Signals 4 · free daily AI digest

Distance generalization in transformers: why bother with positional encoding?

Paper recorded by Signals 4 on 2026-09-10 in cs.CL. Abstract reproduced from arXiv; link to the original below.

Published 2026-09-10 on arXiv · recorded by Signals 4 on 2026-09-11

Category: cs.CL · 自然语言处理 · first seen 2026-09-11

Abstract

Out-of-distribution length generalization, namely to extrapolate a task from short to longer context, has been studied intensively for transformers. Here we focus on distance generalization, which probes performance when inter-token distances are changed between training and inference, while keeping a fixed context length. We construct two synthetic delay copy tasks, both involving finite distance

Read on arXiv →

#66 most recent of 186 cs.CL papers we have recorded · ↑ newer: Doc2FRC: Length-Consistent Document-Level Machine Translation via Fixe · ↓ older: Nuha-Speech: Building General-Purpose Arabic Speech-LLMs
Cite this page: Distance generalization in transformers: why bother with positional encoding?: the #66 most recent of 186 cs.CL papers we have recorded (as of 2026-09-10). Source: Signals 4 (Signals API) — https://data.jiangzhang.ca/signals4/t/papers/distance-generalization-in-transformers-why-bother-with-positional-encoding.html
Free to quote with attribution to “Signals 4 (Signals API)”. Machine-readable: papers.json
Related: More cs.CL papers · arXiv signals · All papers · Today in AI
Get 4 AI signals a day by email — free.
Subscribe free → See all plans →
Get 4 AI signals a day by email — free
All models · All repos · By company · Daily editions