Argilla has released version 2.4 of its open-source data curation platform, enabling users to build fine-tuning and evaluation datasets directly on the Hugging Face Hub without writing any code. The update streamlines the process of collecting, annotating, and curating data for large language model (LLM) training and evaluation.
Key features of Argilla 2.4 include a new no-code interface that allows teams to create datasets via a simple drag-and-drop workflow, integrated support for the Hugging Face Hub for seamless dataset storage and sharing, and enhanced collaboration tools for annotation projects. The platform also introduces pre-built templates for common tasks such as text classification, summarization, and instruction tuning.
“With Argilla 2.4, we’re lowering the barrier for teams to create high-quality datasets for LLM fine-tuning,” said a spokesperson. “You no longer need engineering resources to set up data pipelines; domain experts can now contribute directly.”
The release also includes improvements in search and filtering, making it easier to spot data issues, and integration with popular LLM frameworks like Transformers and PEFT for immediate fine-tuning. Argilla 2.4 is available now as an open-source tool.