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..
Semantic Retrieval Augmented Contrastive Learning for Sequential Recommendation
Authors: Ziqiang Cui, Yunpeng Weng, Xing Tang, Xiaokun Zhang, Shiwei Li, Peiyang Liu, Bowei He, Dugang Liu, Weihong Luo, Xiuqiang He, Chen Ma
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments on four public datasets demonstrate the effectiveness and model-agnostic nature of our approach. Our code is available at https://github.com/ziqiangcui/SRA-CL |
| Researcher Affiliation | Collaboration | 1 City University of Hong Kong 2 Huazhong University of Science and Technology 3 Tencent 4 Shenzhen Technology University 5 Peking University 6 Shenzhen University |
| Pseudocode | Yes | Algorithm 1 Training for SRA-CL Require: Training data {Su} for all u U; hyperparameters α, β, k ... Algorithm 2 Inference for SRA-CL Require: Trained model parameters θ; test data {Su} |
| Open Source Code | Yes | Our code is available at https://github.com/ziqiangcui/SRA-CL |
| Open Datasets | Yes | Following previous studies [21, 34, 22], we conducted experiments on four public realworld datasets: Yelp, Amazon Sports, Beauty, and Office. The statistics for these datasets are presented in Table 3. More details about the datasets are shown in Appendix B.1. ... 3https://www.yelp.com/dataset 4http://jmcauley.ucsd.edu/data/amazon/ |
| Dataset Splits | Yes | The leave-one-out strategy is employed, where the last interaction is used for testing, the second-to-last interaction for validation, and the remaining interactions for training. |
| Hardware Specification | Yes | All experiments are conducted with a single 32G V100 GPU. |
| Software Dependencies | Yes | For the LLM, we select Deep Seek-V3, a robust large language model that demonstrates exceptional performance on both standard benchmarks and openended generation evaluations. ... For the text embedding model M, we use the pre-trained Sim CSE-Ro BERTa7 from Hugging Face. |
| Experiment Setup | Yes | The embedding size is set to 64. We adopt the batch size of 256 and employ the Adam optimizer with a learning rate of 0.001. The dropout rate is set to 0.5 across all datasets. Following previous studies [35], we set the maximum sequence length to 20. The early stopping is applied if the metrics on the validation set do not improve over 10 consecutive epochs. |