Large Language Models Reasoning in a Continuous Latent Space // not just a text based
Large language models (LLMs) are restricted to reason in the “language space”, where they typically express the reasoning process with a chain-of-thought (CoT) to solve a complex reasoning problem. However, we argue that language space may not always be optimal for reasoning. For example, most word tokens are primarily for textual coherence and not essential for reasoning, while some critical tokens require complex planning and pose huge challenges to LLMs. To explore the potential of LLM reasoning in an unrestricted latent space instead of using natural language, we introduce a new paradigm Coconut (Chain of Continuous Thought). We utilize the last hidden state of the LLM as a representation of the reasoning state (termed “continuous thought”). Rather than decoding this into a word token, we feed it back to the LLM as the subsequent input embedding directly in the continuous space. Experiments show that Coconut can effectively augment the LLM on several reasoning tasks. This novel latent reasoning paradigm leads to emergent advanced reasoning patterns: the continuous thought can encode multiple alternative next reasoning steps, allowing the model to perform a breadth-first search (BFS) to solve the problem, rather than prematurely committing to a single deterministic path like CoT. Coconut outperforms CoT in certain logical reasoning tasks that require substantial backtracking during planning, with fewer thinking tokens during inference. These findings demonstrate the promise of latent reasoning and offer valuable insights for future research.
In this paper, we presented Coconut, a novel paradigm for reasoning in continuous latent space. Through extensive experiments, we demonstrated that Coconut significantly enhances LLM reasoning capabilities. Notably, our detailed analysis highlighted how an unconstrained latent space allows the model to develop an effective reasoning pattern similar to breadth-first search (BFS). Future work is needed to further refine and scale latent reasoning methods. One promising direction is pretraining LLMs with continuous thoughts, which may enable models to generalize more effectively across a wider range of reasoning scenarios. We anticipate that our findings will inspire further research into latent reasoning methods, ultimately contributing to the development of more advanced machine reasoning systems.
