σ-GPTs: A New Approach to Autoregressive Models
This addresses the inefficiency of sequential generation in autoregressive models for applications requiring flexible token sampling, though it is an incremental improvement on existing methods.
The paper tackles the fixed-order generation limitation of autoregressive models like GPT by introducing positional encoding for outputs, enabling dynamic per-sample order modulation and sub-linear model evaluations. It demonstrates a tenfold reduction in generation steps across language modeling, path-solving, and aircraft prediction tasks.
Autoregressive models, such as the GPT family, use a fixed order, usually left-to-right, to generate sequences. However, this is not a necessity. In this paper, we challenge this assumption and show that by simply adding a positional encoding for the output, this order can be modulated on-the-fly per-sample which offers key advantageous properties. It allows for the sampling of and conditioning on arbitrary subsets of tokens, and it also allows sampling in one shot multiple tokens dynamically according to a rejection strategy, leading to a sub-linear number of model evaluations. We evaluate our method across various domains, including language modeling, path-solving, and aircraft vertical rate prediction, decreasing the number of steps required for generation by an order of magnitude.