Direct Value Optimization: Improving Chain-of-Thought Reasoning in LLMs with Refined Values
This work addresses the challenge of improving reasoning in LLMs for AI applications, particularly in scenarios lacking explicit human preference information, though it appears incremental as it builds on existing optimization techniques.
The paper tackles the problem of enhancing large language models in complex reasoning tasks by introducing Direct Value Optimization (DVO), a reinforcement learning framework that uses value signals at individual reasoning steps to optimize models, and it shows that DVO consistently outperforms existing offline preference optimization techniques on mathematical and commonsense reasoning tasks.
We introduce Direct Value Optimization (DVO), an innovative reinforcement learning framework for enhancing large language models in complex reasoning tasks. Unlike traditional methods relying on preference labels, DVO utilizes value signals at individual reasoning steps, optimizing models via a mean squared error loss. The key benefit of DVO lies in its fine-grained supervision, circumventing the need for labor-intensive human annotations. Target values within the DVO are estimated using either Monte Carlo Tree Search or an outcome value model. Our empirical analysis on both mathematical and commonsense reasoning tasks shows that DVO consistently outperforms existing offline preference optimization techniques, even with fewer training steps. These findings underscore the importance of value signals in advancing reasoning capabilities and highlight DVO as a superior methodology under scenarios lacking explicit human preference information.