Semantic Candidate-Job Matching: A Comparative Evaluation of Dense Embedding Models in Hybrid Retrieval

Sai Yashwant; Siddhartha Jain; Anurag Dubey; Samaroha Chatterjee; Gantala Thulsiram

cs.IRSep 20, 2026

Semantic Candidate-Job Matching: A Comparative Evaluation of Dense Embedding Models in Hybrid Retrieval

Authors: Sai Yashwant, Siddhartha Jain, Anurag Dubey, Samaroha Chatterjee, Gantala Thulsiram

Organizations: ManpowerGroup Services India Pvt. Ltd. · Indian Institute of Technology, Hyderabad

Abstract

This paper presents a comparative evaluation of dense embedding models for semantic candidate-job matching in high-volume staffing workflows. Incoming job descriptions are converted into structured English search text and language-specific keywords through LLM-based parsing, and candidate profiles are indexed as semantically enriched resume representations. We evaluate EmbeddingGemma (base) against EmbeddingGemma fine-tuned with Cached Multiple Negatives Ranking Loss (MNRL) within a unified hybrid retrieval pipeline that fuses vector similarity and full-text relevance via reciprocal rank fusion (RRF), and benchmark both against the MPNet model on a batch comparative evaluation dataset scored through the deployed job-candidate matching scoring pipeline. We further document, with mathematical detail, the broader set of contrastive fine-tuning objectives considered during model development (including AnglE/CoSENT-style refinement) and the empirical rationale for retaining Cached-MNRL-only adaptation as the preferred configuration. To support reproducible model selection, we define a broader evaluation framework comprising standard information retrieval metrics (Recall@K, mean reciprocal rank, nDCG) under the exact hybrid-retrieval protocol; the metrics used for the evaluation reported in this paper are fine-tuning convergence diagnostics and a batch comparative evaluation using the deployed AI-Match score and an independent LLM-as-a-Judge relevance score, and we state this scope explicitly rather than implying the full framework was measured. The paper addresses the gap between general-purpose embedding benchmarks and enterprise job-candidate matching constraints, providing a structured basis for comparing embedding strategies under realistic job-candidate retrieval conditions.

Figures & tables

Loss	Supervision	Primary effect on space
Cached MNRL	Positive pairs (+ in-batch neg.)	Retrieval ranking; local neighborhoods
Triplet / hard-neg	$(a,p,n)$	Margin separation vs. named neg.
Cosine MSE	Graded $y$	Absolute score calibration
CoSENT	Graded $y$ (pairwise)	Ordinal cosine consistency
ST-AnglE ( AnglELoss )	Graded $y$ (pairwise)	CoSENT-style ranking w/ angle sim

Table 1: Inductive bias of embedding losses considered in this work.

Iter.	Relative corpus size	Runtime (h)	Throughput (samples/s)	Final train loss	Min. val. loss
1	Smallest	$2.60$	$14.02$	$0.418$	$2.192$
2	Medium	$3.71$	$14.30$	$0.399$	$2.144$
3	Largest	$6.51$	$14.13$	$0.350$	$\mathbf{2.063}$

Table 2: Cached-MNRL training runs across the three corpus-scaling iterations. Exact corpus sizes are proprietary and withheld; iterations are ordered by increasing corpus size.

Figure 1: Cached-MNRL training loss (a) and validation loss (b) versus training step for the three corpus-scaling iterations, extracted from Azure ML MLflow chart exports. Validation loss improves monotonically with corpus scale despite a higher final training loss for iteration 3.

Iter.	Split	Pearson	Spearman	MAE	MSE	ROC-AUC	PR-AUC	F1	Jobs	R@1	R@3–10
2	Val	$0.203$	$0.186$	$0.360$	$0.173$	$0.576$	$0.567$	$0.705$	$20$	$0.650$	$\geq 0.95$
2	Test	$0.209$	$0.221$	$0.343$	$0.159$	$0.592$	$0.538$	$0.649$	$22$	$0.636$	$1.00$
3	Val	$0.072$	$0.077$	$0.378$	$0.194$	$0.538$	$0.609$	$0.753$	$12$	$0.917$	$1.00$
3	Test	$0.161$	$0.152$	$0.356$	$0.168$	$0.561$	$0.563$	$0.692$	$18$	$0.556$	$1.00$

Table 3: Post-hoc regression/classification/ranking diagnostics on the fine-tuning validation and test posthoc subsets (

n=512

rows per split).

Metric	MPNet base (FR)	Gemma base (EN)	Gemma+MNRL (EN)
Candidates scored $n$	$9{,}332$	$8{,}543$	$9{,}849$
Weighted mean score	$59.3$	$52.7$	$\mathbf{73.5}$
Score mode (bucket)	$60$ – $70$	$50$ – $60$	$70$ – $80$
Modal bucket share	$34.9\%$	$65.3\%$	$81.7\%$
Share scored $\geq 60$	$38.3\%$	$6.2\%$	$95.0\%$

Table 4: AI-Match score distribution by embedding configuration (batch comparative evaluation). Note: AI-Match scores are not directly comparable across embedding backbones; see Section 4.3 discussion below.

Figure 2: Share of candidates exceeding LLM-as-a-Judge relevance thresholds, by embedding configuration, on the batch comparative evaluation dataset ( $n=9{,}292$ / $8{,}065$ / $9{,}848$ judged candidates for MPNet-FR, Gemma-base-EN, and Gemma+MNRL-EN respectively).

Metric	Base EN	Fine-tuned EN	Change
AI-Match weighted mean score	$52.7$	$73.5$	$+20.8$
AI-Match share $\geq 60$	$6.2\%$	$95.0\%$	$+88.8$ pts
LLM-Judge share $>60$	$88.1\%$	$88.6\%$	$+0.5$ pts
LLM-Judge share $>70$	$69.30\%$	$69.35\%$	$+0.05$ pts
LLM-Judge share $>80$	$43.8\%$	$45.6\%$	$+1.7$ pts
LLM-Judge share $>90$	$16.9\%$	$17.7\%$	$+0.8$ pts

Table 5: Effect of Cached-MNRL fine-tuning on EmbeddingGemma (English), isolated from the MPNet Model (French) comparison.

Job-Candidate Matching

See all 13 papers

Apr 20, 2026

Mira-Embeddings-V1: Domain-Adapted Semantic Reranking for Recruitment via LLM-Synthesized Data

Candidate sourcing for recruiters is best viewed as a two-stage retrieval and reranking pipeline with recall as the primary objective under a limited review budget. An upstream production retriever first returns a candidate shortlist for each job description (JD), and our goal is to rerank that shortlist so that qualified candidates appear as high as possible. We present mira-embeddings-v1, a semantic reranking system for the recruitment domain that reshapes the embedding space with LLM-synthesized training data and corrects boundary confusions with a lightweight reranking head. Starting from real JDs, we build a five-stage prompt pipeline to generate diverse positive and hard negative samples that sculpt the semantic space from multiple angles. We then apply a two-round LoRA adaptation: JD--JD contrastive training followed by JD--CV triplet alignment on a heterogeneous text dataset. Importantly, these gains require no large-scale manually labeled industrial training pairs: a modest set of real JDs is expanded into supervision through LLM synthesis. Finally, a BoundaryHead MLP reranks the Top-K results to distinguish between roles that share the same title but differ in scope. On a local pool of 300 real JDs with candidates from an upstream production retriever, mira-embeddings-v1 improves Recall@50 from 68.89% (baseline) to 77.55% while lifting Precision@10 from 35.77% to 39.62%. On a supportive global pool over 44,138 candidates judged by a Qwen3-32B rubric, it achieves Recall@200 of 0.7047 versus 0.5969 for the baseline. These results show that LLM-synthesized supervision with boundary-aware reranking yields robust gains without a heavy cross-encoder.

Zhaohua Liang, Zhilin Wang, Renjie Cao +1

OpenJobs AI · San Francisico, CA, USA

Learning to Rank Semantic Search Job-Candidate Matching Recommender Systems Text Embedding Models

Ask PDF

Oct 10, 2026

Fine-Tuning Strategy, Not Just Data: A Statistically Rigorous Audit of Gender and Stereotypical Bias in Embedding Models for HR Recommendation Systems

Embedding models underpin modern job-candidate matching: resumes, job descriptions, and queries are encoded as dense vectors, and recommendations come from nearest-neighbor retrieval over that space. Any social bias in the embeddings therefore feeds directly into HR recommendations. We audit six multilingual sentence-embedding models (LaBSE- and E5-family, both public and custom fine-tuned) using Bias-in-Bios (occupational gender bias) and StereoSet (stereotypes across profession, race, gender, and religion). We treat statistical validity as a first-class concern: we identify a non-independence error in a naive significance test, where pooling non-independent pairwise similarities inflated a trivial 0.0046 mean difference to

p \approx 0

, and replace it with a profession-level paired Wilcoxon signed-rank test, individual-resampling bootstrap confidence intervals, and permutation tests for StereoSet. Under the corrected methodology, knowledge-distillation (KD) fine-tuning of E5-Base significantly increases measured bias on Bias-in-Bios (

Δ= -0.177

p < 10^{-8}

; more negative means more bias), although this measure is sensitive to global embedding geometry. The KD result is corroborated by pro-stereotype patterns in the StereoSet profession category (uncorrected

p < 0.05

). Binary-classification fine-tuning of the same base model shows no statistically distinguishable change (

p = 0.218

). Fine-tuning strategy is thus an actionable design lever for practitioners building HR recommenders. We also report which benchmark categories are underpowered, and note that two domain-adapted models could not be included in significance testing due to a software-access constraint, which limits our conclusions.

Ekta Yadav

Randstad N.V.

Text Embedding Models Recommender Systems

Ask PDF

Apr 2, 2026

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

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.

Ansel Kaplan Erol, Seohee Yoon, Keenan Hom +1

Georgia Institute of Technology, Atlanta, GA, USA

Job-Candidate Matching Recommender Systems Large Language Model-Based Recommendation Large Language Model-Guided Optimization Explainable Recommendation

Ask PDF

Abstract

Job-Candidate Matching

Explore similar work

Mira-Embeddings-V1: Domain-Adapted Semantic Reranking for Recruitment via LLM-Synthesized Data

Fine-Tuning Strategy, Not Just Data: A Statistically Rigorous Audit of Gender and Stereotypical Bias in Embedding Models for HR Recommendation Systems

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