Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in Coakley et alK. L. Coakley, T. Snelleman, H. Hoos, and O. E. Gundersen, "The embrace of open science: An analysis of a decade of AI research and 56 800 conference papers," Under Review, 2026..

Addressing Mark Imbalance in Integration-free Marked Temporal Point Processes

Authors: Sishun Liu, KE DENG, Yongli Ren, Yan Wang, Xiuzhen Zhang

NeurIPS 2025 | Venue PDF | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments on real-world datasets demonstrate the superior performance of our solution against various baselines for the next event mark and time prediction.
Researcher Affiliation Academia Sishun Liu RMIT University Melbourne, Victoria 3000 EMAIL Ke Deng RMIT University Melbourne, Victoria 3000 EMAIL Yongli Ren RMIT University Melbourne, Victoria 3000 EMAIL Yan Wang Macquarie University Syndey, New South Wales 2000 EMAIL Xiuzhen Zhang RMIT University Melbourne, Victoria 3000 EMAIL
Pseudocode Yes The pseudo code is below:
Open Source Code Yes The code is available at https://github.com/undes1red/IFNMTPP.
Open Datasets Yes Retweet, Stack Overflow, Taobao, and USearthquake are released under Apache-2.0 license[38]. Four real-world datasets include Retweet [42], Stack Overflow(SO) [20], Taobao User Behavior Data(Taobao) [2], and earthquake events over the Conterminous US(USearthquake) [38].
Dataset Splits No The paper refers to training and test sets but does not provide specific split percentages, absolute sample counts, or a detailed methodology for partitioning the datasets into training, validation, and test sets. It mentions splitting marks into frequent and rare subsets, but not the dataset itself for training/testing.
Hardware Specification Yes We conduct all experiments on an internal cluster. It includes Intel Xeon CPUs and NVIDIA A100-PCIE GPUs.
Software Dependencies No We rewrote Fully NN in Py Torch[30] based on the official implementation available at https://github.com/omitakahiro/ Neural Network Point Process, which is publicly accessible without any license. (The reference [30] states "Py Torch" but does not give a specific version number).
Experiment Setup Yes This section introduces the hyperparameter settings for all MTPP models used in this paper. The two values of Steps refer to the number of warm-up steps and total training steps, respectively. BS refers to batch size, and LR refers to the learning rate. Unless otherwise specified, we repeatedly train a model 3 times with different random seeds and report the mean and standard deviation of the results. We conduct all experiments on an internal cluster. It includes Intel Xeon CPUs and NVIDIA A100-PCIE GPUs. All codes will be release upon acceptance under the MIT license. Table 10 lists the hyperparameter settings for IFNMTPP.