THE CIRCUITS RESEARCH LANDSCAPE: RESULTS AND PERSPECTIVES

Summary from the TLDR AI newsletter

Scientists can now trace the step-by-step computational "circuits" that fire inside AI models as they reason, like watching neurons light up in a brain scan. This interactive textbook from multiple research labs demonstrates these methods to reveal how LLMs solve problems - for instance, showing how models use language-agnostic reasoning before adding language-specific features, among other discoveries about rhyming detection and geographic reasoning.