Create and Train a 25M-Parameter LLM from Scratch Using Only Your CPU
A tutorial shows building, pre-training, and fine-tuning a 25M-parameter LLM on CPU, covering architecture, loss, and reinforcement learning.
reinforcement learning
Accelerating AI and Quantum Deployment from MIT to IBM
MIT-IBM Computing Research Lab helps former MIT researchers bring AI, RL, quantum ML, and trustworthy AI ideas from academia into IBM products.
Collecting feedback from humans
RL-Teacher is an open-source tool that trains AI agents through occasional human feedback instead of hand-crafted reward functions. It addresses the difficulty of specifying rewards for complex real-w…
Imperfect reward functions in real-world settings
This article introduces reward misspecification in reinforcement learning: reward functions are only proxies for our true goals, and powerful optimizers exploit gaps between what we write and what we …
Building Your First World Model from the Ground Up
Learn to build a world model in Python, daydream CartPole trajectories, and measure when imagined states diverge from real dynamics.
Using TRL and OpenEnv to teach a code-generation model the art of watercolor painting
A viral project showed a language model producing watercolor paintings by writing JavaScript with p5.brush rather than generating images directly. This article recreates Surya Narreddi's recipe fully …