RLAD: Training LLMs to Discover Abstractions | Best AI papers explained | Podwise