-domain [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Retrieval performance (Recall@2) on single-hop out-of-domain datasets. [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Model scaling comparison of GLM-RAG vs capacity-matched GFM-RAG* and GFM-RAG+. To conclude, we showed that GLM-RAG’s trans￾ferability claims still hold under matched visibil￾ity and capacity, implying that the generalizability gains stem from utilizing the existing large-scale pretrained LMs. The scalability potential of GLMs is an additional benefit that strongly indicates po￾tential for further improveme… view at source ↗
Figure 7
Figure 7. Figure 7: shows how both the GFM-RAG* and GLM-RAG models scale with increasing amounts of training data. Following this trend, and in order to reduce computational costs, we use a smaller subset of the data consisting of approximately 60k queries and 700k documents, instead of approxi￾mately 277k queries and 2.9M documents, for the following ablation analyses: 15k 30k 60k 277k Number of Q-Doc Pairs 0.58 0.60 0.62 0.… view at source ↗
Figure 8
Figure 8. Figure 8: An example test question from 2Wiki. D.2 Retrieval Performance by Subgraph Budget In order to analyze the subgraph budget behavior on different neighborhoods, we separate questions into three categories, namely sparse, medium and dense, using the number of edges around the seed entities. The density distribution of the queries are more on the sparse side, however we account for this by separating the queri… view at source ↗
Figure 9
Figure 9. Figure 9: Retrieval performance (Recall@2 and Recall@5) on in-domain Wikipedia datasets. [PITH_FULL_IMAGE:figures/full_fig_p020_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Retrieval performance (Recall@2 and Recall@5) on 7 out-of-domain single-hop datasets. [PITH_FULL_IMAGE:figures/full_fig_p021_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Retrieval performance (Recall@2 and Recall@5) on MultiHopRAG, an out-of-domain dataset with [PITH_FULL_IMAGE:figures/full_fig_p022_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Comparision of GFM-RAG, GFM-RAG+ and GLM-RAG’s retrieval quality on the three in-domain [PITH_FULL_IMAGE:figures/full_fig_p024_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Recall degradation with increasing number of supporting documents for multi-hop datasets. [PITH_FULL_IMAGE:figures/full_fig_p025_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Recall degradation with increasing number of supporting documents for single-hop datasets. [PITH_FULL_IMAGE:figures/full_fig_p026_14.png] view at source ↗
Figure 15
Figure 15. Figure 15: Recall degradation with increasing entity-level distance between golden and retrieved entities for [PITH_FULL_IMAGE:figures/full_fig_p026_15.png] view at source ↗
Figure 16
Figure 16. Figure 16: Recall degradation with increasing entity-level distance between golden and retrieved entities for [PITH_FULL_IMAGE:figures/full_fig_p026_16.png] view at source ↗
Figure 17
Figure 17. Figure 17: Golden document reachability of GLM-RAG split into 3 question categories regarding their neighborhood [PITH_FULL_IMAGE:figures/full_fig_p027_17.png] view at source ↗
Figure 18
Figure 18. Figure 18: Retrieval recall of GLM-RAG and GFM-RAG* split into 3 question categories regarding their neighbor [PITH_FULL_IMAGE:figures/full_fig_p028_18.png] view at source ↗
Figure 19
Figure 19. Figure 19: QA performance (Exact Match and F1) on in-domain Wikipedia datasets with [PITH_FULL_IMAGE:figures/full_fig_p028_19.png] view at source ↗
Figure 20
Figure 20. Figure 20: The one-shot prompt template for 2Wiki. MuSiQue and HotpotQA have similar templates, adjusted to [PITH_FULL_IMAGE:figures/full_fig_p029_20.png] view at source ↗

discussion (0)

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

Reference graph

Works this paper leans on

55 extracted references · 8 canonical work pages

  1. [1]

    Aijun Bai, Rolf Jagerman, Zhen Qin, Le Yan, Pratyush Kar, Bing-Rong Lin, Xuanhui Wang, Michael Bendersky, and Marc Najork. 2023. http://arxiv.org/abs/2211.01494 Regression compatible listwise objectives for calibrated ranking with binary relevance

    Pith/arXiv arXiv 2023

  2. [2]

    Payal Bajaj, Daniel Campos, Nick Craswell, Li Deng, Jianfeng Gao, Xiaodong Liu, Rangan Majumder, Andrew McNamara, Bhaskar Mitra, Tri Nguyen, Mir Rosenberg, Xia Song, Alina Stoica, Saurabh Tiwary, and Tong Wang. 2018. http://arxiv.org/abs/1611.09268 Ms marco: A human generated machine reading comprehension dataset

    Pith/arXiv arXiv 2018

  3. [3]

    Vittorio Castelli, Rishav Chakravarti, Saswati Dana, Anthony Ferritto, Radu Florian, Martin Franz, Dinesh Garg, Dinesh Khandelwal, Scott McCarley, Michael McCawley, Mohamed Nasr, Lin Pan, Cezar Pendus, John Pitrelli, Saurabh Pujar, Salim Roukos, Andrzej Sakrajda, Avi Sil, Rosario Uceda-Sosa, Todd Ward, and Rong Zhang. 2020. https://doi.org/10.18653/v1/202...

    work page doi:10.18653/v1/2020.acl-main.117 2020

  4. [4]

    Tianlong Chen, Kaixiong Zhou, Keyu Duan, Wenqing Zheng, Peihao Wang, Xia Hu, and Zhangyang Wang. 2022. https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=9773017 Bag of tricks for training deeper graph neural networks: A comprehensive benchmark study . IEEE Transactions on Pattern Analysis and Machine Intelligence, 45(3):2769--2781

    2022

  5. [5]

    Mostafa Dehghani, Josip Djolonga, Basil Mustafa, Piotr Padlewski, Jonathan Heek, Justin Gilmer, Andreas Peter Steiner, Mathilde Caron, Robert Geirhos, Ibrahim Alabdulmohsin, Rodolphe Jenatton, Lucas Beyer, Michael Tschannen, Anurag Arnab, Xiao Wang, Carlos Riquelme Ruiz, Matthias Minderer, Joan Puigcerver, Utku Evci, Manoj Kumar, Sjoerd Van Steenkiste, Ga...

    2023

  6. [6]

    Darren Edge, Ha Trinh, Newman Cheng, Joshua Bradley, Alex Chao, Apurva Mody, Steven Truitt, Dasha Metropolitansky, Robert Osazuwa Ness, and Jonathan Larson. 2024. https://arxiv.org/abs/2404.16130 From local to global: A graph rag approach to query-focused summarization . arXiv preprint arXiv:2404.16130

    Pith/arXiv arXiv 2024

  7. [7]

    Beni Egressy and Jan Stühmer. 2025. http://arxiv.org/abs/2505.15433 Set-llm: A permutation-invariant llm

    Pith/arXiv arXiv 2025

  8. [8]

    Mikhail Galkin, Xinyu Yuan, Hesham Mostafa, Jian Tang, and Zhaocheng Zhu. 2024. https://openreview.net/forum?id=jVEoydFOl9 Towards foundation models for knowledge graph reasoning . In The Twelfth International Conference on Learning Representations

    2024

  9. [9]

    Yunfan Gao, Yun Xiong, Xinyu Gao, Kangxiang Jia, Jinliu Pan, Yuxi Bi, Yi Dai, Jiawei Sun, Meng Wang, and Haofen Wang. 2024. http://arxiv.org/abs/2312.10997 Retrieval-augmented generation for large language models: A survey

    Pith/arXiv arXiv 2024

  10. [10]

    Justin Gilmer, Samuel S Schoenholz, Patrick F Riley, Oriol Vinyals, and George E Dahl. 2017. https://arxiv.org/abs/1704.01212 Neural message passing for quantum chemistry . In International conference on machine learning, pages 1263--1272. Pmlr

    Pith/arXiv arXiv 2017

  11. [11]

    Haisong Gong, Zhibo Liu, Qiang Liu, Shu Wu, and Liang Wang. 2026. https://arxiv.org/abs/2601.22657 Nag: A unified native architecture for encoder-free text-graph modeling in language models . arXiv preprint arXiv:2601.22657

    arXiv 2026

  12. [12]

    Zirui Guo, Lianghao Xia, Yanhua Yu, Tu Ao, and Chao Huang. 2025. http://arxiv.org/abs/2410.05779 Lightrag: Simple and fast retrieval-augmented generation

    Pith/arXiv arXiv 2025

  13. [13]

    Bernal Jiménez Gutiérrez, Yiheng Shu, Yu Gu, Michihiro Yasunaga, and Yu Su. 2024. https://openreview.net/forum?id=hkujvAPVsg Hipporag: Neurobiologically inspired long-term memory for large language models . In The Thirty-eighth Annual Conference on Neural Information Processing Systems

    2024

  14. [14]

    Bernal Jiménez Gutiérrez, Yiheng Shu, Weijian Qi, Sizhe Zhou, and Yu Su. 2025. http://arxiv.org/abs/2502.14802 From rag to memory: Non-parametric continual learning for large language models

    Pith/arXiv arXiv 2025

  15. [15]

    Xiaoxin He, Yijun Tian, Yifei Sun, Nitesh Chawla, Thomas Laurent, Yann LeCun, Xavier Bresson, and Bryan Hooi. 2024. https://arxiv.org/abs/2402.07630 G-retriever: Retrieval-augmented generation for textual graph understanding and question answering . Advances in Neural Information Processing Systems, 37:132876--132907

    Pith/arXiv arXiv 2024

  16. [16]

    Xanh Ho, Anh-Khoa Duong Nguyen, Saku Sugawara, and Akiko Aizawa. 2020. https://doi.org/10.18653/v1/2020.coling-main.580 Constructing a multi-hop QA dataset for comprehensive evaluation of reasoning steps . In Proceedings of the 28th International Conference on Computational Linguistics, pages 6609--6625, Barcelona, Spain (Online). International Committee ...

    doi:10.18653/v1/2020.coling-main.580 2020

  17. [17]

    Qiao Jin, Bhuwan Dhingra, Zhengping Liu, William Cohen, and Xinghua Lu. 2019. https://doi.org/10.18653/v1/D19-1259 P ub M ed QA : A dataset for biomedical research question answering . In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-I...

    doi:10.18653/v1/d19-1259 2019

  18. [18]

    Ehsan Kamalloo, Aref Jafari, Xinyu Zhang, Nandan Thakur, and Jimmy Lin. 2023. https://arxiv.org/abs/2307.16883 Hagrid: A human-llm collaborative dataset for generative information-seeking with attribution . arXiv preprint arXiv:2307.16883

    Pith/arXiv arXiv 2023

  19. [19]

    Jared Kaplan, Sam McCandlish, Tom Henighan, Tom B Brown, Benjamin Chess, Rewon Child, Scott Gray, Alec Radford, Jeffrey Wu, and Dario Amodei. 2020. https://arxiv.org/pdf/2001.08361/1000 Scaling laws for neural language models . arXiv preprint arXiv:2001.08361

    Pith/arXiv arXiv 2020

  20. [20]

    Vladimir Karpukhin, Barlas Oguz, Sewon Min, Patrick Lewis, Ledell Wu, Sergey Edunov, Danqi Chen, and Wen-tau Yih. 2020. https://doi.org/10.18653/v1/2020.emnlp-main.550 Dense passage retrieval for open-domain question answering . In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 6769--6781, Online. Ass...

    doi:10.18653/v1/2020.emnlp-main.550 2020

  21. [21]

    u ttler, Mike Lewis, Wen-tau Yih, Tim Rockt\

    Patrick Lewis, Ethan Perez, Aleksandra Piktus, Fabio Petroni, Vladimir Karpukhin, Naman Goyal, Heinrich K\" u ttler, Mike Lewis, Wen-tau Yih, Tim Rockt\" a schel, Sebastian Riedel, and Douwe Kiela. 2020. https://dl.acm.org/doi/10.5555/3495724.3496517 Retrieval-augmented generation for knowledge-intensive nlp tasks . In Proceedings of the 34th Internationa...

    arXiv 2020

  22. [22]

    Junyi Li, Tianyi Tang, Wayne Xin Zhao, Zhicheng Wei, Nicholas Jing Yuan, and Ji-Rong Wen. 2021. https://aclanthology.org/2021.findings-acl.136/ F ew-shot K nowledge G raph-to- T ext G eneration with P retrained L anguage M odels . In ACL Findings

    2021

  23. [23]

    Mufei Li, Siqi Miao, and Pan Li. 2025. Simple is effective: The roles of graphs and large language models in knowledge-graph-based retrieval-augmented generation. In The Thirteenth International Conference on Learning Representations

    2025

  24. [24]

    Qimai Li, Zhichao Han, and Xiao-ming Wu. 2018. https://doi.org/10.1609/aaai.v32i1.11604 Deeper insights into graph convolutional networks for semi-supervised learning . Proceedings of the AAAI Conference on Artificial Intelligence, 32(1)

    doi:10.1609/aaai.v32i1.11604 2018

  25. [25]

    Lei Liang, Zhongpu Bo, Zhengke Gui, Zhongshu Zhu, Ling Zhong, Peilong Zhao, Mengshu Sun, Zhiqiang Zhang, Jun Zhou, Wenguang Chen, Wen Zhang, and Huajun Chen. 2025. https://doi.org/10.1145/3701716.3715240 Kag: Boosting llms in professional domains via knowledge augmented generation . In Companion Proceedings of the ACM on Web Conference 2025, WWW '25, page...

    arXiv 2025

  26. [26]

    Bill Yuchen Lin, Xinyue Chen, Jamin Chen, and Xiang Ren. 2019. https://doi.org/10.18653/v1/D19-1282 K ag N et: Knowledge-aware graph networks for commonsense reasoning . In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP), pages 2...

    work page doi:10.18653/v1/d19-1282 2019

  27. [27]

    Linhao Luo, Zicheng Zhao, Gholamreza Haffari, Dinh Phung, Chen Gong, and Shirui Pan. 2025. https://openreview.net/forum?id=0QNmAvQQqj GFM - RAG : Graph foundation model for retrieval augmented generation . In The Thirty-ninth Annual Conference on Neural Information Processing Systems

    2025

  28. [28]

    Linhao Luo, Zicheng Zhao, Junnan Liu, Zhangchi Qiu, Junnan Dong, Serge Panev, Chen Gong, Thuy-Trang Vu, Gholamreza Haffari, Dinh Phung, Alan Wee-Chung Liew, and Shirui Pan. 2026. https://openreview.net/forum?id=zJm9nmoahk G-reasoner: Foundation models for unified reasoning over graph-structured knowledge . In The Fourteenth International Conference on Lea...

    2026

  29. [29]

    Chaitanya Malaviya, Chandra Bhagavatula, Antoine Bosselut, and Yejin Choi. 2020. https://ojs.aaai.org/index.php/AAAI/article/view/5684/5540&ved=2ahUKEwjLju2_htWUAxWv8LsIHZWSJl8QFnoECBoQAQ&usg=AOvVaw0ZnUq9R655--px5T6vFOMX Commonsense knowledge base completion with structural and semantic context . Proceedings of the 34th AAAI Conference on Artificial Intelligence

    2020

  30. [30]

    Chaitanya Malaviya, Subin Lee, Sihao Chen, Elizabeth Sieber, Mark Yatskar, and Dan Roth. 2024. https://doi.org/10.18653/v1/2024.naacl-long.167 E xpert QA : Expert-curated questions and attributed answers . In Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volu...

    doi:10.18653/v1/2024.naacl-long.167 2024

  31. [31]

    Costas Mavromatis and George Karypis. 2025. https://doi.org/10.18653/v1/2025.findings-acl.856 GNN - RAG : Graph neural retrieval for efficient large language model reasoning on knowledge graphs . In Findings of the Association for Computational Linguistics: ACL 2025, pages 16682--16699, Vienna, Austria. Association for Computational Linguistics

    doi:10.18653/v1/2025.findings-acl.856 2025

  32. [32]

    Abhilash Nandy, Soumya Sharma, Shubham Maddhashiya, Kapil Sachdeva, Pawan Goyal, and NIloy Ganguly. 2021. https://doi.org/10.18653/v1/2021.findings-emnlp.392 Question answering over electronic devices: A new benchmark dataset and a multi-task learning based QA framework . In Findings of the Association for Computational Linguistics: EMNLP 2021, pages 4600...

    work page doi:10.18653/v1/2021.findings-emnlp.392 2021

  33. [33]

    Kenta Oono and Taiji Suzuki. 2020. https://openreview.net/forum?id=S1ldO2EFPr Graph neural networks exponentially lose expressive power for node classification . In International Conference on Learning Representations

    2020

  34. [34]

    Moritz Plenz and Anette Frank. 2024. https://doi.org/10.18653/v1/2024.acl-long.245 Graph language models . In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 4477--4494, Bangkok, Thailand. Association for Computational Linguistics

    work page doi:10.18653/v1/2024.acl-long.245 2024

  35. [35]

    Moritz Plenz, Juri Opitz, Philipp Heinisch, Philipp Cimiano, and Anette Frank. 2023. https://doi.org/10.18653/v1/2023.acl-long.338 Similarity-weighted construction of contextualized commonsense knowledge graphs for knowledge-intense argumentation tasks . In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: ...

    work page doi:10.18653/v1/2023.acl-long.338 2023

  36. [36]

    Nils Reimers and Iryna Gurevych. 2019. https://doi.org/10.18653/v1/D19-1410 Sentence- BERT : Sentence embeddings using S iamese BERT -networks . In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP), pages 3982--3992, Hong Kong, Chi...

    doi:10.18653/v1/d19-1410 2019

  37. [37]

    Leonardo F. R. Ribeiro, Martin Schmitt, Hinrich Sch \"u tze, and Iryna Gurevych. 2021. https://doi.org/10.18653/v1/2021.nlp4convai-1.20 Investigating pretrained language models for graph-to-text generation . In Proceedings of the 3rd Workshop on Natural Language Processing for Conversational AI, pages 211--227, Online. Association for Computational Linguistics

    work page doi:10.18653/v1/2021.nlp4convai-1.20 2021

  38. [38]

    Stephen E Robertson and Steve Walker. 1994. https://www.staff.city.ac.uk/ sbrp622/papers/robertson_walker_sigir94.pdf Some simple effective approximations to the 2-poisson model for probabilistic weighted retrieval . In SIGIR’94: Proceedings of the Seventeenth Annual International ACM-SIGIR Conference on Research and Development in Information Retrieval, ...

    1994

  39. [39]

    Yu Rong, Wenbing Huang, Tingyang Xu, and Junzhou Huang. 2020. https://openreview.net/forum?id=Hkx1qkrKPr Dropedge: Towards deep graph convolutional networks on node classification . In International Conference on Learning Representations

    2020

  40. [40]

    Mobashir Sadat, Zhengyu Zhou, Lukas Lange, Jun Araki, Arsalan Gundroo, Bingqing Wang, Rakesh Menon, Md Parvez, and Zhe Feng. 2023. https://doi.org/10.18653/v1/2023.findings-emnlp.59 D elucion QA : Detecting hallucinations in domain-specific question answering . In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 822--835, Singa...

    work page doi:10.18653/v1/2023.findings-emnlp.59 2023

  41. [41]

    Keshav Santhanam, Omar Khattab, Jon Saad-Falcon, Christopher Potts, and Matei Zaharia. 2022. https://doi.org/10.18653/v1/2022.naacl-main.272 C ol BERT v2: Effective and efficient retrieval via lightweight late interaction . In Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language ...

    doi:10.18653/v1/2022.naacl-main.272 2022

  42. [42]

    Parth Sarthi, Salman Abdullah, Aditi Tuli, Shubh Khanna, Anna Goldie, and Christopher D Manning. 2024. https://arxiv.org/abs/2401.18059 Raptor: Recursive abstractive processing for tree-organized retrieval . In The Twelfth International Conference on Learning Representations

    Pith/arXiv arXiv 2024

  43. [43]

    Martin Schmitt, Sahand Sharifzadeh, Volker Tresp, and Hinrich Sch \"u tze. 2020. https://doi.org/10.18653/v1/2020.emnlp-main.577 An unsupervised joint system for text generation from knowledge graphs and semantic parsing . In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 7117--7130, Online. Associati...

    work page doi:10.18653/v1/2020.emnlp-main.577 2020

  44. [44]

    Yixuan Tang and Yi Yang. 2024. https://openreview.net/forum?id=t4eB3zYWBK Multihop- RAG : Benchmarking retrieval-augmented generation for multi-hop queries . In First Conference on Language Modeling

    2024

  45. [45]

    Harsh Trivedi, Niranjan Balasubramanian, Tushar Khot, and Ashish Sabharwal. 2022. https://doi.org/10.1162/tacl_a_00475 M u S i Q ue: Multihop questions via single-hop question composition . Transactions of the Association for Computational Linguistics, 10:539--554

    doi:10.1162/tacl_a_00475 2022

  46. [46]

    Dario Vajda. 2026. http://arxiv.org/abs/2605.10247 Teaching llms to see graphs: Unifying text and structural reasoning

    Pith/arXiv arXiv 2026

  47. [47]

    Ning Wang, Kuanyan Zhu, Daniel Yuehwoon Yee, Yitang Gao, Shiying Huang, Zirun Xu, and Sainyam Galhotra. 2026. https://arxiv.org/pdf/2602.04926 Pruning minimal reasoning graphs for efficient retrieval-augmented generation . arXiv preprint arXiv:2602.04926

    arXiv 2026

  48. [48]

    Zhishang Xiang, Chuanjie Wu, Qinggang Zhang, Shengyuan Chen, Zijin Hong, Xiao Huang, and Jinsong Su. 2025. When to use graphs in rag: A comprehensive analysis for graph retrieval-augmented generation. arXiv preprint arXiv:2506.05690

    arXiv 2025

  49. [49]

    Yilin Xiao, Junnan Dong, Chuang Zhou, Su Dong, Qian wen Zhang, Di Yin, Xing Sun, and Xiao Huang. 2025. http://arxiv.org/abs/2506.02404 Graphrag-bench: Challenging domain-specific reasoning for evaluating graph retrieval-augmented generation

    Pith/arXiv arXiv 2025

  50. [50]

    Masatsugu Yamada and Mahito Sugiyama. 2026. http://arxiv.org/abs/2605.06239 When graph language models go beyond memorization

    Pith/arXiv arXiv 2026

  51. [51]

    Zhilin Yang, Peng Qi, Saizheng Zhang, Yoshua Bengio, William Cohen, Ruslan Salakhutdinov, and Christopher D Manning. 2018. https://aclanthology.org/D18-1259 Hotpotqa: A dataset for diverse, explainable multi-hop question answering . In Proceedings of the 2018 conference on empirical methods in natural language processing, pages 2369--2380

    2018

  52. [52]

    Michihiro Yasunaga, Antoine Bosselut, Hongyu Ren, Xikun Zhang, Christopher D Manning, Percy Liang, and Jure Leskovec. 2022. https://openreview.net/forum?id=4NpoSrT8uU- Deep bidirectional language-knowledge graph pretraining . In Advances in Neural Information Processing Systems

    2022

  53. [53]

    Yanzhao Zhang, Mingxin Li, Dingkun Long, Xin Zhang, Huan Lin, Baosong Yang, Pengjun Xie, An Yang, Dayiheng Liu, Junyang Lin, et al. 2025. https://arxiv.org/abs/2506.05176 Qwen3 embedding: Advancing text embedding and reranking through foundation models . arXiv preprint arXiv:2506.05176

    Pith/arXiv arXiv 2025

  54. [54]

    Yue Zhang, Yafu Li, Leyang Cui, Deng Cai, Lemao Liu, Tingchen Fu, Xinting Huang, Enbo Zhao, Yu Zhang, Yulong Chen, et al. 2023. https://arxiv.org/abs/2309.01219 Siren's song in the ai ocean: A survey on hallucination in large language models . arXiv preprint arXiv:2309.01219

    Pith/arXiv arXiv 2023

  55. [55]

    Jianan Zhao, Meng Qu, Chaozhuo Li, Hao Yan, Qian Liu, Rui Li, Xing Xie, and Jian Tang. 2023. https://openreview.net/forum?id=q0nmYciuuZN Learning on large-scale text-attributed graphs via variational inference . In The Eleventh International Conference on Learning Representations

    2023