Interpretable Representation Learning for Motion Forecasting

We address interpretable representation learning for motion forecasting in self-driving cars. Rather than treating transformers as black boxes, we develop methods to interpret and modify learned representations. We introduce self-supervised pre-training with interpretable objectives. Moreover, we pr...

Cur síos iomlán

Sábháilte in:
Sonraí bibleagrafaíochta
Príomhchruthaitheoir: Wagner, Royden
Formáid: Online
Teanga:Béarla
Foilsithe / Cruthaithe: KIT Scientific Publishing 2026
Ábhair:
Rochtain ar líne:ONIX_20260519T105721_9783731514749_14
Clibeanna: Cuir clib leis
Níl clibeanna ann, Bí ar an gcéad duine le clib a chur leis an taifead seo!
_version_ 1869528448057212928
author Wagner, Royden
author_browse Wagner, Royden
author_facet Wagner, Royden
author_sort Wagner, Royden
collection Directory of Open Access Books
description We address interpretable representation learning for motion forecasting in self-driving cars. Rather than treating transformers as black boxes, we develop methods to interpret and modify learned representations. We introduce self-supervised pre-training with interpretable objectives. Moreover, we probe latent spaces of forecasting models and reveal interpretable features, allowing us to make targeted interventions. Finally, we uncover retrocausal mechanisms, which enable goal-based instructions.
format Online
id doab-20.500.12854ir-176606
institution Directory of Open Access Books
language eng
publishDate 2026
publishDateRange 2026
publishDateSort 2026
publisher KIT Scientific Publishing
publisherStr KIT Scientific Publishing
record_format ojs
spelling doab-20.500.12854ir-1766062026-05-20T06:05:23Z Interpretable Representation Learning for Motion Forecasting Wagner, Royden Autonomes Fahren Bewegungsvorhersage Maschinelles Lernen Mechanistic interpretability Motion forecasting Representation learning Self-driving cars Transformer models thema EDItEUR::U Computing and Information Technology::UY Computer science We address interpretable representation learning for motion forecasting in self-driving cars. Rather than treating transformers as black boxes, we develop methods to interpret and modify learned representations. We introduce self-supervised pre-training with interpretable objectives. Moreover, we probe latent spaces of forecasting models and reveal interpretable features, allowing us to make targeted interventions. Finally, we uncover retrocausal mechanisms, which enable goal-based instructions. 2026-05-20T06:05:16Z 2026-05-20T06:05:16Z 2026-05-19T12:51:58Z 2026 book ONIX_20260519T105721_9783731514749_14 1613-4214 (Online) https://library.oapen.org/handle/20.500.12657/113239 9783731514749 https://directory.doabooks.org/handle/20.500.12854/176606 eng Schriftenreihe / Institut für Mess- und Regelungstechnik, Karlsruher Institut für Technologie open access image/jpeg Attribution 4.0 International https://library.oapen.org/bitstream/20.500.12657/113239/1/9783731514749.pdf KIT Scientific Publishing KIT Scientific Publishing 10.5445/KSP/1000191275 10.5445/KSP/1000191275 68fffc18-8f7b-44fa-ac7e-0b7d7d979bd2 9783731514749 KIT Scientific Publishing 172 Karlsruhe, Germany open access
spellingShingle Autonomes Fahren
Bewegungsvorhersage
Maschinelles Lernen
Mechanistic interpretability
Motion forecasting
Representation learning
Self-driving cars
Transformer models
thema EDItEUR::U Computing and Information Technology::UY Computer science
Wagner, Royden
Interpretable Representation Learning for Motion Forecasting
title Interpretable Representation Learning for Motion Forecasting
title_full Interpretable Representation Learning for Motion Forecasting
title_fullStr Interpretable Representation Learning for Motion Forecasting
title_full_unstemmed Interpretable Representation Learning for Motion Forecasting
title_short Interpretable Representation Learning for Motion Forecasting
title_sort interpretable representation learning for motion forecasting
topic Autonomes Fahren
Bewegungsvorhersage
Maschinelles Lernen
Mechanistic interpretability
Motion forecasting
Representation learning
Self-driving cars
Transformer models
thema EDItEUR::U Computing and Information Technology::UY Computer science
topic_facet Autonomes Fahren
Bewegungsvorhersage
Maschinelles Lernen
Mechanistic interpretability
Motion forecasting
Representation learning
Self-driving cars
Transformer models
thema EDItEUR::U Computing and Information Technology::UY Computer science
url ONIX_20260519T105721_9783731514749_14
work_keys_str_mv AT wagnerroyden interpretablerepresentationlearningformotionforecasting