TII launches CrowdLabel challenge
Innovators worldwide invited to redefine human data-labelling
#UAE #datalabelling – Technology Innovation Institute (TII), the applied research arm of Abu Dhabi’s Advanced Technology Research Council (ATRC), has launched a challenge to revolutionise human data labeling for large language models (LLMs). The TII CrowdLabel Challenge invites global innovators to design cost-effective, scalable crowdsourced methods for human labeling of samples for use in LLM alignment, a process used to make models helpful, safe, and reliable. Submissions can be submitted up until 28 February 2025.
SO WHAT? – The TII CrowdLabel Challenge aims to address the high costs of human labeling for LLM alignment. Large language models finetuned by Reinforcement Learning with Human Feedback (RLHF) consistently outperform their competitors, hiring human labelers to produce the annotated data samples required can be costly for AI companies. This new challenge is part of TII’s effort to identify innovative crowdsourced methods for human labeling of samples, that can reduce these high costs and allow more LLM builders to align their models.
Here are some key points about the CrowdLabel Challenge:
Abu Dhabi’s Technology Innovation Institute (TII) has announced the TII CrowdLabel Challenge, which invites global innovators to design cost-effective, scalable crowdsourced methods for human labeling of samples for use in LLM alignment.
The challenge seeks passive human labeling methods that integrate seamlessly into daily tasks or processes. Solutions must work in at least two languages, support multimodal content, and handle diverse topics.
The $50,000 prize pool for the challenge includes an initial award for the Best Idea ($5,000), followed by an award for Best Implementation ($15,000) and an award for the Best Overall submission ($30,000).
The TII CrowdLabel Challenge is open to individuals, teams, startups, universities, and research institutions. Submissions will be judged on novelty, cost-efficiency, and labeling scalability.
Submissions for Stage 1 of the TII CrowdLabel Challenge must be received by 11:59 PM (US Eastern Time) on February 28th, 2025.
LINKS
TII CrowdLabel Challenge (website)