Towards Understanding the Universality of Transformers for Next-Token Prediction

Can a mere next-token predictor faithfully model human intelligence? We crystallize this emerging concern and correct popular misconceptions surrounding it, and advocate a simple multi-token objective. As a starting point, we argue that the two often-conflated phases of next-token predictionautoregressive inference and teacher-forced training -must be treated distinctly. The popular criticism that errors can compound during autoregressive inference, crucially assumes that teacherforcing has learned an accurate next-token predictor. This assumption sidesteps a more deep-rooted problem we expose: in certain classes of tasks, teacher-forcing can simply fail to learn an accurate next-token predictor in the first place. We describe a general mechanism of how teacherforcing can fail, and design a minimal planning task where both the Transformer and the Mamba architecture empirically fail in that manner -remarkably, despite the task being straightforward to learn. Finally, we provide preliminary evidence that this failure can be resolved using teacherless training, a simple modification using dummy tokens that predicts multiple tokens in advance. We hope this finding can ground future debates and inspire explorations beyond the next-token prediction paradigm. We make our code available under https://github.com/gregorbachmann/ Next-Token-Failures * Equal contribution 1 ETH Zürich, Switzerland 2 Google Research, US.