Adaption has introduced AutoScientist, a new AI tool designed to help models learn specific capabilities quickly through automated fine-tuning, aiming to revolutionize AI model training.
Adaption’s New Approach
Launched on Wednesday, AutoScientist aims to simplify training and fine-tuning of AI models by co-optimizing both data and models. This new method, according to co-founder and CEO Sara Hooker, signifies a shift in how AI training is approached. “What’s super exciting about it is that it co-optimizes both the data and the model, and learns the best way to basically learn any capability,” Hooker stated.
Building on Existing Technologies
AutoScientist builds upon Adaption’s Adaptive Data, a service that facilitates the creation of high-quality datasets over time. The new tool is designed to transform these evolving datasets into continuously improving AI models. Hooker emphasizes that the whole stack should be adaptable, optimizing on the fly for any given task.
Performance Metrics and Accessibility
Adaption claims that AutoScientist has more than doubled win-rates across different models, a result that is difficult to quantify due to the unique nature of the tool. Conventional benchmarks like SWE-Bench or ARC-AGI aren’t applicable, but Adaption remains confident in user satisfaction. The company offers the tool free for the first 30 days post-release.
Unlocking Innovation
Hooker compares AutoScientist to code generation in terms of its potential to unlock innovation in various fields. The tool is expected to significantly impact frontier-level AI model training, making advanced AI capabilities more accessible outside of research labs.
AutoScientist is a promising advancement for AI researchers and developers, providing a new solution for efficient model training and fine-tuning.
Source: https://techcrunch.com/2026/05/13/adaption-aims-big-with-autoscientist-an-ai-tool-that-helps-models-train-themselves/
