README.md

metadata

title: RLM Training - Needle in Haystack
sdk: docker
hardware: t4-small

RLM Training - Recursive Language Model Skills

Training Qwen3-0.6B-Base to find needles in haystacks using GRPO.

Task

• Long context with hidden facts
• Model learns to extract specific information
• 20 steps quick test

Based on

• RLM Paper (arXiv:2512.24601)
• Sebastian Raschka's GRPO insights