Scientific American

Can you solve these language puzzles? Test your skills with these problems from North America’s biggest linguistics competition

For 20 years, this computational linguistics competition has inspired new generations of innovators in AI and language ...
GitHub

Infinite RL loop that trains an open-source LLM into a SOTA coding agent — self-generating problems, multi-language sandboxed execution, zero human labels.

Train Qwen/Qwen2.5-Coder-7B-Instruct inside an infinite reinforcement learning loop powered by GRPO, AI-generated coding problems, and real code execution across six runtimes. Generator LLM │ ...