Evaluate and improve models and agents using environments

Marin fork of SkyRL (NovaSky-AI/SkyRL) — snapshot of penfever/working

Scalable toolkit for efficient model reinforcement

Harbor is a framework for running agent evaluations and creating and using RL environments.

Data recipes and robust infrastructure for training AI agents

Official Python client library for kubernetes

Post-training with Tinker

Terminal-Bench 2.1

🚧 Accepting Task Submissions 🚧

Autoresearcher + wandb + datadog + overleaf + claude code. All local, all free. Boot a node, git clone, fill the form, let it rip.

slime is an LLM post-training framework for RL Scaling.

The best ChatGPT that $100 can buy.

Terminal Bench for Science

A benchmark for LLMs on complicated tasks in the terminal

The simplest, fastest repository for training/finetuning medium-sized GPTs.

Accelerated First Order Parallel Associative Scan

Understand and test language model architectures on synthetic tasks.

Code for the paper "ViperGPT: Visual Inference via Python Execution for Reasoning"

High-quality implementations of standard and SOTA methods on a variety of tasks.