Salvato in:
Dettagli Bibliografici
Autori principali: Hankemeier, Victoria, Schilling, Malte
Natura: Preprint
Pubblicazione: 2026
Soggetti:
Accesso online:https://arxiv.org/abs/2602.10956
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866917317341872128
author Hankemeier, Victoria
Schilling, Malte
author_facet Hankemeier, Victoria
Schilling, Malte
contents Spatio-temporal models analyze spatial structures and temporal dynamics, which makes them prone to information degeneration among space and time. Prior literature has demonstrated that over-squashing in causal attention or temporal convolutions creates a bias on the first tokens. To analyze whether such a bias is present in temporal attention mechanisms, we derive sensitivity bounds on the expected value of the Jacobian of a temporal attention layer. We theoretically show how off-diagonal attention scores depend on the sequence length, and that temporal attention matrices suffer a diagonal attention sink. We suggest regularization methods, and experimentally demonstrate their effectiveness.
format Preprint
id arxiv_https___arxiv_org_abs_2602_10956
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Stochastic Parroting in Temporal Attention -- Regulating the Diagonal Sink
Hankemeier, Victoria
Schilling, Malte
Machine Learning
Spatio-temporal models analyze spatial structures and temporal dynamics, which makes them prone to information degeneration among space and time. Prior literature has demonstrated that over-squashing in causal attention or temporal convolutions creates a bias on the first tokens. To analyze whether such a bias is present in temporal attention mechanisms, we derive sensitivity bounds on the expected value of the Jacobian of a temporal attention layer. We theoretically show how off-diagonal attention scores depend on the sequence length, and that temporal attention matrices suffer a diagonal attention sink. We suggest regularization methods, and experimentally demonstrate their effectiveness.
title Stochastic Parroting in Temporal Attention -- Regulating the Diagonal Sink
topic Machine Learning
url https://arxiv.org/abs/2602.10956