Motivation
In today’s rapidly changing socio-technological landscape, industries and workplaces are transforming quickly. Technological advancements, such as task automation and Artificial Intelligence (AI), are reshaping the labor market by creating new roles that demand specialized skills, often difficult to source. The rise of remote hiring, fueled by technological innovation, has expanded the labor market to a global and multilingual scale. Simultaneously, social progress is narrowing ethnic and gender disparities within companies, fostering more inclusive workplaces.
Simultaneously, there has been rapid progress in the development and deployment of language-based systems, driven in part by the creation of the Large Language Models (LLMs). These advances are revolutionizing the use of technology in Human Capital Management (HCM) and Human Resources (HR), enabling the generation of systems able to process large volumes of data and facilitate the identification of the best candidates for specific roles based on their education, work experience and professional skills.
Integrating language technologies into HCM significantly enhances key areas. In sourcing and hiring, these tools improve candidate matching by analyzing their skills and experience. During onboarding and training, they create personalized learning materials tailored to individual employee needs. For strategic workforce planning, NLP tools predict market skill trends and future company demands. Additionally, in career development, these technologies monitor employee progress, supporting targeted upskilling and reskilling aligned with both organizational goals and personal aspirations.
Despite all these benefits, the development and implementation of these systems present challenges such as:
Multilingualism: The global nature of modern workforces means that companies often need to manage employees and candidates who speak multiple languages. This requires language-based systems to not only understand and process various languages accurately but also to maintain the context and cultural nuances inherent in each. Developing systems that effectively handle multiple languages is a complex task that involves significant computational resources and sophisticated NLP techniques.
Fair models and transparency: Ensuring fairness and reducing bias is a critical challenge in HCM. Given the potential semi-automation of some of these systems, it is necessary to build models that do not introduce biases that may discriminate against people of different ethnicities or genders, guaranteeing equity and diversity in decision making. Beyond deciding who fits a vacancy, systems increasingly need to explain why. Recruiters need to understand a ranking, organizations need to audit it, and a growing body of regulation on automated decision-making expects hiring tools to be explainable, not merely accurate.
Cross-Industry Adaptability: The needs and HR selection criteria vary significantly across industries. NLP systems must be flexible enough to align with the unique requirements, standards, and practices of each sector, from healthcare to technology to retail, ensuring they are effective and relevant in various contexts.
Goals of the lab
TalentCLEF seeks to promote technological development taking into account these critical aspects for the correct implementation of systems in real environments. Specifically, TalentCLEF aims to:
Establish a public benchmark of NLP models in the HR field: setting up a common framework for the evaluation of NLP models applied to human resources. This allows the comparison of methodologies and facilitates technological development in the area, ensuring a fair and consistent evaluation of the models, incorporating mechanisms to assess both the bias of the models, allowing to evaluate their fairness regardless of sociodemographic factors, and their cross-industry adaptability.
Create a meeting point for the discussion and evaluation of NLP systems applied to HR: providing a unified space where researchers and practitioners can discuss and evaluate NLP models applied to human resources. Although different workshops in this area have emerged in recent years (i.e. NLP4HR, RecSys in HR, AI4HR & PES, and TMC), there is a lack of environments in which to carry out comparative evaluation of the models.
Pushing the state-of-the-art: community assessment campaigns such as BioCreative or BioASQ have proved to be effective in areas such as information extraction and retrieval applied to bioinformatics and biomedicine. TalentCLEF’s goal is to drive similar innovations in the area of HCM, promoting the creation of more advanced and effective solutions in a competitive environment.
Previous editions and lessons learned
The inaugural edition of TalentCLEF (2025) attracted 76 registered teams, 15 submitted working notes, and 280 runs across both tasks. The second edition (2026) consolidated and expanded that community, with 115 registered teams, 18 working notes, and more than 400 submissions. In this third edition our goal is to consolidate and scale up the community created, and to continue building public benchmarks for NLP in Human Capital Management while fostering the development of fair models.
From these two editions we have extracted insights that shape the 2027 design. In the first edition, to avoid privacy concerns, the dataset did not include contextual information; many participants nonetheless enriched the available data using LLMs, which motivated the context-rich, synthetically generated profiles introduced in 2026. A further lesson is that most submitted systems have focused on retrieval-based methods, and that more open-ended tasks tend to see lower initial uptake but encourage a wider variety of system designs. TalentCLEF 2027 builds directly on these observations by retaining the ranking benchmark that participants already know, while introducing a new and more open-ended explainability layer on top of it, designed so that participating in it adds no submission burden to those who only wish to compete on ranking.