Looking beyond the next token
This addresses a fundamental limitation in language modeling for tasks requiring long-term planning and reasoning, though it appears incremental as it builds on known data-processing ideas.
The paper tackles the mismatch between causal language model training and human writing by proposing Trelawney, a technique that rearranges and processes training data sequences to improve performance on benchmarks for planning, algorithmic reasoning, and story generation without architectural changes.
The structure of causal language model training assumes that each token can be accurately predicted from the previous context. This contrasts with humans' natural writing and reasoning process, where goals are typically known before the exact argument or phrasings. While this mismatch has been well studied in the literature, the working assumption has been that architectural changes are needed to address this mismatch. We argue that rearranging and processing the training data sequences can allow models to more accurately imitate the true data-generating process, and does not require any other changes to the architecture or training infrastructure. We demonstrate that this technique, Trelawney, and the inference algorithms derived from it allow us to improve performance on several key benchmarks that span planning, algorithmic reasoning, and story generation tasks. Finally, our method naturally enables the generation of long-term goals at no additional cost. We investigate how using the model's goal-generation capability can further improve planning and reasoning. Additionally, we believe Trelawney could potentially open doors to new capabilities beyond the current language modeling paradigm.