Can AI Tell Real Seniority from Inflated Titles? Researchers Built Traps to Find Out
Reading Between the Lines: Classifying Resume Seniority with Large Language Models
Matan Cohen, Shira Shani, Eden Menahem, Yehudit Aperstein, Alexander Apartsin · arXiv:2509.09229 (PDF) · submitted September 11, 2025
Summary by Andrew Johnson, OneTwo Resume editorial team · updated · We are not affiliated with the authors. arXiv papers are preprints and may not be peer-reviewed.
Seniority is one of the harder things to read off a resume — titles inflate, scope gets embellished, and self-presentation varies by culture and confidence. This study evaluates LLMs (including fine-tuned BERT models) on classifying candidate seniority, using a hybrid dataset that mixes real resumes with synthetically generated hard cases designed to simulate exaggerated qualifications and understated seniority.
The adversarial construction is the interesting part: the benchmark specifically tests whether models can read between the lines — catching the padded resume and crediting the modest one.
What the paper reports
- Introduces a hybrid real + synthetic-hard-case dataset for seniority classification, explicitly modeling exaggeration and understatement.
- Evaluates LLMs and fine-tuned BERT architectures on the task.
- Frames automated seniority assessment as a robustness problem, not just a classification problem.
What this means for your resume
Our editorial interpretation — the paper does not give job-seeker advice.
- Screeners are being trained on datasets built to catch inflation — the "everyone exaggerates, so I must too" strategy is aging badly. Precise, verifiable scope statements ("led a team of 4", "owned the release process") beat grand titles.
- The mirror image matters more: if you understate — common for career changers and returners — the machine may file you a level below where you belong. State your scope plainly.
Read it with these caveats
arXiv preprint (v1). Synthetic hard cases test robustness but may not distribute like real-world exaggeration; per-model accuracy figures are in the paper.
Primary source: Reading Between the Lines: Classifying Resume Seniority with Large Language Models — always read the paper before citing it. Spotted an error in our summary? Tell us and we'll fix it with a visible correction.
More in how ai screeners work — and fail
Forget Bias for a Moment — Do LLM Screeners Even Rank Correctly?
Bias audits ask if screeners are fair. This Princeton-line study asks the prior question: are their rankings valid at all?
General-Purpose LLMs vs a Purpose-Built Hiring Model, on 10,000 Real Pairs
OpenAI, Anthropic, Google, Meta, and DeepSeek models benchmarked on ~10,000 real candidate-job pairs — against a domain-specific model that beat them all.
A Name for the "200 Applications, Zero Replies" Problem
High vacancies AND prolonged unemployment, together — this framework blames deterministic screening rejecting qualified candidates via semantic misinterpretation.