Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Enhancing Talent Search Ranking with Role-Aware Expert Mixtures and LLM-based Fine-Grained Job Descriptions

Jihang Li, Bing Xu, Zulong Chen, Chuanfei Xu, Minping Chen, Suyu Liu, Ying Zhou, Zeyi Wen

TL;DR

The paper tackles talent search by addressing nuanced job-specific preferences, recruiter-role heterogeneity, and noisy subjective signals. It introduces an LLM-based fine-grained job description encoding, a role-aware multi-gate MoE to capture recruiter behavior, and a multi-task objective aligning CTR, CVR, and resume relevance. Empirical results show offline AUC gains (1.70% for CTR, 5.97% for CVR) and a 17.29% uplift in CTCVR, translating to significant annual cost savings. The framework demonstrates practical value by reducing dependence on external sourcing channels while improving match quality on a production-scale platform.

Abstract

Talent search is a cornerstone of modern recruitment systems, yet existing approaches often struggle to capture nuanced job-specific preferences, model recruiter behavior at a fine-grained level, and mitigate noise from subjective human judgments. We present a novel framework that enhances talent search effectiveness and delivers substantial business value through two key innovations: (i) leveraging LLMs to extract fine-grained recruitment signals from job descriptions and historical hiring data, and (ii) employing a role-aware multi-gate MoE network to capture behavioral differences across recruiter roles. To further reduce noise, we introduce a multi-task learning module that jointly optimizes click-through rate (CTR), conversion rate (CVR), and resume matching relevance. Experiments on real-world recruitment data and online A/B testing show relative AUC gains of 1.70% (CTR) and 5.97% (CVR), and a 17.29% lift in click-through conversion rate. These improvements reduce dependence on external sourcing channels, enabling an estimated annual cost saving of millions of CNY.

Enhancing Talent Search Ranking with Role-Aware Expert Mixtures and LLM-based Fine-Grained Job Descriptions

TL;DR

The paper tackles talent search by addressing nuanced job-specific preferences, recruiter-role heterogeneity, and noisy subjective signals. It introduces an LLM-based fine-grained job description encoding, a role-aware multi-gate MoE to capture recruiter behavior, and a multi-task objective aligning CTR, CVR, and resume relevance. Empirical results show offline AUC gains (1.70% for CTR, 5.97% for CVR) and a 17.29% uplift in CTCVR, translating to significant annual cost savings. The framework demonstrates practical value by reducing dependence on external sourcing channels while improving match quality on a production-scale platform.

Abstract

Talent search is a cornerstone of modern recruitment systems, yet existing approaches often struggle to capture nuanced job-specific preferences, model recruiter behavior at a fine-grained level, and mitigate noise from subjective human judgments. We present a novel framework that enhances talent search effectiveness and delivers substantial business value through two key innovations: (i) leveraging LLMs to extract fine-grained recruitment signals from job descriptions and historical hiring data, and (ii) employing a role-aware multi-gate MoE network to capture behavioral differences across recruiter roles. To further reduce noise, we introduce a multi-task learning module that jointly optimizes click-through rate (CTR), conversion rate (CVR), and resume matching relevance. Experiments on real-world recruitment data and online A/B testing show relative AUC gains of 1.70% (CTR) and 5.97% (CVR), and a 17.29% lift in click-through conversion rate. These improvements reduce dependence on external sourcing channels, enabling an estimated annual cost saving of millions of CNY.

Paper Structure

This paper contains 32 sections, 5 equations, 8 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Methodology
  3. Problem Statement
  4. Method Overview
  5. Encoding Module
  6. Personalized MMoE Module
  7. Multi-task Learning Module
  8. Experiments
  9. Experiment Settings
  10. Datasets
  11. Baselines
  12. Metrics
  13. Offline Evaluation
  14. Online A/B Test
  15. Ablation Study
  16. ...and 17 more sections

Figures (8)

  • Figure 1: Recruitment process of our online system.
  • Figure 2: Overview of proposed framework.
  • Figure 3: A/B test result. Values are concealed due to confidentiality.
  • Figure 4: Impact of the number of experts in the MMoE module and the historical sequence length in the JD encoding module on the final performance. The performance is the average AUC result of the CTR prediction task and the CVR prediction task.
  • Figure 5: Case study of candidate ranking for a "Search Product Manager" position.
  • ...and 3 more figures