HUMANS IN THE LOOP: 'ALIA Challenge'
Apply before February 22, 2026. Spaces are limited.
JOIN THE ALIA LANGUAGE MODEL ALIGNMENT PROCESS AND HELP BUILD A SAFER, FAIRER, AND MORE INCLUSIVE PUBLIC AI.
Take part in Humans in the Loop: ALIA Challenge, a 10-hour hackathon focused on the alignment process of ALIA, the public language model launched by the Ministry for Digital Transformation and Public Administration, through the State Secretariat for Digitization and AI, and developed by the Barcelona Supercomputing Center – National Supercomputing Center (BSC-CNS). The hackathon represents the final stage of the ALIA Challenge, where a community of collaborators has annotated real interactions with ALIA, and is a key step in the overall alignment process.
A critical phase: evaluate and identify the correlation between AI responses and human ones.
Through the Humans in the Loop challenges, you will contribute to data curation and build evaluation pipelines that compare human judgment with an AI-based assessor. This is your chance to demonstrate how to align and evaluate public AI while keeping humans actively in the loop.
SHAPE A PUBLIC AI MODEL
Contribute directly to the alignment and evaluation process of ALIA. Through your work, you will help build a safer, fairer, and more inclusive public AI
IMMERSION IN AI SYSTEMS
The value lies not in the solution, but in the process. As a participant, you will dive into AI safety, governance, and transparency, and explore how these aspects translate into technical design and development.
TAKE YOUR SKILLS TO THE NEXT LEVEL
You will receive prior guidance on how to use ALIA’s resources and gain hands-on experience in critical areas of AI development, such as alignment, evaluation, bias, and robustness testing.
Two main challenges, including:
- Human-criteria data validation: You will have the opportunity to validate and select an evaluation dataset to help establish a reliable human benchmark for alignment—a crucial step in confirming human judgment.
- Your LLM, your AI-based judge!::In this second phase, you will be asked to design and test an AI-based “judge” according to human preferences. Experiment with prompts, models, and optimizations to maximize the correlation between machine judgment and human evaluation. Discover how AI systems can learn to assess alignment and see where they fall short.
If you want to learn more about the ALIA Project, explore its objectives and check out the ALIA Kit.
IN COLLABORATION WITH
Event Highlights
Date
2026
Prize
+ Recognition for the top teams and gift cards
Challenge
Collaborate in creating an LLM as a judge using data from telecom operators.
Testing environment
Join the Talent Arena community today and be part of THIS ACTIVITY!
Questions? Our team is here to help:
[email protected]
