7 papers
Don't Forget! Decomposing the Training Dynamics of Memorization in Language Models
Florian Eichin, Philipp Mondorf, Andrei Mircea +3
Memorization has been proposed as a mechanism to explain how language models fit the tail of their training distributions, but its training dynamics are not understood well. In thi…
Copy First, Translate Later: Interpreting Translation Dynamics in Multilingual Pretraining
Felicia Körner, Maria Matveev, Florian Eichin +3
Large language models exhibit impressive cross-lingual capabilities. However, prior work analyzes this phenomenon through isolated factors and at sparse points during training, lim…
Linear Script Representations in Speech Foundation Models Enable Zero-Shot Transliteration
Ryan Soh-Eun Shim, Kwanghee Choi, Kalvin Chang +8
Multilingual speech foundation models such as Whisper are trained on web-scale data, where data for each language consists of a myriad of regional varieties. However, different reg…
What's the Difference? Supporting Users in Identifying the Effects of Prompt and Model Changes Through Token Patterns
Michael A. Hedderich, Anyi Wang, Raoyuan Zhao +3
Prompt engineering for large language models is challenging, as even small prompt perturbations or model changes can significantly impact the generated output texts. Existing evalu…
ExPLAIND: Unifying Model, Data, and Training Attribution to Study Model Behavior
Florian Eichin, Yupei Du, Philipp Mondorf +3
Post-hoc interpretability methods typically attribute a model's behavior to its components, data, or training trajectory in isolation, and are often tied to a particular level of g…
Probing LLMs for Multilingual Discourse Generalization Through a Unified Label Set
Florian Eichin, Yang Janet Liu, Barbara Plank +1
Discourse understanding is essential for many NLP tasks, yet most existing work remains constrained by framework-dependent discourse representations. This work investigates whether…
Semantic Component Analysis: Introducing Multi-Topic Distributions to Clustering-Based Topic Modeling
Florian Eichin, Carolin M. Schuster, Georg Groh +1
Topic modeling is a key method in text analysis, but existing approaches fail to efficiently scale to large datasets or are limited by assuming one topic per document. Overcoming t…