Program-of-Thought
A method in which language models generate executable programs, such as Python code, to solve reasoning problems, deriving final answers through program execution. It achieves greater precision than Chain-of-Thought on mathematical and logical tasks.
Timeline (last 90 days)
For comparison, the Program-of-Thought baseline achieves 72.0% overall accuracy and the tool-calling agent achieves 58.7%.
In the 30-token configuration, the approach is 8.8× faster than Program-of-Thought and 2.8× more energy-efficient.
Program-of-Thought is a prompting strategy that has language models generate executable code to solve mathematical problems.
All Facts (3)
For comparison, the Program-of-Thought baseline achieves 72.0% overall accuracy and the tool-calling agent achieves 58.7%.
50%UnverifiedIn the 30-token configuration, the approach is 8.8× faster than Program-of-Thought and 2.8× more energy-efficient.
50%UnverifiedProgram-of-Thought is a prompting strategy that has language models generate executable code to solve mathematical problems.
50%