Learning to Reason and Memorize with Self-Notes
This addresses a key limitation in language models for tasks requiring complex reasoning and memory, though it appears incremental as an extension of existing prompting techniques.
The paper tackles the problem of large language models struggling with multi-step reasoning and memory retention by introducing Self-Notes, a method that allows models to write down thoughts while reading, resulting in outperformance over chain-of-thought and scratchpad methods across various tasks.
Large language models have been shown to struggle with multi-step reasoning, and do not retain previous reasoning steps for future use. We propose a simple method for solving both of these problems by allowing the model to take Self-Notes. Unlike recent chain-of-thought or scratchpad approaches, the model can deviate from the input context at any time to explicitly think and write down its thoughts. This allows the model to perform reasoning on the fly as it reads the context and even integrate previous reasoning steps, thus enhancing its memory with useful information and enabling multi-step reasoning. Experiments across a wide variety of tasks demonstrate that our method can outperform chain-of-thought and scratchpad methods by taking Self-Notes that interleave the input text.