A Formal Comparison Between Chain-of-Thought and Latent Thought
This work provides practical guidance for researchers and practitioners in AI on choosing between reasoning methods, but it is incremental as it builds on existing paradigms without introducing a new one.
The paper tackled the problem of comparing reasoning paradigms in large language models, specifically Chain-of-Thought (CoT) and Latent Thought, by formally analyzing their computational properties, finding that Latent Thought enables more efficient parallel computation while CoT is better for approximating solutions to intractable problems.
Chain-of-Thought (CoT) elicits reasoning in large language models by explicitly generating intermediate steps in natural language. In contrast, Latent Thought in looped models operates directly in the continuous latent space, enabling computation beyond discrete linguistic representations. While both approaches exploit iterative computation, their comparative capabilities remain underexplored. In this work, we present a formal analysis showing that Latent Thought in Looped Transformers enables parallel computation, which is more efficient than the inherently sequential process of CoT. In contrast, CoT leverages stochastic decoding to approximate solutions to problems where exact computation is intractable. These separations suggest the tasks for which depth-driven recursion is more suitable, thereby offering practical guidance for choosing between reasoning paradigms. Code is available at https://github.com/kevin671/cot-vs-loop.