Reproduce “EZInterviewer: To Improve Job Interview Performance with Mock Interview Generator” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Mingzhe Li, Xiuying Chen, Weiheng Liao, Yang Song, Tao Zhang, Dongyan Zhao, Rui Yan
EZInterviewer, a mock interview generator, leverages a disentangled architecture and pre-training to produce personalized and relevant interview questions based on candidate resumes and job descriptions. The system achieved superior performance across automatic and human evaluation metrics, including a 6.92% higher BLEU-1 score, and demonstrated robust performance even with limited training data.
Key insights
- Disentangling the processes of dialogue generation and knowledge selection significantly reduces the model's reliance on scarce knowledge-grounded interview data, enhancing robustness in low-resource settings.
- Pre-training individual components on readily available, large-scale ungrounded dialogues and job-resume matching data proves effective for building specialized, knowledge-grounded dialogue systems.
- A fusion gate in the Decoding Manager allows dynamic weighting and integration of context-driven and knowledge-driven information, enabling the generation of highly relevant and coherent interview questions.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.