Synapse: Evolving Job-Person Fit with Explainable Two-phase Retrieval and LLM-guided Genetic Resume Optimization

arXiv cs.LG / 4/6/2026

💬 OpinionSignals & Early TrendsIdeas & Deep AnalysisModels & Research

共有:

Key Points

Synapse is presented as a multi-stage recruitment recommender that splits the workflow into high-recall candidate generation and high-precision semantic reranking to better match job requirements at scale and under cost constraints.
The system uses dense retrieval with FAISS plus an ensemble approach based on contrastive learning and LLM reasoning, improving recommendation ranking quality substantially over embedding-only baselines (nDCG@10 up by 22%).
To address transparency needs, Synapse adds a retrieval-augmented explanation layer that grounds recommendations in explicit evidence from retrieved materials.
It also introduces an “evolutionary resume optimization” method that refines candidate representations as a black-box optimization problem using Differential Evolution with LLM-guided mutation operators, achieving monotonic gains in recommender scores (over 60% relative in evaluated profiles) without labeled data.
The authors indicate they plan to release code and data after publication, suggesting practical adoption potential for the research pipeline.

Abstract

Modern recruitment platforms operate under severe information imbalance: job seekers must search over massive, rapidly changing collections of postings, while employers are overwhelmed by high-volume, low-relevance applicant pools. Existing recruitment recommender systems typically rely on keyword matching or single-stage semantic retrieval, which struggle to capture fine-grained alignment between candidate experience and job requirements under real-world scale and cost constraints. We present Synapse, a multi-stage semantic recruitment system that separates high-recall candidate generation from high-precision semantic reranking, combining efficient dense retrieval using FAISS with an ensemble of contrastive learning and Large Language Model (LLM) reasoning. To improve transparency, Synapse incorporates a retrieval-augmented explanation layer that grounds recommendations in explicit evidence. Beyond retrieval, we introduce a novel evolutionary resume optimization framework that treats resume refinement as a black-box optimization problem. Using Differential Evolution with LLM-guided mutation operators, the system iteratively modifies candidate representations to improve alignment with screening objectives, without any labeled data. Evaluation shows that the proposed ensemble improves nDCG@10 by 22% over embedding-only retrieval baselines, while the evolutionary optimization loop consistently yields monotonic improvements in recommender scores, exceeding 60% relative gain across evaluated profiles. We plan to release code and data upon publication.