pith. sign in

arxiv: 2605.18570 · v1 · pith:23MJ3KXUnew · submitted 2026-05-18 · 💻 cs.AI

Query-Conditioned Knowledge Alignment for Reliable Cross-System Medical Reasoning

Yan Jiao , Jingran Xu , Pin-Han Ho , Limei Peng This is my paper

Pith reviewed 2026-05-20 10:51 UTC · model grok-4.3

classification 💻 cs.AI
keywords entity alignmentknowledge graphsmedical reasoningquery-conditioned alignmentcross-domain integrationTCM-WMretrieval-augmented generation
0
0 comments X

The pith

Treating source entity text as a query allows context-dependent ranking of matches in target medical knowledge graphs.

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

The paper proposes Query-Conditioned Entity Alignment to move beyond static mappings that ignore context, direction, and non-bijective relations when integrating different medical systems. It reframes alignment as ranking candidates in one graph based on a textual query from the other, using semantic encoding, graph learning, and a direction-aware module. This setup is tested on TCM-WM graphs for symptom and herb-molecule tasks, yielding gains in ranking scores and better performance when the alignments feed into retrieval-augmented generation. A reader would care because reliable cross-system medical reasoning depends on how accurately evidence from one tradition can be located and used in the other.

Core claim

QCEA reformulates entity alignment as a query-conditioned correspondence problem. The textual description of a source entity is treated as a query that ranks candidate entities in the target graph, with the framework combining semantic encoding, graph-based representation learning, and a direction-aware transformation module to capture asymmetric and many-to-many correspondences across heterogeneous medical knowledge systems.

What carries the argument

Query-Conditioned Entity Alignment (QCEA), which ranks target-graph candidates by treating a source entity's text as a query to enable context-dependent and direction-sensitive matching.

If this is right

  • Higher Hit@K and MRR scores on both symptom alignment and herb-molecule alignment tasks from TCM-WM graphs.
  • Improved evidence retrieval quality when the aligned entities are used inside retrieval-augmented generation pipelines.
  • Stronger grounding and measurably higher answer accuracy in cross-system medical reasoning experiments.
  • Alignment viewed as an active contributor to knowledge accessibility rather than a passive preprocessing step.

Where Pith is reading between the lines

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

  • The same query-conditioned ranking pattern could be applied to align other heterogeneous knowledge bases where relations are context-dependent and non-bijective.
  • Medical AI systems that combine multiple traditions might reduce retrieval errors by adopting query context as a standard alignment step.
  • Scalability tests on larger or noisier medical graphs would show whether the direction-aware module remains effective without extra adjustments.

Load-bearing premise

Semantic encoding together with graph representations and a direction-aware transformation can reliably capture asymmetric and many-to-many correspondences without dataset-specific tuning that would change the reported rank metrics.

What would settle it

Evaluating QCEA on a fresh pair of medical knowledge graphs and finding no gains over standard alignment baselines on Hit@K, MRR, or downstream RAG answer accuracy would falsify the central claim.

Figures

Figures reproduced from arXiv: 2605.18570 by Jingran Xu, Limei Peng, Pin-Han Ho, Yan Jiao.

Figure 1
Figure 1. Figure 1: Cross-system semantic misalignment and query￾conditioned entity alignment. (a) Context-dependent ambigu￾ity in cross-system correspondence. (b) Description-agnostic entity-based alignment produces identical rankings for different descriptions. (c) QCEA enables query-conditioned alignment with context-dependent outputs. or one-to-one matching, highlighting the need for more flexible alignment mechanisms. Th… view at source ↗
Figure 2
Figure 2. Figure 2: Overview of the proposed QCEA framework. Query descriptions are encoded into query representations. TCM and WM entities are mapped into graph-aware embeddings. A direction-aware Tucker projection module, conditioned on the alignment direction 𝑠, produces target representations for scoring and top-𝑘 ranking under a direction-weighted, many-to-many-aware contrastive objective. What is alignment? Alignment is… view at source ↗
Figure 3
Figure 3. Figure 3: Training dynamics and retrieval performance on the Symptom and Herb datasets. (a) Training and validation loss with selected best epochs. (b) Overall Hit@K and Recall@K. (c)-(d) Directional performance (TCM→WM and WM→TCM). (e)-(f) Performance under different ground-truth cardinalities (GT=1 vs. GT>1). Solid lines denote Symptom, dashed lines denote Herb, and 𝐾 is shown on a logarithmic scale. parametric kn… view at source ↗
Figure 4
Figure 4. Figure 4: Impact of seed alignment ratio on (a) Symptom and (b) Herb tasks. Performance improves with increasing supervision. Recall@100 is used for Herb. NoAlign. The retriever remains within the TCM subgraph and fails to access cross-system alignment edges, resulting in missing evidence and ungrounded answers. QCEA. QCEA retrieves multiple cross-system candidates (e.g., Luteolin, Apigenin, Protocatechuic Acid), en… view at source ↗
Figure 5
Figure 5. Figure 5: Downstream RAG evaluation under different alignment settings. (a) Overall comparison in terms of retrieval-level (evidence recall@K, cross-system hit rate), generation-level (answer accuracy), and end-to-end metrics (groundedness, end-to￾end accuracy). (b) Category-wise end-to-end accuracy for Symptom and Herb questions. (c) Effect of confidence-based top-𝑘 truncation of first-hop alignment candidates. (d)… view at source ↗
read the original abstract

Cross-domain knowledge alignment is essential for integrating heterogeneous medical systems, yet existing approaches typically treat entity alignment as a static matching problem, ignoring query context and cross-system asymmetry. This limitation is particularly critical in integrative medical settings, where correspondence between concepts is inherently context-dependent, non-bijective, and direction-sensitive. In this paper, we propose Query-Conditioned Entity Alignment (QCEA), which reformulates entity alignment as a query-conditioned correspondence problem. Instead of learning a fixed mapping between entity representations, QCEA treats the textual description of a source entity as a query and ranks candidate entities in the target graph, enabling context-dependent alignment. The framework integrates semantic encoding, graph-based representation learning, and a direction-aware transformation module to capture asymmetric and many-to-many correspondence across heterogeneous knowledge systems. We evaluate QCEA on TCM--WM knowledge graphs derived from SymMap, covering both symptom alignment and herb--molecule alignment tasks. Experimental results show consistent improvements over representative baselines, particularly on rank-sensitive metrics such as Hit@K and MRR. Furthermore, downstream retrieval-augmented generation (RAG) experiments demonstrate that improved alignment leads to better evidence retrieval, stronger grounding, and higher answer accuracy. These findings highlight that alignment is not merely a data integration step, but a key factor that shapes knowledge accessibility and reliability in cross-system medical reasoning.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The manuscript proposes Query-Conditioned Entity Alignment (QCEA) for integrating heterogeneous medical knowledge systems such as TCM and WM. It reformulates entity alignment as a query-conditioned ranking problem in which the textual description of a source entity serves as a query to rank candidate entities in the target graph. The framework combines semantic encoding, graph-based representation learning, and a direction-aware transformation module to handle context-dependent, asymmetric, and many-to-many correspondences. Experiments on TCM-WM knowledge graphs derived from SymMap report consistent gains over baselines on Hit@K and MRR for symptom and herb-molecule alignment tasks, with additional improvements shown in downstream RAG experiments for evidence retrieval, grounding, and answer accuracy.

Significance. If the reported gains prove robust, the work could meaningfully advance reliable cross-system medical reasoning by shifting alignment from static mappings to query-conditioned, direction-sensitive processes. The downstream RAG results provide a concrete link between improved alignment and practical gains in retrieval-augmented medical QA, addressing a recognized bottleneck in integrative medicine AI applications.

major comments (2)
  1. Abstract and experimental section: the abstract reports consistent gains on rank metrics and RAG accuracy, yet provides no error bars, no details on data splits or exclusion rules, and no ablation isolating the direction-aware module. These omissions leave the central claim only partially supported, as the evaluation depends on the specific TCM-WM graphs without clear controls for variability or module contribution.
  2. Method description (direction-aware transformation): the claim that the module reliably captures asymmetric and many-to-many correspondences is central, but the manuscript supplies no equations, architecture details, or ablation results showing its incremental effect on direction sensitivity. If the module reduces to a standard linear map or attention, the observed Hit@K and MRR gains could be driven by query reformulation or graph learning alone.
minor comments (1)
  1. Abstract: the phrase 'direction-aware transformation module' would benefit from a one-sentence clarification of its concrete implementation (e.g., explicit direction modeling versus standard attention) to help readers assess novelty relative to prior alignment work.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed comments. We address each major point below with clarifications from the manuscript and commit to targeted revisions that will improve transparency and support for the central claims.

read point-by-point responses

  1. Referee: Abstract and experimental section: the abstract reports consistent gains on rank metrics and RAG accuracy, yet provides no error bars, no details on data splits or exclusion rules, and no ablation isolating the direction-aware module. These omissions leave the central claim only partially supported, as the evaluation depends on the specific TCM-WM graphs without clear controls for variability or module contribution.

    Authors: We acknowledge that the abstract is concise and omits these details. The full experimental section describes the TCM-WM graphs derived from SymMap, including the construction process, entity filtering, and the 70/15/15 train/validation/test splits used for both symptom and herb-molecule tasks. To address the concern directly, the revised manuscript will add (i) error bars from five independent runs with different random seeds for all Hit@K and MRR results, (ii) explicit statements of exclusion rules (e.g., removal of entities with fewer than three relations), and (iii) a dedicated ablation table isolating the direction-aware module. These additions will provide clearer controls for variability and module contribution. revision: yes

  2. Referee: Method description (direction-aware transformation): the claim that the module reliably captures asymmetric and many-to-many correspondences is central, but the manuscript supplies no equations, architecture details, or ablation results showing its incremental effect on direction sensitivity. If the module reduces to a standard linear map or attention, the observed Hit@K and MRR gains could be driven by query reformulation or graph learning alone.

    Authors: The direction-aware transformation is presented in Section 3.3 as a learnable module that applies a direction-specific projection followed by an attention mechanism conditioned on the query embedding to model asymmetry and many-to-many mappings. We agree that the current draft lacks explicit equations and a focused ablation. In the revision we will insert the full mathematical formulation (including the directional transformation matrix and attention equations) and add an ablation study that reports performance with and without the module on both alignment and downstream RAG tasks. This will demonstrate its incremental contribution beyond query reformulation and graph learning. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical framework with independent experimental validation

full rationale

The paper defines QCEA as an architectural integration of semantic encoding, graph representation learning, and a direction-aware transformation module, then evaluates it empirically on TCM-WM graphs derived from SymMap for symptom and herb-molecule alignment tasks. Reported gains on Hit@K, MRR, and downstream RAG metrics are presented as experimental outcomes rather than closed-form derivations or predictions that reduce by construction to fitted parameters or self-citations. No equations appear that equate outputs to inputs tautologically, and the central claims rest on observable performance differences against baselines, rendering the chain self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The framework rests on standard graph representation learning and semantic encoding techniques drawn from prior work, plus the novel direction-aware transformation whose effectiveness is shown only empirically on the chosen datasets.

axioms (1)
  • domain assumption Entity descriptions can be treated as effective queries for ranking in a heterogeneous target graph
    Invoked in the reformulation of alignment as query-conditioned correspondence

pith-pipeline@v0.9.0 · 5773 in / 1212 out tokens · 23176 ms · 2026-05-20T10:51:51.641681+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

46 extracted references · 46 canonical work pages · 7 internal anchors

  1. [1]

    Healthcare knowledge graph construction: A systematic review of the state-of-the-art, open issues, and opportuni- ties

    Abu-Salih,B.,Al-Qurishi,M.,Alweshah,M.,Al-Smadi,M.,Alfayez, R., Saadeh, H., 2023. Healthcare knowledge graph construction: A systematic review of the state-of-the-art, open issues, and opportuni- ties. Journal of Big Data 10, 81

    work page 2023

  2. [2]

    Achiam, J., Adler, S., Agarwal, S., Ahmad, L., Akkaya, I., Aleman, F.L., Almeida, D., Altenschmidt, J., Altman, S., Anadkat, S., et al.,

    work page

  3. [3]

    GPT-4 Technical Report

    Gpt-4 technical report. arXiv preprint arXiv:2303.08774 . Y. Jiao et al.:Preprint submitted to ElsevierPage 13 of 15 Query-Conditioned Knowledge Alignment

    work page internal anchor Pith review Pith/arXiv arXiv

  4. [4]

    gpt-oss-120b & gpt-oss-20b Model Card

    Agarwal, S., Ahmad, L., Ai, J., Altman, S., Applebaum, A., Arbus, E.,Arora,R.K.,Bai,Y.,Baker,B.,Bao,H.,etal.,2025. gpt-oss-120b & gpt-oss-20b model card. arXiv preprint arXiv:2508.10925

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  5. [5]

    The unified medical language system (UMLS): integrating biomedical terminology

    Bodenreider, O., 2004. The unified medical language system (UMLS): integrating biomedical terminology. Nucleic Acids Re- search 32, D267–D270

    work page 2004

  6. [6]

    Multilingualknowl- edge graph embeddings for cross-lingual knowledge alignment, in: Proceedings of the 26th International Joint Conference on Artificial Intelligence (IJCAI), pp

    Chen,M.,Tian,Y.,Yang,M.,Zaniolo,C.,2017. Multilingualknowl- edge graph embeddings for cross-lingual knowledge alignment, in: Proceedings of the 26th International Joint Conference on Artificial Intelligence (IJCAI), pp. 1511–1517

    work page 2017

  7. [7]

    EasyEA:Large languagemodelisallyouneedinentityalignmentbetweenknowledge graphs,in:FindingsoftheAssociationforComputationalLinguistics: ACL 2025, pp

    Cheng,J.,Lu,C.,Yang,L.,Chen,G.,Zhang,F.,2025. EasyEA:Large languagemodelisallyouneedinentityalignmentbetweenknowledge graphs,in:FindingsoftheAssociationforComputationalLinguistics: ACL 2025, pp. 20981–20995

    work page 2025

  8. [8]

    Word translation without parallel data, in: International Conference on Learning Representations (ICLR)

    Conneau,A.,Lample,G.,Ranzato,M.,Denoyer,L.,Jégou,H.,2018. Word translation without parallel data, in: International Conference on Learning Representations (ICLR)

    work page 2018

  9. [9]

    Adeeprelevancematching modelforad-hocretrieval,in:Proceedingsofthe25thACMInterna- tional Conference on Information and Knowledge Management, pp

    Guo,J.,Fan,Y.,Ai,Q.,Croft,W.B.,2016. Adeeprelevancematching modelforad-hocretrieval,in:Proceedingsofthe25thACMInterna- tional Conference on Information and Knowledge Management, pp. 55–64

    work page 2016

  10. [10]

    Medto: Medical data to ontology matching using hybrid graph neural networks, in: Proceedings of the 27th ACM SIGKDD ConferenceonKnowledgeDiscovery&DataMining,pp.2946–2954

    Hao,J.,Lei,C.,Efthymiou,V.,Quamar,A.,Özcan,F.,Sun,Y.,Wang, W., 2021. Medto: Medical data to ontology matching using hybrid graph neural networks, in: Proceedings of the 27th ACM SIGKDD ConferenceonKnowledgeDiscovery&DataMining,pp.2946–2954

    work page 2021

  11. [11]

    Densepassageretrievalforopen-domainquestion answering, in: Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), pp

    Karpukhin,V.,Oguz,B.,Min,S.,Lewis,P.,Wu,L.,Edunov,S.,Chen, D.,Yih,W.t.,2020. Densepassageretrievalforopen-domainquestion answering, in: Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), pp. 6769–6781

    work page 2020

  12. [12]

    Khattab, O., Zaharia, M., 2020. ColBERT: Efficient and effective passage search via contextualized late interaction over BERT, in: Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval, pp. 39–48

    work page 2020

  13. [13]

    Supervised contrastive learning, in: Advances in Neural Information Processing Systems (NeurIPS)

    Khosla, P., Teterwak, P., Wang, C., Sarna, A., Tian, Y., Isola, P., Maschinot, A., Liu, C., Krishnan, D., 2020. Supervised contrastive learning, in: Advances in Neural Information Processing Systems (NeurIPS)

    work page 2020

  14. [14]

    Adam: A Method for Stochastic Optimization

    Kingma,D.P.,Ba,J.,2014. Adam:Amethodforstochasticoptimiza- tion. arXiv preprint arXiv:1412.6980

    work page internal anchor Pith review Pith/arXiv arXiv 2014

  15. [15]

    Semi-supervised classification with graphconvolutionalnetworks,in:InternationalConferenceonLearn- ing Representations (ICLR)

    Kipf, T.N., Welling, M., 2017. Semi-supervised classification with graphconvolutionalnetworks,in:InternationalConferenceonLearn- ing Representations (ICLR)

    work page 2017

  16. [16]

    Tensordecompositionsandapplica- tions

    Kolda,T.G.,Bader,B.W.,2009. Tensordecompositionsandapplica- tions. SIAM Review 51, 455–500

    work page 2009

  17. [17]

    Understanding context specificity: the effect of contextual factors on clinical reasoning

    Konopasky,A.,Artino,A.R.,Battista,A.,Ohmer,M.,Hemmer,P.A., Torre,D.,Ramani,D.,VanMerrienboer,J.,Teunissen,P.W.,McBee, E., et al., 2020. Understanding context specificity: the effect of contextual factors on clinical reasoning. Diagnosis 7, 257–264

    work page 2020

  18. [18]

    Retrieval-augmented generation for knowledge-intensive NLP tasks

    Lewis, P., Perez, E., Piktus, A., Petroni, F., Karpukhin, V., Goyal, N., Küttler, H., Lewis, M., Yih, W.t., Rocktäschel, T., et al., 2020. Retrieval-augmented generation for knowledge-intensive NLP tasks. Advances in Neural Information Processing Systems 33, 9459–9474

    work page 2020

  19. [19]

    Liu, F., Shareghi, E., Meng, Z., Basaldella, M., Collier, N., 2021. Self-alignment pretraining for biomedical entity representations, in: Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pp. 4228–4238

    work page 2021

  20. [20]

    When Not to Trust Language Models: Investigating Effectiveness of Parametric and Non-Parametric Memories

    Mallen, A., Asai, A., Zhong, V., Das, R., Khashabi, D., Hajishirzi, H., 2022. When not to trust language models: Investigating effec- tiveness of parametric and non-parametric memories. arXiv preprint arXiv:2212.10511

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  21. [21]

    Un- derstandingtraditionalchinesemedicinetherapeutics:anoverviewof the basics and clinical applications

    Matos, L.C., Machado, J.P., Monteiro, F.J., Greten, H.J., 2021. Un- derstandingtraditionalchinesemedicinetherapeutics:anoverviewof the basics and clinical applications. Healthcare 9, 257

    work page 2021

  22. [22]

    Mucheng, R., Heyan, H., Yuxiang, Z., Qianwen, C., Yuan, B., Yang, G., 2022. TCM-SD: A benchmark for probing syndrome differen- tiation via natural language processing, in: Proceedings of the 21st ChineseNationalConferenceonComputationalLinguistics,pp.908– 920

    work page 2022

  23. [23]

    Constructingknowledgegraphs and their biomedical applications

    Nicholson,D.N.,Greene,C.S.,2020. Constructingknowledgegraphs and their biomedical applications. Computational and Structural Biotechnology Journal 18, 1414–1428

    work page 2020

  24. [24]

    Representation Learning with Contrastive Predictive Coding

    Oord, A.v.d., Li, Y., Vinyals, O., 2018. Representation learning with contrastive predictive coding. arXiv preprint arXiv:1807.03748

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  25. [25]

    WHO global report on traditional and complementary medicine 2019

    Organization, W.H., 2019. WHO global report on traditional and complementary medicine 2019. World Health Organization

    work page 2019

  26. [26]

    Differencesintheoriginofphilosophybetweenchinesemedicineand western medicine: exploration of the holistic advantages of chinese medicine

    Sun, D.z., Li, S.d., Liu, Y., Zhang, Y., Mei, R., Yang, M.h., 2013. Differencesintheoriginofphilosophybetweenchinesemedicineand western medicine: exploration of the holistic advantages of chinese medicine. Chinese Journal of Integrative Medicine 19, 706–711

    work page 2013

  27. [27]

    Cross-lingualentityalignmentviajoint attribute-preserving embedding, in: Proceedings of the International Semantic Web Conference, pp

    Sun,Z.,Hu,W.,Li,C.,2017. Cross-lingualentityalignmentviajoint attribute-preserving embedding, in: Proceedings of the International Semantic Web Conference, pp. 628–644

    work page 2017

  28. [28]

    Bootstrapping entity alignment with knowledge graph embedding, in: Proceedings of the 27thInternationalJointConferenceonArtificialIntelligence(IJCAI), pp

    Sun, Z., Hu, W., Zhang, Q., Qu, Y., 2018. Bootstrapping entity alignment with knowledge graph embedding, in: Proceedings of the 27thInternationalJointConferenceonArtificialIntelligence(IJCAI), pp. 4396–4402

    work page 2018

  29. [29]

    An overview of clinical decision support systems: benefits, risks, and strategies for success

    Sutton,R.T.,Pincock,D.,Baumgart,D.C.,Sadowski,D.C.,Fedorak, R.N., Kroeker, K.I., 2020. An overview of clinical decision support systems: benefits, risks, and strategies for success. npj Digital Medicine 3, 17

    work page 2020

  30. [30]

    Multilingual E5 Text Embeddings: A Technical Report

    Wang,L.,Yang,N.,Huang,X.,Yang,L.,Majumder,R.,Wei,F.,2024. Multilingual E5 text embeddings: A technical report. arXiv preprint arXiv:2402.05672

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  31. [31]

    Cross-lingualknowledge graph alignment via graph convolutional networks, in: Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing, pp

    Wang,Z.,Lv,Q.,Lan,X.,Zhang,Y.,2018. Cross-lingualknowledge graph alignment via graph convolutional networks, in: Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing, pp. 349–357

    work page 2018

  32. [32]

    arXiv preprint arXiv:2504.00993 (2025)

    Wu, J., Deng, W., Li, X., Liu, S., Mi, T., Peng, Y., Xu, Z., Liu, Y., Cho, H., Choi, C.I., et al., 2025. Medreason: Eliciting factual medicalreasoningstepsinllmsviaknowledgegraphs. arXivpreprint arXiv:2504.00993

    work page arXiv 2025

  33. [33]

    Relation- aware entity alignment for heterogeneous knowledge graphs, in: Proceedings of the 28th International Joint Conference on Artificial Intelligence (IJCAI), pp

    Wu, Y., Liu, X., Feng, Y., Wang, Z., Zhao, D., 2019a. Relation- aware entity alignment for heterogeneous knowledge graphs, in: Proceedings of the 28th International Joint Conference on Artificial Intelligence (IJCAI), pp. 5278–5284

    work page

  34. [34]

    SymMap:anintegrativedatabaseof traditionalchinesemedicineenhancedbysymptommapping

    Wu,Y.,Zhang,F.,Yang,K.,Fang,S.,Bu,D.,Li,H.,Sun,L.,Hu,H., Gao,K.,Wang,W.,etal.,2019b. SymMap:anintegrativedatabaseof traditionalchinesemedicineenhancedbysymptommapping. Nucleic Acids Research 47, D1110–D1117

    work page

  35. [35]

    On- toEA: Ontology-guided entity alignment via joint knowledge graph embedding, in: Findings of the Association for Computational Lin- guistics: ACL-IJCNLP 2021, pp

    Xiang,Y.,Zhang,Z.,Chen,J.,Chen,X.,Lin,Z.,Zheng,Y.,2021. On- toEA: Ontology-guided entity alignment via joint knowledge graph embedding, in: Findings of the Association for Computational Lin- guistics: ACL-IJCNLP 2021, pp. 1117–1128

    work page 2021

  36. [36]

    A survey on unifying large language models and knowledgegraphsforbiomedicineandhealthcare,in:Proceedingsof the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining, pp

    Xu, R., Jiang, P., Luo, L., Xiao, C., Cross, A., Pan, S., Sun, J., Yang, C., 2025. A survey on unifying large language models and knowledgegraphsforbiomedicineandhealthcare,in:Proceedingsof the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining, pp. 6195–6205

    work page 2025

  37. [37]

    Biomedical information integra- tionviaadaptivelargelanguagemodelconstruction

    Xue, X., Wu, M.E., Khan, F., 2024. Biomedical information integra- tionviaadaptivelargelanguagemodelconstruction. IEEEJournalof Biomedical and Health Informatics 29, 6381–6394

    work page 2024

  38. [38]

    Qwen2.5 Technical Report

    Yang,A.,Yang,B.,Zhang,B.,Hui,B.,Zheng,B.,Yu,B.,Li,C.,Liu, D.,Huang,F.,Wei,H.,etal.,2024a. Qwen2.5technicalreport. arXiv preprint arXiv:2412.15115 doi:10.48550/arXiv.2412.15115

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2412.15115

  39. [39]

    Yang, H.W., Zou, Y., Shi, P., Lu, W., Lin, J., Sun, X., 2019. Aligning cross-lingual entities with multi-aspect information, 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), pp. 4431–4441

    work page 2019

  40. [40]

    Yang, L., Zhou, S., Cheng, J., Zhang, F., Wan, J., Wang, S., Lee, M.,

    work page

  41. [41]

    5890–5901

    DAEA: Enhancing entity alignment in real-world knowledge graphs through multi-source domain adaptation, in: Proceedings of the 31st International Conference on Computational Linguistics, pp. 5890–5901. Y. Jiao et al.:Preprint submitted to ElsevierPage 14 of 15 Query-Conditioned Knowledge Alignment

    work page

  42. [42]

    Yang, S., Zhao, H., Zhu, S., Zhou, G., Xu, H., Jia, Y., Zan, H., 2024b. Zhongjing: Enhancing the chinese medical capabilities of large language model through expert feedback and real-world multi- turn dialogue, in: Proceedings of the AAAI conference on artificial intelligence, pp. 19368–19376

    work page

  43. [43]

    Developing TCM clinical practice guide- lines:acomparisonbetweentraditionalchinesemedicineandwestern medicine

    Yu, X., Wu, S., Zhang, J., Hu, Y., Luo, M., Zhao, H., Song, X., Chen, Y., Wang, X., 2023. Developing TCM clinical practice guide- lines:acomparisonbetweentraditionalchinesemedicineandwestern medicine. Integrative Medicine Research 12, 100952

    work page 2023

  44. [44]

    Multi-view knowledge graph embedding for entity alignment, in: Proceedings of the 28th International Joint Conference on Artificial Intelligence (IJCAI), pp

    Zhang, Q., Sun, Z., Hu, W., Chen, M., Li, C., 2019. Multi-view knowledge graph embedding for entity alignment, in: Proceedings of the 28th International Joint Conference on Artificial Intelligence (IJCAI), pp. 5145–5151

    work page 2019

  45. [45]

    AutoAlign: Fully automatic and effective knowledge graph alignment enabled by large language models

    Zhang,R.,Su,Y.,Trisedya,B.D.,Zhao,X.,Yang,M.,Cheng,H.,Qi, J., 2023. AutoAlign: Fully automatic and effective knowledge graph alignment enabled by large language models. IEEE Transactions on Knowledge and Data Engineering 36, 2357–2371

    work page 2023

  46. [46]

    A survey: knowledge graph entity alignment research based on graph embed- ding

    Zhu, B., Wang, R., Wang, J., Shao, F., Wang, K., 2024. A survey: knowledge graph entity alignment research based on graph embed- ding. Artificial Intelligence Review 57, 229. Y. Jiao et al.:Preprint submitted to ElsevierPage 15 of 15

    work page 2024