Patience Is The Key to Large Language Model Reasoning
This work addresses the challenge of improving reasoning in large language models for complex tasks, but it is incremental as it builds on existing Chain of Thought approaches.
The paper tackles the problem of large language models sacrificing detailed reasoning for brevity or requiring extensive training data, by proposing a method to encourage a more patient reasoning style through preference optimization, resulting in a performance increase of up to 2.1% on GSM8k with lightweight dataset training.
Recent advancements in the field of large language models, particularly through the Chain of Thought (CoT) approach, have demonstrated significant improvements in solving complex problems. However, existing models either tend to sacrifice detailed reasoning for brevity due to user preferences, or require extensive and expensive training data to learn complicated reasoning ability, limiting their potential in solving complex tasks. To bridge this gap, following the concept of scaling test-time, we propose a simple method by encouraging models to adopt a more patient reasoning style without the need of introducing new knowledge or skills. To employ a preference optimization approach, we generate detailed reasoning processes as positive examples and simple answers as negative examples, thereby training the model to favor thoroughness in its responses. Our results demonstrate a performance increase of up to 2.1% on GSM8k with training just on a lightweight dataset.