It's incredible how quickly you can go from an idea to a fully-trained model with this system. And the resulting models are really good! The grammar model in the demo video works really well, yet it was trained for just minutes, and was only 1.5B params 🙂
@mattshumer_
-
AutoRL: Simplest Way to Train Task-Specific LLMs
By
–
Introducing `AutoRL` 📈
— Matt Shumer (@mattshumer_) 30 juillet 2025
The world's simplest way to train a task-specific LLM with RL.
*Just write a SENTENCE describing the model you want.*
A chain of AI systems will generate data + rubrics and train a model for you.
Powered by ART, it's open source.
Link in thread: pic.twitter.com/OVxe2hWTZYIntroducing `AutoRL` The world's simplest way to train a task-specific LLM with RL. *Just write a SENTENCE describing the model you want.* A chain of AI systems will generate data + rubrics and train a model for you. Powered by ART, it's open source. Link in thread:
-
AutoRL: Automated Reinforcement Learning for Model Generation
By
–
How AutoRL works, in a nutshell: – The user describes the model they want
Ex: "A model that detects spelling and grammar errors" – OSS models generate a system prompt that will be used a) to generate, and b) for RULER to rank outputs – We generate input data, and RL a model! -
Training weights in machine learning models explained
By
–
sorry, by train i mean actually updating weights, but i appreciate the comment!
-
Training and Fine-tuning Models: Community Discussion Opportunity
By
–
If you train/fine-tune models, I'd love to chat with you. Please comment or DM and I'll reach out!
-
Building a Personal Product: From Problem to Solution
By
–
For the first time in a long time, I just built something *I* wanted. It's a really different feeling than building something other people want. I know the problem so well, and will be using the product from day one. This may become a new Otherside product… we'll see!
-
XML Prompts Remain Superior for Serious LLM Practitioners
By
–
Sorry, but it's not. I've been prompting/fine-tuning LLMs since 2019, and have tried just about everything, ran tests, etc. etc. There's a reason most serious practitioners use XML prompts for almost everything.
-
Tool Call ID Mismatch Errors Fixed in AI Framework
By
–
It's SO good, and doesn't break with annoying tool call ID mismatch errors
-
Building Custom AI Tools Without Built-in Tool Calling
By
–
This is what I do! I just build it myself instead of relying on built-in tool calling 🙂
-
Optimized Prompts for Advanced AI Models
By
–
Oh, I meant that that looked like an optimized-for-o3 prompt!