Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Generative Job Recommendations with Large Language Model

Zhi Zheng, Zhaopeng Qiu, Xiao Hu, Likang Wu, Hengshu Zhu, Hui Xiong

TL;DR

This paper introduces GIRL, a generative, LLM-based framework for job recommendation that produces personalized job descriptions from CVs to improve explainability and broaden beyond retrieval. It trains the JD generator via a three-stage pipeline: supervised fine-tuning on CV-JD pairs, a reward model trained on recruiter feedback, and PPO-based reinforcement learning to align outputs with market needs. Additionally, GIRL demonstrates how generated JDs can enhance conventional discriminative recommender systems, boosting performance in both typical and cold-start scenarios. Extensive experiments on a real-world recruitment dataset show that generation quality improves with RL fine-tuning and that generation-enhanced recommendations yield tangible gains over strong baselines. The work suggests a pathway toward more holistic, advisory, and explainable career AI tools.

Abstract

The rapid development of online recruitment services has encouraged the utilization of recommender systems to streamline the job seeking process. Predominantly, current job recommendations deploy either collaborative filtering or person-job matching strategies. However, these models tend to operate as "black-box" systems and lack the capacity to offer explainable guidance to job seekers. Moreover, conventional matching-based recommendation methods are limited to retrieving and ranking existing jobs in the database, restricting their potential as comprehensive career AI advisors. To this end, here we present GIRL (GeneratIve job Recommendation based on Large language models), a novel approach inspired by recent advancements in the field of Large Language Models (LLMs). We initially employ a Supervised Fine-Tuning (SFT) strategy to instruct the LLM-based generator in crafting suitable Job Descriptions (JDs) based on the Curriculum Vitae (CV) of a job seeker. Moreover, we propose to train a model which can evaluate the matching degree between CVs and JDs as a reward model, and we use Proximal Policy Optimization (PPO)-based Reinforcement Learning (RL) method to further fine-tine the generator. This aligns the generator with recruiter feedback, tailoring the output to better meet employer preferences. In particular, GIRL serves as a job seeker-centric generative model, providing job suggestions without the need of a candidate set. This capability also enhances the performance of existing job recommendation models by supplementing job seeking features with generated content. With extensive experiments on a large-scale real-world dataset, we demonstrate the substantial effectiveness of our approach. We believe that GIRL introduces a paradigm-shifting approach to job recommendation systems, fostering a more personalized and comprehensive job-seeking experience.

Generative Job Recommendations with Large Language Model

TL;DR

This paper introduces GIRL, a generative, LLM-based framework for job recommendation that produces personalized job descriptions from CVs to improve explainability and broaden beyond retrieval. It trains the JD generator via a three-stage pipeline: supervised fine-tuning on CV-JD pairs, a reward model trained on recruiter feedback, and PPO-based reinforcement learning to align outputs with market needs. Additionally, GIRL demonstrates how generated JDs can enhance conventional discriminative recommender systems, boosting performance in both typical and cold-start scenarios. Extensive experiments on a real-world recruitment dataset show that generation quality improves with RL fine-tuning and that generation-enhanced recommendations yield tangible gains over strong baselines. The work suggests a pathway toward more holistic, advisory, and explainable career AI tools.

Abstract

The rapid development of online recruitment services has encouraged the utilization of recommender systems to streamline the job seeking process. Predominantly, current job recommendations deploy either collaborative filtering or person-job matching strategies. However, these models tend to operate as "black-box" systems and lack the capacity to offer explainable guidance to job seekers. Moreover, conventional matching-based recommendation methods are limited to retrieving and ranking existing jobs in the database, restricting their potential as comprehensive career AI advisors. To this end, here we present GIRL (GeneratIve job Recommendation based on Large language models), a novel approach inspired by recent advancements in the field of Large Language Models (LLMs). We initially employ a Supervised Fine-Tuning (SFT) strategy to instruct the LLM-based generator in crafting suitable Job Descriptions (JDs) based on the Curriculum Vitae (CV) of a job seeker. Moreover, we propose to train a model which can evaluate the matching degree between CVs and JDs as a reward model, and we use Proximal Policy Optimization (PPO)-based Reinforcement Learning (RL) method to further fine-tine the generator. This aligns the generator with recruiter feedback, tailoring the output to better meet employer preferences. In particular, GIRL serves as a job seeker-centric generative model, providing job suggestions without the need of a candidate set. This capability also enhances the performance of existing job recommendation models by supplementing job seeking features with generated content. With extensive experiments on a large-scale real-world dataset, we demonstrate the substantial effectiveness of our approach. We believe that GIRL introduces a paradigm-shifting approach to job recommendation systems, fostering a more personalized and comprehensive job-seeking experience.
Paper Structure (27 sections, 15 equations, 6 figures, 4 tables, 1 algorithm)

This paper contains 27 sections, 15 equations, 6 figures, 4 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. RELATED WORK
  3. Job Recommendation
  4. Large Language Models
  5. LLMs for Recommendation
  6. Problem Definition
  7. Generative Recommendation Framework
  8. Supervised Fine-tuning
  9. Reward Model Training
  10. Reinforcement Learning
  11. Job Description Generation
  12. KL Divergence Computation
  13. Reward and Advantages Computation
  14. Actor Model Optimization
  15. Critic Model Optimization
  16. ...and 12 more sections

Figures (6)

  • Figure 1: Schematic diagram of three distinct job recommendation paradigms.
  • Figure 2: The training framework of the generative recommendation model.
  • Figure 3: The prompt template of training step one.
  • Figure 4: The prompt template for generation quality evaluation.
  • Figure 5: Performance of different models with different generation number.
  • ...and 1 more figures

Theorems & Definitions (2)

  • Definition 1: Generative Job Recommendation
  • Definition 2: Generation-Enhanced Job Recommendation