The Marginal Cost of Thought: Optimality and Adaptation in Large Reasoning Models

Introduction

The emergence of large reasoning models has validated a fundamental bet in modern artificial intelligence: namely, that inference time compute represents a lever