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..
Device-Cloud Collaborative Correction for On-Device Recommendation
Authors: Tianyu Zhan, Shengyu Zhang, Zheqi Lv, Jieming Zhu, Jiwei Li, Fan Wu, Fei Wu
IJCAI 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments on multiple datasets show that Co Corr Rec outperforms existing Transformer-based and RNN-based device recommendation models in terms of performance, with fewer parameters and lower FLOPs, thereby achieving a balance between real-time performance and high efficiency. Code is available at https: //github.com/Yuzt-zju/Co Corr Rec. In this section, we conduct extensive experiments on three realworld datasets to answer the following research questions: RQ1: What is the resource consumption required for various basic recommendation models (FLOPs and maximum memory usage)? RQ2: What is the Time delay caused by device-cloud communication of the Co Corr Rec? RQ3: How does our Co Corr Rec perform compared to the baselines under various experimental settings and how do the designs of it affect the performance? |
| Researcher Affiliation | Collaboration | 1 Zhejiang University 2 Huawei Noah s Ark Lab 3 Shanghai Jiao Tong University EMAIL, EMAIL, EMAIL, EMAIL |
| Pseudocode | Yes | Algorithm 1 TTT Block of SCN Input: H Output: Wb, Z Process: Wb, Z XQ, XK, XV θQH, θKH, θV H XQ, XK Ro PE(XQ, XV ), Ro PE(XK, XV ) XKl W0XK XV Zl σ ( XKl) σ is the derivative of sigmod function W0l Zl(XK)T Wb W0 W0l b b0 mask( Zl) attn mask(XT KXQ) Z W0XQ Zl attn + b return Wb, Z |
| Open Source Code | Yes | Code is available at https: //github.com/Yuzt-zju/Co Corr Rec. |
| Open Datasets | Yes | We conduct experiments on three publicly available datasets in different scenarios. Amazon-beauty and Amazonelectronic are from an e-commerce platform Amazon1, which https://jmcauley.ucsd.edu/data/amazon/ covers various product categories such as books, electronics, home goods, and more. We also evaluated on Yelp2 which is a representative business dataset containing user reviews for different restaurants. 2https://www.yelp.com/dataset/ |
| Dataset Splits | No | For all datasets, we first sort all in-interactions chronologically according to the timestamps. And then discard users and items with interactions 10. All user-item pairs in the dataset are treated as positive samples. In the training and test sets, the user-item pairs that do not exist in the dataset are sampled at 1:4 and 1:100, respectively, as negative samples. |
| Hardware Specification | Yes | All models are trained using a single RTX-4090 GPU. |
| Software Dependencies | No | Optimizer Adam W |
| Experiment Setup | Yes | We tune the learning rates of all models in {1e 3, 2e 3, 5e 3, 1e 2} and select the best performance. As to model-specific hyper-parameters, we select default parameters from open-source code. Additionally, unless otherwise specified, the length of the user sequence is 10. All models are trained using a single RTX-4090 GPU. Other hyperparameters of training are shown in table ?? and the number of head is for transformer-based models. Dataset Model Hyperparameter Setting Beauty Electronic Yelp DIN GRU4Rec SASRec BERT4Rec Mamba4Rec Co Corr Rec GPU RTX-4090 GPU(24G) Optimizer Adam W Learning rate {1e 3, 2e 3, 5e 3, 1e 2} Batch size 1024 Sequence length 10 Dimension of Embedding 1 64 Number of Head 4 Table 3: Hyperparameters of training. |