COMPOSED PROMPTS WITH RLVR

Summary from the TLDR AI newsletter

Composition-RL reused “too-easy” pass-rate-1 prompts by automatically combining multiple problems into new verifiable questions for RLVR training. Across 4B–30B models, the approach improved reasoning over training on the original dataset.