arxiv: 2605.13153 · v1 · submitted 2026-05-13 · 💻 cs.AI

Recognition: no theorem link

Strikingness-Aware Evaluation for Temporal Knowledge Graph Reasoning

Rikui Huang, Shengzhe Zhang, Wei Wei

Authors on Pith no claims yet

Pith reviewed 2026-05-14 20:22 UTC · model grok-4.3

classification 💻 cs.AI

keywords temporal knowledge graphsreasoningevaluation metricsstrikingnessrule-based frameworkMRRHits@kensemble evaluation

0 comments

The pith

A strikingness measure shows that temporal knowledge graph models degrade on rare outstanding events and that ensemble gains often come from fitting trivial repetitions instead.

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

The paper argues that standard evaluation in temporal knowledge graph reasoning treats every event the same, which inflates apparent progress because most events are just repetitive and easy to guess from history. It introduces a rule-based strikingness measuring framework that scores how unusual an event is by contrasting its expected frequency against similar events generated from temporal rules mined from the same data. When strikingness is used to reweight metrics such as MRR and Hits@k, every tested model performs progressively worse as strikingness rises. Path-based models do relatively better on low-strikingness events while representation-based models hold up better on high-strikingness ones. An ensemble method that appears strong under uniform evaluation turns out to derive its advantage mainly from correctly predicting the trivial cases rather than from improved reasoning on the hard ones.

Core claim

By weighting evaluation metrics according to event strikingness, the framework shows that representative temporal knowledge graph models all lose accuracy as events become more striking, that path-based and representation-based approaches have complementary strengths across strikingness levels, and that ensemble improvements largely reflect better handling of repetitive events rather than genuine advances in reasoning about unusual future events.

What carries the argument

The rule-based strikingness measuring framework (RSMF) that assigns a strikingness score to each event by comparing its observed occurrence against the expected frequency of peer events generated from temporal rules extracted from the training data.

If this is right

All tested models lose accuracy on high-strikingness events, so uniform metrics have been overstating reasoning ability.
Path-based methods are relatively stronger on low-strikingness repetitive events while representation-based methods are relatively stronger on high-strikingness ones.
Ensemble performance gains under standard metrics largely disappear once trivial events are down-weighted, indicating the gains come from fitting common patterns.
Future model comparisons should report both uniform and strikingness-weighted scores to separate frequency fitting from reasoning improvement.

Where Pith is reading between the lines

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

Models could be trained with an auxiliary loss that explicitly rewards accurate prediction of high-strikingness events.
The same strikingness idea could be applied to other temporal prediction tasks such as user-behavior forecasting or scientific event prediction.
Re-evaluating previously published TKGR results with strikingness weighting might reorder the current leaderboard.
The framework suggests that progress in the field has been measured more by memorizing frequent patterns than by learning to anticipate unusual ones.

Load-bearing premise

The rule-based method for calculating strikingness correctly identifies events whose correct prediction requires deeper reasoning rather than simple frequency matching.

What would settle it

Re-running the four benchmark experiments after replacing the rule-derived strikingness scores with random weights and checking whether the reported performance gaps between models and between strikingness levels disappear.

Figures

Figures reproduced from arXiv: 2605.13153 by Rikui Huang, Shengzhe Zhang, Wei Wei.

**Figure 2.** Figure 2: Group performances on ICEWS14 and ICEWS18. In each [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗

**Figure 3.** Figure 3: WMRR with different bias b on ICEWS18. ported by sufficient historical evidence remain learnable. The findings validate that our strikingness measure aligns not only with event rarity but also with predictive difficulty rooted in evidence scarcity. The definition of NOf and more analysis are provided in the Appendix E. 4.3 Strikingness-Aware Evaluation for TKGR To mitigate the low-strikingness bias in the … view at source ↗

**Figure 4.** Figure 4: Comparing with other strikingness baselines. [PITH_FULL_IMAGE:figures/full_fig_p012_4.png] view at source ↗

**Figure 5.** Figure 5: Group performances of different models and data volume [PITH_FULL_IMAGE:figures/full_fig_p012_5.png] view at source ↗

**Figure 6.** Figure 6: Relation between strikingness and novelty. [PITH_FULL_IMAGE:figures/full_fig_p014_6.png] view at source ↗

**Figure 7.** Figure 7: Strikingness and count of events with different relations [PITH_FULL_IMAGE:figures/full_fig_p014_7.png] view at source ↗

**Figure 8.** Figure 8: WMRR with different bias b on ICEWS14, ICEWS05-15, and GDELT. 0 3500 = 0.2 0 3500 = 0.15 0 3500 = 0.1 0 3500 = 0.05 0.0 0.2 0.4 0.6 0.8 1.0 0 3500 = 0.0 Strikingness Count 0 3500 =0.25 0 3500 =0.2 0 3500 =0.15 0 3500 =0.1 0.0 0.2 0.4 0.6 0.8 1.0 0 3500 =0.05 Strikingness Count 0 3500 s=0.25 0 3500 s=0.3 0 3500 s=0.35 0 3500 s=0.4 0.0 0.2 0.4 0.6 0.8 1.0 0 3500 s=0.45 Strikingness Count 0 3500 w=50 0 3500 w… view at source ↗

**Figure 9.** Figure 9: Distribution on ICEWS14 with different hyperparameters. [PITH_FULL_IMAGE:figures/full_fig_p015_9.png] view at source ↗

**Figure 11.** Figure 11: Ensemble results of different baseline models. The lower triangle (including the diagonal) represents the original metrics and the [PITH_FULL_IMAGE:figures/full_fig_p016_11.png] view at source ↗

read the original abstract

Temporal Knowledge Graph Reasoning (TKGR) aims at inferring missing (especially future) events from historical data. Current evaluation in TKGR uniformly weights all events, ignoring that most are trivial repetitions, which overestimate the true reasoning ability. Therefore, the rare outstanding events, whose prediction demands deeper reasoning, should be distinguished and emphasized. To this end, we propose a strikingness-aware evaluation framework, which introduces a rule-based strikingness measuring framework (RSMF) to quantify event strikingness by comparing its expected occurrence with peer events derived from temporal rules. Strikingness is then integrated as a weighting factor into metrics like weighted MRR and Hits@k. Experiments on four TKG benchmarks reveal: 1) All representative models perform worse as event strikingness increases, 2) Path-based methods excel on low-strikingness events and representation-based ones on high-strikingness events, 3) We design an ensemble method whose gains stem from fitting trivial events rather than reasoning improvement. Our framework provides a more rigorous evaluation, refocusing the field on predicting outstanding events.

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.

Desk Editor's Note private letter to a colleague

The paper adds a rule-based strikingness weight to TKGR metrics and shows models degrade on rare events, but rule extraction details and possible leakage need checking before the claims land.

read the letter

The main thing here is a new weighting scheme for TKGR evaluation that down-weights trivial repeating events and up-weights rare ones using temporal rules mined from the data. They call the measure RSMF and fold it into weighted MRR and Hits@k. On four standard benchmarks the results show every model gets worse as strikingness rises, path-based methods handle the common stuff better while embedding methods do better on the outliers, and an ensemble they built mostly wins by nailing the easy cases rather than improving reasoning overall. That last point is useful because it pushes back on the usual claim that ensembles are automatically better at the hard part of the task. The experiments are run across multiple datasets and the trends line up, which is the strongest part of the work. The soft spot is the rule extraction step. If the temporal rules are mined from the full graph including test events, then the strikingness scores for held-out cases are contaminated by patterns that should not be available at test time. The abstract does not spell out whether mining is restricted to the training split, and without that the drop in performance on high-strikingness events could partly reflect leakage rather than genuine reasoning difficulty. They also skip statistical significance tests and exact formulas for the weighting, so the three numbered findings rest on moderate evidence. This is aimed at people already working on temporal knowledge graphs who care about evaluation realism. A reader who wants to know whether current models are actually solving the forecasting problem rather than memorizing repeats will get something concrete from it. The central idea is internally consistent and addresses a real gap, so it deserves a serious referee even if the leakage question forces revisions.

Referee Report

3 major / 2 minor

Summary. The manuscript proposes a strikingness-aware evaluation framework for temporal knowledge graph reasoning (TKGR). It introduces the Rule-based Strikingness Measuring Framework (RSMF) to quantify the strikingness of events by comparing their expected occurrence against peer events derived from temporal rules extracted from the data. Strikingness scores are then used as weights in standard metrics such as MRR and Hits@k. Through experiments on four TKG benchmarks, the authors report three main findings: (1) all models perform worse as strikingness increases, (2) path-based methods excel on low-strikingness events while representation-based methods perform better on high-strikingness ones, and (3) gains from an ensemble method arise from better fitting of trivial (low-strikingness) events rather than improved reasoning.

Significance. If the strikingness measure is valid and free from data leakage, this framework could provide a more rigorous way to evaluate TKGR models by emphasizing events that require deeper reasoning rather than trivial repetitions. This has the potential to shift the field towards developing methods that handle outstanding events better, addressing a possible overestimation in current uniform-weight evaluations. The experimental trends across benchmarks are a positive step toward falsifiable claims about model behavior.

major comments (3)

[Section 3] Section 3 (RSMF description): the rule extraction process is not specified as using only the training split; if temporal rules are mined from the full TKG (train+val+test), then strikingness scores for held-out test events incorporate patterns from those events, creating leakage that directly contaminates the weighted MRR/Hits@k and undermines the interpretation of performance drops on high-strikingness events as reflecting genuine reasoning difficulty.
[Section 4] Section 4 (Experiments and findings): the three numbered claims rest on observed trends, but no statistical significance tests, confidence intervals, or exact weighting formulas (e.g., how expected occurrence is normalized against peers) are reported; without these, the support for 'all models perform worse as strikingness increases' and the path- vs. representation-based distinction remains moderate.
[Section 4.3] Section 4.3 (ensemble analysis): the claim that ensemble gains stem from fitting trivial events rather than reasoning improvement is load-bearing for finding (3), yet lacks ablation results on high-strikingness subsets or comparison of per-strikingness performance before/after ensembling.

minor comments (2)

[Abstract] The abstract and introduction should explicitly name the four TKG benchmarks used.
[Section 3] The precise mathematical definition of the strikingness score (expected occurrence vs. peer events) should be given as a numbered equation for reference.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive and detailed comments. We address each major point below and will incorporate the suggested clarifications and additions into the revised manuscript.

read point-by-point responses

Referee: [Section 3] Section 3 (RSMF description): the rule extraction process is not specified as using only the training split; if temporal rules are mined from the full TKG (train+val+test), then strikingness scores for held-out test events incorporate patterns from those events, creating leakage that directly contaminates the weighted MRR/Hits@k and undermines the interpretation of performance drops on high-strikingness events as reflecting genuine reasoning difficulty.

Authors: We appreciate this important observation. In the RSMF implementation, temporal rules are extracted exclusively from the training split to avoid any leakage from validation or test data. We will revise Section 3 to explicitly state this restriction and provide the precise rule-mining procedure using only training data, thereby ensuring the strikingness scores for test events remain uncontaminated. revision: yes
Referee: [Section 4] Section 4 (Experiments and findings): the three numbered claims rest on observed trends, but no statistical significance tests, confidence intervals, or exact weighting formulas (e.g., how expected occurrence is normalized against peers) are reported; without these, the support for 'all models perform worse as strikingness increases' and the path- vs. representation-based distinction remains moderate.

Authors: We agree that greater statistical rigor is needed. In the revision we will add paired statistical significance tests (e.g., Wilcoxon signed-rank) for performance trends across strikingness bins, report 95% confidence intervals on all weighted metrics, and include the exact formulas for expected occurrence and peer normalization in an expanded Section 3. These additions will provide firmer quantitative support for the reported trends. revision: yes
Referee: [Section 4.3] Section 4.3 (ensemble analysis): the claim that ensemble gains stem from fitting trivial events rather than reasoning improvement is load-bearing for finding (3), yet lacks ablation results on high-strikingness subsets or comparison of per-strikingness performance before/after ensembling.

Authors: This is a fair critique of the current evidence. We will augment Section 4.3 with new ablation tables that isolate ensemble performance on high-strikingness subsets and directly compare per-strikingness MRR/Hits@k values before versus after ensembling. These results will substantiate that the gains derive predominantly from low-strikingness events. revision: yes

Circularity Check

0 steps flagged

No significant circularity detected in the strikingness-aware evaluation derivation

full rationale

The paper defines a rule-based strikingness measuring framework (RSMF) that computes event strikingness via comparison of expected occurrence against peer events from temporal rules, then applies these scores as weights to standard metrics such as weighted MRR and Hits@k. No equations, definitions, or self-citations in the abstract or described framework reduce the final weighted metrics to fitted parameters or inputs by construction; strikingness functions as an independent, externally derived weighting factor rather than a self-referential or tautological quantity. The reported experimental observations (performance degradation with increasing strikingness, differential method strengths, and ensemble analysis) remain empirical claims that do not collapse into the measurement process itself. The derivation chain is therefore self-contained against external benchmarks and does not exhibit any of the enumerated circularity patterns.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

The framework rests on the domain assumption that temporal rules extracted from historical data can reliably define expected peer events for strikingness calculation; no free parameters or invented entities are explicitly introduced in the abstract.

axioms (1)

domain assumption Temporal rules extracted from the data can define expected occurrences for peer events used to compute strikingness.
Invoked to quantify how much an event deviates from what the rules predict.

invented entities (1)

Strikingness score no independent evidence
purpose: To quantify how outstanding an event is relative to rule-derived peers
New scalar introduced to reweight standard metrics; no external falsifiable handle is described in the abstract.

pith-pipeline@v0.9.0 · 5484 in / 1359 out tokens · 59849 ms · 2026-05-14T20:22:24.189556+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

73 extracted references · 49 canonical work pages

[1]

ACM Computing Surveys , volume =

Liang Zhao , title =. ACM Computing Surveys , volume =. 2022 , url =. doi:10.1145/3450287 , timestamp =

work page doi:10.1145/3450287 2022
[2]

International studies review , volume=

Crisis early warning and decision support: Contemporary approaches and thoughts on future research , author=. International studies review , volume=. 2010 , publisher=

2010
[3]

Coutinho and Sagi Eppel and Jacob Gates Foster and Andrew Gritsevskiy and Harlin Lee and Yichao Lu and Jo

Mario Krenn and Lorenzo Buffoni and Bruno C. Coutinho and Sagi Eppel and Jacob Gates Foster and Andrew Gritsevskiy and Harlin Lee and Yichao Lu and Jo. Forecasting the future of artificial intelligence with machine learning-based link prediction in an exponentially growing knowledge network , journal =. 2023 , url =. doi:10.1038/S42256-023-00735-0 , timestamp =

work page doi:10.1038/s42256-023-00735-0 2023
[4]

ISA annual convention , volume=

Gdelt: Global data on events, location, and tone, 1979--2012 , author=. ISA annual convention , volume=. 2013 , organization=

1979
[5]

IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI) , volume =

Ke Liang and Lingyuan Meng and Meng Liu and Yue Liu and Wenxuan Tu and Siwei Wang and Sihang Zhou and Xinwang Liu and Fuchun Sun and Kunlun He , title =. IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI) , volume =. 2024 , url =. doi:10.1109/TPAMI.2024.3417451 , timestamp =

work page doi:10.1109/tpami.2024.3417451 2024
[6]

2024 , booktitle=

Impact4Cast: Forecasting high-impact research topics via machine learning on evolving knowledge graphs , author=. 2024 , booktitle=

2024
[7]

Proceedings of the 2018 International Conference on Management of Data (SIGMOD) , pages =

Gensheng Zhang and Damian Jimenez and Chengkai Li , title =. Proceedings of the 2018 International Conference on Management of Data (SIGMOD) , pages =. 2018 , url =. doi:10.1145/3183713.3183730 , timestamp =

work page doi:10.1145/3183713.3183730 2018
[8]

Yueji Yang and Yuchen Li and Panagiotis Karras and Anthony K. H. Tung , title =. The 27th. 2021 , url =. doi:10.1145/3447548.3467272 , timestamp =

work page doi:10.1145/3447548.3467272 2021
[9]

Proceedings of the 30th

Hanhua Xiao and Yuchen Li and Yanhao Wang and Panagiotis Karras and Kyriakos Mouratidis and Natalia Rozalia Avlona , title =. Proceedings of the 30th. 2024 , url =. doi:10.1145/3637528.3671763 , timestamp =

work page doi:10.1145/3637528.3671763 2024
[10]

2010 , publisher=

Black swan , author=. 2010 , publisher=

2010
[11]

Safety science , volume=

On the meaning of a black swan in a risk context , author=. Safety science , volume=. 2013 , publisher=

2013
[12]

Journal of Business Continuity & Emergency Planning , volume=

Planning for grey rhino risks: How to prepare for the ‘unforeseeable’ , author=. Journal of Business Continuity & Emergency Planning , volume=. 2024 , publisher=

2024
[13]

Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP) , year =

Woojeong Jin and Meng Qu and Xisen Jin and Xiang Ren , title =. Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP) , year =. doi:10.18653/V1/2020.EMNLP-MAIN.541 , timestamp =

work page doi:10.18653/v1/2020.emnlp-main.541 2020
[14]

The 44th International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR) , year =

Zixuan Li and Xiaolong Jin and Wei Li and Saiping Guan and Jiafeng Guo and Huawei Shen and Yuanzhuo Wang and Xueqi Cheng , title =. The 44th International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR) , year =. doi:10.1145/3404835.3462963 , timestamp =

work page doi:10.1145/3404835.3462963
[15]

Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (ACL) , year =

Rikui Huang and Wei Wei and Xiaoye Qu and Shengzhe Zhang and Dangyang Chen and Yu Cheng , title =. Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (ACL) , year =. doi:10.18653/V1/2024.ACL-LONG.580 , timestamp =

work page doi:10.18653/v1/2024.acl-long.580 2024
[16]

Findings of the Association for Computational Linguistics (NAACL) , year =

Ruotong Liao and Xu Jia and Yangzhe Li and Yunpu Ma and Volker Tresp , title =. Findings of the Association for Computational Linguistics (NAACL) , year =. doi:10.18653/V1/2024.FINDINGS-NAACL.268 , timestamp =

work page doi:10.18653/v1/2024.findings-naacl.268 2024
[17]

Machine Learning and Knowledge Discovery in Databases: Research Track - European Conference (ECML-PKDD) , series =

Julia Gastinger and Timo Sztyler and Lokesh Sharma and Anett Schuelke and Heiner Stuckenschmidt , title =. Machine Learning and Knowledge Discovery in Databases: Research Track - European Conference (ECML-PKDD) , series =. 2023 , url =. doi:10.1007/978-3-031-43418-1\_32 , timestamp =

work page doi:10.1007/978-3-031-43418-1 2023
[18]

36th Conference on Artificial Intelligence (AAAI) , year =

Yushan Liu and Yunpu Ma and Marcel Hildebrandt and Mitchell Joblin and Volker Tresp , title =. 36th Conference on Artificial Intelligence (AAAI) , year =. doi:10.1609/AAAI.V36I4.20330 , timestamp =

work page doi:10.1609/aaai.v36i4.20330
[19]

ACM Transactions on Database Systems , volume =

Fabrizio Angiulli and Fabio Fassetti and Luigi Palopoli , title =. ACM Transactions on Database Systems , volume =. 2009 , url =. doi:10.1145/1508857.1508864 , timestamp =

work page doi:10.1145/1508857.1508864 2009
[20]

Temporal Knowledge Graph Forecasting Without Knowledge Using In-Context Learning , booktitle =

Dong. Temporal Knowledge Graph Forecasting Without Knowledge Using In-Context Learning , booktitle =. 2023 , url =. doi:10.18653/V1/2023.EMNLP-MAIN.36 , timestamp =

work page doi:10.18653/v1/2023.emnlp-main.36 2023
[21]

History Repeats Itself:

Julia Gastinger and Christian Meilicke and Federico Errica and Timo Sztyler and Anett Sch. History Repeats Itself:. Proceedings of the 33rd International Joint Conference on Artificial Intelligence (IJCAI) , year =
[22]

Proceedings of the 31st International Joint Conference on Artificial Intelligence (IJCAI) , year =

Yujia Li and Shiliang Sun and Jing Zhao , title =. Proceedings of the 31st International Joint Conference on Artificial Intelligence (IJCAI) , year =. doi:10.24963/IJCAI.2022/299 , timestamp =

work page doi:10.24963/ijcai.2022/299 2022
[23]

Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing (EMNLP) , year =

Haohai Sun and Jialun Zhong and Yunpu Ma and Zhen Han and Kun He , title =. Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing (EMNLP) , year =. doi:10.18653/V1/2021.EMNLP-MAIN.655 , timestamp =

work page doi:10.18653/v1/2021.emnlp-main.655 2021
[24]

Wei Chen and Huaiyu Wan and Yuting Wu and Shuyuan Zhao and Jiayaqi Cheng and Yuxin Li and Youfang Lin , title =. 40th. 2024 , url =. doi:10.1109/ICDE60146.2024.00062 , timestamp =

work page doi:10.1109/icde60146.2024.00062 2024
[25]

Translating Embeddings for Modeling Multi-relational Data , booktitle =

Antoine Bordes and Nicolas Usunier and Alberto Garc. Translating Embeddings for Modeling Multi-relational Data , booktitle =. 2013 , url =

2013
[26]

2023 , url =

Saiping Guan and Xueqi Cheng and Long Bai and Fujun Zhang and Zixuan Li and Yutao Zeng and Xiaolong Jin and Jiafeng Guo , title =. 2023 , url =. doi:10.1109/TKDE.2022.3180362 , timestamp =

work page doi:10.1109/tkde.2022.3180362 2023
[27]

2022 , url =

Guansong Pang and Chunhua Shen and Longbing Cao and Anton van den Hengel , title =. 2022 , url =. doi:10.1145/3439950 , timestamp =

work page doi:10.1145/3439950 2022
[28]

2021 , url =

Ali Bou Nassif and Manar Abu Talib and Qassim Nasir and Fatima Mohamad Dakalbab , title =. 2021 , url =. doi:10.1109/ACCESS.2021.3083060 , timestamp =

work page doi:10.1109/access.2021.3083060 2021
[29]

Annals of Data Science , volume=

A comprehensive survey of anomaly detection algorithms , author=. Annals of Data Science , volume=. 2023 , publisher=

2023
[30]

Neurocomputing , volume =

Enchang Zhu and Zhengtao Yu and Yuxin Huang and Shengxiang Gao and Yantuan Xian , title =. Neurocomputing , volume =. 2025 , url =. doi:10.1016/J.NEUCOM.2024.128639 , timestamp =

work page doi:10.1016/j.neucom.2024.128639 2025
[31]

Findings of the Association for Computational Linguistics:

Yincen Qu and Ningyu Zhang and Hui Chen and Zelin Dai and Chengming Wang and Xiaoyu Wang and Qiang Chen and Huajun Chen , title =. Findings of the Association for Computational Linguistics:. 2022 , url =. doi:10.18653/V1/2022.FINDINGS-EMNLP.2 , timestamp =

work page doi:10.18653/v1/2022.findings-emnlp.2 2022
[32]

Human-Centric Cyber Social Computing Model for Hot-Event Detection and Propagation , journal =

Lei. Human-Centric Cyber Social Computing Model for Hot-Event Detection and Propagation , journal =. 2019 , url =. doi:10.1109/TCSS.2019.2913783 , timestamp =

work page doi:10.1109/tcss.2019.2913783 2019
[33]

2022 , url =

Zuowu Zheng and Xiaofeng Gao and Xiao Ma and Guihai Chen , title =. 2022 , url =. doi:10.1109/TKDE.2020.2994946 , timestamp =

work page doi:10.1109/tkde.2020.2994946 2022
[34]

Jiacheng Luo and Wenyi Xu and Xiaofeng Gao and Guihai Chen , title =. Knowl. Inf. Syst. , volume =. 2022 , url =. doi:10.1007/S10115-022-01733-1 , timestamp =

work page doi:10.1007/s10115-022-01733-1 2022
[35]

9th International Conference on Learning Representations (ICLR) , year =

Zhen Han and Peng Chen and Yunpu Ma and Volker Tresp , title =. 9th International Conference on Learning Representations (ICLR) , year =
[36]

Talukdar and Yiming Yang , title =

Zhiqing Sun and Shikhar Vashishth and Soumya Sanyal and Partha P. Talukdar and Yiming Yang , title =. Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics (ACL) , year =. doi:10.18653/V1/2020.ACL-MAIN.489 , timestamp =

work page doi:10.18653/v1/2020.acl-main.489 2020
[37]

2023 , url =

Yuyue Zhao and Xiang Wang and Jiawei Chen and Yashen Wang and Wei Tang and Xiangnan He and Haiyong Xie , title =. 2023 , url =. doi:10.1145/3531267 , timestamp =

work page doi:10.1145/3531267 2023
[38]

Proceedings of the 34th International Conference on Machine Learning,

Rakshit Trivedi and Hanjun Dai and Yichen Wang and Le Song , title =. Proceedings of the 34th International Conference on Machine Learning,. 2017 , url =

2017
[39]

Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (ACL) , pages =

Mengqi Zhang and Yuwei Xia and Qiang Liu and Shu Wu and Liang Wang , title =. Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (ACL) , pages =. 2023 , url =. doi:10.18653/V1/2023.ACL-LONG.705 , timestamp =

work page doi:10.18653/v1/2023.acl-long.705 2023
[40]

Context-aware Event Forecasting via Graph Disentanglement , booktitle =

Yunshan Ma and Chenchen Ye and Zijian Wu and Xiang Wang and Yixin Cao and Tat. Context-aware Event Forecasting via Graph Disentanglement , booktitle =. 2023 , url =. doi:10.1145/3580305.3599285 , timestamp =

work page doi:10.1145/3580305.3599285 2023
[41]

MM-Forecast:

Haoxuan Li and Zhengmao Yang and Yunshan Ma and Yi Bin and Yang Yang and Tat. MM-Forecast:. Proceedings of the 32nd. 2024 , url =. doi:10.1145/3664647.3681593 , timestamp =

work page doi:10.1145/3664647.3681593 2024
[42]

In Proceedings of 38th Conferenceon Neural Information Processing Systems (NeurIPS) , year =

Julia Gastinger and Shenyang Huang and Michael Galkin and Erfan Loghmani and Ali Parviz and Farimah Poursafaei and Jacob Danovitch and Emanuele Rossi and Ioannis Koutis and Heiner Stuckenschmidt and Reihaneh Rabbany and Guillaume Rabusseau , title =. In Proceedings of 38th Conferenceon Neural Information Processing Systems (NeurIPS) , year =
[43]

Proceedings of the 32nd Conference on Artificial Intelligence, (AAAI) , year =

Tim Dettmers and Pasquale Minervini and Pontus Stenetorp and Sebastian Riedel , title =. Proceedings of the 32nd Conference on Artificial Intelligence, (AAAI) , year =. doi:10.1609/AAAI.V32I1.11573 , timestamp =

work page doi:10.1609/aaai.v32i1.11573
[44]

Findings of the Association for Computational Linguistics,

Yuxiang Cai and Qiao Liu and Yanglei Gan and Changlin Li and Xueyi Liu and Run Lin and Da Luo and JiayeYang JiayeYang , title =. Findings of the Association for Computational Linguistics,. 2024 , url =. doi:10.18653/V1/2024.FINDINGS-ACL.343 , timestamp =

work page doi:10.18653/v1/2024.findings-acl.343 2024
[45]

Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers),

Qika Lin and Jun Liu and Rui Mao and Fangzhi Xu and Erik Cambria , title =. Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers),. 2023 , url =. doi:10.18653/V1/2023.ACL-LONG.71 , timestamp =

work page doi:10.18653/v1/2023.acl-long.71 2023
[46]

Zixuan Li and Xiaolong Jin and Saiping Guan and Wei Li and Jiafeng Guo and Yuanzhuo Wang and Xueqi Cheng , title =. Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing, (ACL/IJCNLP) , year =. doi:10.18653/V1/2021.ACL-LONG.365 , timestamp =

work page doi:10.18653/v1/2021.acl-long.365 2021
[47]

Proceedings of the 46th International

Shangfei Zheng and Hongzhi Yin and Tong Chen and Quoc Viet Hung Nguyen and Wei Chen and Lei Zhao , title =. Proceedings of the 46th International. 2023 , url =. doi:10.1145/3539618.3591671 , timestamp =

work page doi:10.1145/3539618.3591671 2023
[48]

Findings of the Association for Computational Linguistics,

Yuwei Xia and Ding Wang and Qiang Liu and Liang Wang and Shu Wu and Xiaoyu Zhang , title =. Findings of the Association for Computational Linguistics,. 2024 , url =. doi:10.18653/V1/2024.FINDINGS-ACL.955 , timestamp =

work page doi:10.18653/v1/2024.findings-acl.955 2024
[49]

8th International Conference on Learning Representations (ICLR) , year =

Daniel Ruffinelli and Samuel Broscheit and Rainer Gemulla , title =. 8th International Conference on Learning Representations (ICLR) , year =
[50]

Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP) , year =

Prachi Jain and Sushant Rathi and Mausam and Soumen Chakrabarti , title =. Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP) , year =. doi:10.18653/V1/2020.EMNLP-MAIN.305 , timestamp =

work page doi:10.18653/v1/2020.emnlp-main.305 2020
[51]

Proceedings of the VLDB Endowment , volume =

Naeemul Hassan and Afroza Sultana and You Wu and Gensheng Zhang and Chengkai Li and Jun Yang and Cong Yu , title =. Proceedings of the VLDB Endowment , volume =. 2014 , url =. doi:10.14778/2733004.2733029 , timestamp =

work page doi:10.14778/2733004.2733029 2014
[52]

Agarwal and Chengkai Li and Jun Yang and Cong Yu , title =

You Wu and Pankaj K. Agarwal and Chengkai Li and Jun Yang and Cong Yu , title =. The 18th. 2012 , url =. doi:10.1145/2339530.2339762 , timestamp =

work page doi:10.1145/2339530.2339762 2012
[53]

RNNLogic: Learning Logic Rules for Reasoning on Knowledge Graphs , booktitle =

Meng Qu and Jun. RNNLogic: Learning Logic Rules for Reasoning on Knowledge Graphs , booktitle =. 2021 , url =

2021
[54]

Findings of the Association for Computational Linguistics (EMNLP) , year =

Zhengtao Liu and Lei Tan and Mengfan Li and Yao Wan and Hai Jin and Xuanhua Shi , title =. Findings of the Association for Computational Linguistics (EMNLP) , year =. doi:10.18653/V1/2023.FINDINGS-EMNLP.249 , timestamp =

work page doi:10.18653/v1/2023.findings-emnlp.249 2023
[55]

Large Language Models-guided Dynamic Adaptation for Temporal Knowledge Graph Reasoning , booktitle =

Jiapu Wang and Kai Sun and Linhao Luo and Wei Wei and Yongli Hu and Alan Wee. Large Language Models-guided Dynamic Adaptation for Temporal Knowledge Graph Reasoning , booktitle =. 2024 , url =

2024
[56]

3rd Conference on Automated Knowledge Base Construction (AKBC) , year =

Christian Meilicke and Patrick Betz and Heiner Stuckenschmidt , title =. 3rd Conference on Automated Knowledge Base Construction (AKBC) , year =. doi:10.24432/C5PK5V , timestamp =

work page doi:10.24432/c5pk5v
[57]

Simon Ott and Patrick Betz and Daria Stepanova and Mohamed H. Gad. Rule-based Knowledge Graph Completion with Canonical Models , booktitle =. 2023 , url =. doi:10.1145/3583780.3615042 , timestamp =

work page doi:10.1145/3583780.3615042 2023
[58]

Learning Sequence Encoders for Temporal Knowledge Graph Completion , booktitle =

Alberto Garc. Learning Sequence Encoders for Temporal Knowledge Graph Completion , booktitle =. 2018 , url =

2018
[59]

Kangzheng Liu and Feng Zhao and Guandong Xu and Xianzhi Wang and Hai Jin , title =. 39th. 2023 , url =. doi:10.1109/ICDE55515.2023.00138 , timestamp =

work page doi:10.1109/icde55515.2023.00138 2023
[60]

2021 , url =

Corentin Kervadec and Grigory Antipov and Moez Baccouche and Christian Wolf , title =. 2021 , url =. doi:10.1109/CVPR46437.2021.00280 , timestamp =

work page doi:10.1109/cvpr46437.2021.00280 2021
[61]

Proceedings of the 3rd Workshop on Continuous Vector Space Models and their Compositionality , pages =

Kristina Toutanova and Danqi Chen , title =. Proceedings of the 3rd Workshop on Continuous Vector Space Models and their Compositionality , pages =. 2015 , url =. doi:10.18653/V1/W15-4007 , timestamp =

work page doi:10.18653/v1/w15-4007 2015
[62]

35th Conference on Artificial Intelligence (AAAI) , year =

Cunchao Zhu and Muhao Chen and Changjun Fan and Guangquan Cheng and Yan Zhang , title =. 35th Conference on Artificial Intelligence (AAAI) , year =. doi:10.1609/AAAI.V35I5.16604 , timestamp =

work page doi:10.1609/aaai.v35i5.16604
[63]

Kangzheng Liu and Feng Zhao and Hai Jin , title =. Int. J. Softw. Informatics , volume =. 2023 , url =. doi:10.21655/IJSI.1673-7288.00308 , timestamp =

work page doi:10.21655/ijsi.1673-7288.00308 2023
[64]

Proceedings of the 32nd International Joint Conference on Artificial Intelligence (IJCAI) , year =

Hao Dong and Zhiyuan Ning and Pengyang Wang and Ziyue Qiao and Pengfei Wang and Yuanchun Zhou and Yanjie Fu , title =. Proceedings of the 32nd International Joint Conference on Artificial Intelligence (IJCAI) , year =. doi:10.24963/IJCAI.2023/232 , timestamp =

work page doi:10.24963/ijcai.2023/232 2023
[65]

37th Conference on Artificial Intelligence (AAAI) , year =

Yi Xu and Junjie Ou and Hui Xu and Luoyi Fu , title =. 37th Conference on Artificial Intelligence (AAAI) , year =. doi:10.1609/AAAI.V37I4.25601 , timestamp =

work page doi:10.1609/aaai.v37i4.25601
[66]

Structure and Interpretation of Computer Programs

Harold Abelson and Gerald Jay Sussman and Julie Sussman. Structure and Interpretation of Computer Programs. 1985

1985
[67]

Visual Information Extraction with Lixto

Robert Baumgartner and Georg Gottlob and Sergio Flesca. Visual Information Extraction with Lixto. Proceedings of the 27th International Conference on Very Large Databases. 2001

2001
[68]

Brachman and James G

Ronald J. Brachman and James G. Schmolze. An overview of the KL-ONE knowledge representation system. Cognitive Science. 1985

1985
[69]

Complexity results for nonmonotonic logics

Georg Gottlob. Complexity results for nonmonotonic logics. Journal of Logic and Computation. 1992

1992
[70]

Hypertree Decompositions and Tractable Queries

Georg Gottlob and Nicola Leone and Francesco Scarcello. Hypertree Decompositions and Tractable Queries. Journal of Computer and System Sciences. 2002

2002
[71]

Levesque

Hector J. Levesque. Foundations of a functional approach to knowledge representation. Artificial Intelligence. 1984

1984
[72]

Levesque

Hector J. Levesque. A logic of implicit and explicit belief. Proceedings of the Fourth National Conference on Artificial Intelligence. 1984

1984
[73]

On the compilability and expressive power of propositional planning formalisms

Bernhard Nebel. On the compilability and expressive power of propositional planning formalisms. Journal of Artificial Intelligence Research. 2000

2000