Powering Job Search at Scale: LLM-Enhanced Query Understanding in Job Matching Systems
- Paper: https://arxiv.org/abs/2509.09690
- Authors: Ping Liu, Jianqiang Shen, Qianqi Shen, Chunnan Yao, Kevin Kao, Dan Xu, et al. (12 authors total)
- Affiliation: LinkedIn
- Date: August 2025
- Topic area: Query Understanding / Job Search
- A unified perspective on query understanding for job matching, organizing the problem space of mapping free-text job seeker queries to structured intents.
- A clear decomposition of method components — representation learning, retrieval, reranking, and generation/feedback mechanisms — suitable for engineering deployment.
- Industrial-scale integration of LLM-based query understanding into an existing retrieval and ranking stack, under real-world latency, cost, and quality constraints.
- Discussion of interfaces with emerging paradigms such as LLM tool use, reinforcement learning, and multi-agent collaboration, along with the path from research prototype to production system.
- Identification of open problems: evaluation trustworthiness, latency and cost, hallucination and safety, and cross-lingual / multimodal extension.
- Datasets: public IR benchmarks (e.g., MS MARCO, BEIR, Natural Questions) and proprietary domain corpora.
- Metrics: nDCG@10, MRR, Recall@k, Hit@k, human preference, task success rate, latency, and token cost.
- Baselines: BM25, dense retrieval, cross-encoder reranking, retrieval-free LLMs, and commercial search APIs.
- Ablations: contribution of retrieval steps, reranking depth, and training data scale.
- Aligned Query Expansion: Efficient Query Expansion for Information Retrieval
- Beyond the Limitation of a Single Query: Train Your LLM for Query Expansion
- Decomposing Complex Queries for Tip-of-the-tongue Retrieval
- Few-Shot Generative Conversational Query Rewriting, SIGIR 2020, DOI: 10.1145/3397271.3401323
- Hierarchical Query Classification in E-commerce Search, WWW 2024
- Hypothetical Documents or Knowledge Leakage? Rethinking LLM-based Query Expansion
- Original paper: Powering Job Search at Scale: LLM-Enhanced Query Understanding in Job Matching Systems, arXiv, https://arxiv.org/abs/2509.09690
Overview
This paper presents LinkedIn's work on applying LLM-enhanced query understanding to job matching and job search at production scale. In large-scale search, recommendation, and personalization systems, information retrieval has long faced challenges around efficiency, scalability, and user intent understanding. Traditional pipeline approaches tend to separate retrieval, ranking, and generation, which makes it difficult to meet users' expectations for natural language interaction, multi-hop reasoning, and real-time knowledge in the LLM era.
The paper addresses this gap in the context of job search, where user queries are often ambiguous, incomplete, or phrased in non-standard ways that do not directly map to structured job titles, skills, or locations in a job taxonomy.
Core Contributions
Method / System Architecture
The general technical route follows a four-step pattern:
1. Input and representation: encode queries, documents, and user context as dense or sparse representations, or as structured prompts. 2. Core modules: retrievers, rerankers, planners, and tool interfaces connected in series or parallel. 3. Learning strategies: supervised fine-tuning, contrastive learning, distillation, reinforcement learning (including process rewards), and bootstrapped data synthesis. 4. Inference strategies: single-pass retrieval, iterative retrieval, parallel sub-queries, early stopping, and budget control.
Evaluation
Typical experimental protocols in this problem domain include:
Specific quantitative results should be verified against the original PDF tables; this summary is based on the abstract and public metadata.
Key Takeaways for Search / Rec / Personalization
1. Architecture: cascaded retrieval + reranking + generation remains mainstream, but agentic paradigms are making retrieval strategy itself a learnable object. 2. Data: high-quality instruction data and click/session logs are both critical; synthetic data requires safeguards against knowledge leakage and distribution shift. 3. Evaluation: the gap between offline metrics and online satisfaction is widening; LLM-as-judge should be cross-validated with human evaluation. 4. Product: latency, cost, explainability, and safety policies are hard constraints in industrial deployment — optimizing academic benchmarks alone is insufficient.
Related Work
Glossary
| Term | Meaning | |------|---------| | IR | Information Retrieval | | RAG | Retrieval-Augmented Generation | | LTR | Learning to Rank | | nDCG | Normalized Discounted Cumulative Gain, a ranking quality metric | | Agentic Search | Modeling search as sequential decision-making and tool invocation | | Gen-IR | Generative Information Retrieval |