2024/12/31 by Kazuki Irie, Irie, Kazuki · 2 citations
Arts and Humanities · Computer Science · Earth and Planetary Sciences · #3D Surveying and Cultural Heritage #Archaeological and Geological Studies #Computation and Language (cs.CL) #FOS: Computer and information sciences #Image Processing and 3D Reconstruction #Machine Learning (cs.LG)
paper · pdf · doi:10.48550/arxiv.2501.00659
openalex publication_date 2024/12/31 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
Do autoregressive Transformer language models require explicit positional encodings (PEs)? The answer is 'no' provided they have more than one layer -- they can distinguish sequences with permuted tokens without the need for explicit PEs. This follows from the fact that a cascade of (permutation invariant) set processors can collectively exhibit sequence-sensitive behavior in the autoregressive setting. This property has been known since early efforts (contemporary with GPT-2) adopting the Transformer for language modeling. However, this result does not appear to have been well disseminated, leading to recent rediscoveries. This may be partially due to a sudden growth of the language modeling community after the advent of GPT-2/3, but perhaps also due to the lack of a clear explanation in prior work, despite being commonly understood by practitioners in the past. Here we review the long-forgotten explanation why explicit PEs are nonessential for multi-layer autoregressive Transformers (in contrast, one-layer models require PEs to discern order information of their inputs), as well as the origin of this result, and hope to re-establish it as a common knowledge.