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..

Path-Enhanced Contrastive Learning for Recommendation

Authors: Haoran Sun, Fei Xiong, Yuanzhe Hu, Liang Wang

NeurIPS 2025 | Venue PDF | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments on three real-world datasets demonstrate the effectiveness of our model.
Researcher Affiliation Academia Haoran Sun Beijing Jiaotong University EMAIL Fei Xiong Beijing Jiaotong University EMAIL Yuanzhe Hu Institute of Software Chinese Academy of Sciences EMAIL Liang Wang Northwestern Polytechnical University EMAIL
Pseudocode Yes Algorithm 1: PECL Framework Training Algorithm
Open Source Code Yes Answer: [Yes] Justification: Yes, all experimental settings, codes, pesudo code and datasets are provided in paper and github.
Open Datasets Yes Our experiments utilize three publicly available, real-world datasets ML-1M, Ciao and Amazon that offer rich information, including both user-item ratings and the corresponding timestamps of user interactions.
Dataset Splits Yes Each dataset is divided into training and test sets by randomly selecting 80% of the rating entries for training purposes, while the leftover 20% is reserved to evaluate the model s performance during testing.
Hardware Specification Yes All experiments we conducted are performed with Intel(R) Xeon(R) Bronze 3204 CPU @ 1.90GHz, Tesla A100, and 256GB memory, running the Ubuntu 20.04.4 LTS operating system.
Software Dependencies No The paper mentions the operating system (Ubuntu 20.04.4 LTS), but does not specify programming languages or library versions like Python, PyTorch, or CUDA, which are common for such research. Therefore, it does not provide specific ancillary software details with version numbers.
Experiment Setup Yes To assess the effectiveness of our recommendation approach, we rely on two widely adopted evaluation metrics: Recall@K and NDCG@K. In our experiment, K is set to 10 and 20. ... To comprehensively evaluate the impact of hyperparameters on PECL performance, we performed a series of controlled experiments by varying key hyperparameters. Specifically, we focus on three crucial hyperparameters: α (Eq.6), β (Eq.9), and τ (Eqs.7, 14).