Fast Thinking for Large Language Models
This addresses the problem of high latency and token usage in reasoning-oriented LLMs for users needing efficient and controllable reasoning, representing an incremental improvement by combining existing techniques with novel components.
The paper tackles the inefficiency of explicit reasoning tokens in large language models by introducing a framework that uses concise CoT sketches during training to learn a codebook of discrete strategy priors, enabling fast inference with continuous thinking vectors and adaptive routing, achieving competitive or superior accuracy while substantially lowering inference cost across multiple reasoning benchmarks.
Reasoning-oriented Large Language Models (LLMs) often rely on generating explicit tokens step by step, and their effectiveness typically hinges on large-scale supervised fine-tuning or reinforcement learning. While Chain-of-Thought (CoT) techniques substantially enhance performance on complex reasoning tasks, they remain inefficient, requiring long reasoning traces that increase latency and token usage. In this work, we introduce Latent Codebooks for Fast Thinking, a framework that uses concise CoT sketches only during training to learn a codebook of discrete strategy priors. At inference, the model conditions on a handful of continuous thinking vectors distilled from the codebook in a single pass, enabling strategy-level guidance without producing explicit reasoning tokens. To complement this design, we propose GainRouter, a lightweight routing mechanism that adaptively switches between fast codebook guided inference and slow explicit reasoning, thereby suppressing overthinking and reducing unnecessary token generation. Experiments across multiple reasoning benchmarks show that our approach achieves competitive or superior accuracy while substantially lowering inference cost, offering a practical path toward efficient and controllable reasoning in large language models.