Enhancing Job Recommendation through LLM-Based Generative Adversarial Networks

Authors: Yingpeng Du, Di Luo, Rui Yan, Xiaopei Wang, Hongzhi Liu, Hengshu Zhu, Yang Song, Jie Zhang

AAAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments on three large real-world recruitment datasets demonstrate the effectiveness of our proposed method.
Researcher Affiliation Collaboration 1School of Computer Science and Engineering, Nanyang Technological University, Singapore 2Gaoling School of Artificial Intelligence, Renmin University of China, Beijing, China 3School of Languages and Communication Studies, Beijing Jiaotong University, Beijing, China 4School of Software and Microelectronics, Peking University, Beijing, China 5Career Science Lab, BOSS Zhipin, Beijing, China 6NLP Center, BOSS Zhipin, Beijing, China
Pseudocode No The paper includes mathematical formulas and an architecture diagram (Figure 2) but does not present any pseudocode or algorithm blocks.
Open Source Code No The paper does not provide a specific repository link or an explicit statement about releasing the source code for the methodology.
Open Datasets No We evaluated the proposed method on three real-world data sets, which were provided by a popular online recruiting platform.
Dataset Splits Yes We spitted the interaction records into training, validation, and test sets equally.
Hardware Specification No The paper does not mention any specific GPU models, CPU models, or other hardware specifications used for running the experiments.
Software Dependencies No We adopted the Chat GLM-6B (Du et al. 2022) as the LLM model in this paper. For a fair comparison, all methods were optimized by the Adam W optimizer...
Experiment Setup Yes all methods were optimized by the Adam W optimizer with the same latent space dimension (i.e., 64), batch size (i.e., 1024), learning rate (i.e., 5 10 5), and regularization coefficient (i.e., 1 10 4). We set d = 768, de = 128, de = 64, and dc = ds = dg = 256 for the proposed method.