Speculative Decoding Across Languages
Nirajan Paudel*, Michael Ginn*, Luc De Nardi, Alexis Palmer · 2026
Read Speculative Decoding Across Languages by Nirajan Paudel*, Michael Ginn*, Luc De Nardi, Alexis Palmer (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-30580.txt.
Description
Paudel and colleagues test speculative decoding, drafting tokens with a small model for a large one to check, across eleven languages. The speedup, well studied in English, largely evaporates for low-resource languages whose small draft models lack multilingual skill. Distilling the large model into the draft model generalized poorly across tasks in the same language, suggesting a fully representative task-agnostic dataset is infeasible there. Weaker n-gram draft models, by contrast, delivered moderate speedups simply because they cost almost nothing to run.
Catalog record
| CREATOR | Nirajan Paudel*, Michael Ginn*, Luc De Nardi, Alexis Palmer |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 24020 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.30580