Die Beiträge decken Themen wie mechanistische Interpretierbarkeit, mehrsprachige und ressourcenarme NLP, Tokenisierung sowie Erklärbarkeit ab und spiegeln damit die Bandbreite der Forschungsarbeiten im gesamten Labor wider. Herzlichen Glückwunsch an alle Autoren!
Main Conference:
Tracing Stereotypes from Representation to Output in Multilingual LLMs
Ariun-Erdene Tumurchuluun, Yusser Al Ghussin, Pinzhen Chen, Josef van Genabith, Koel Dutta Chowdhury
Want Better Synthetic Data? Steer It: Activation Steering for Low-Resource Language Generation
Jan Cegin, Daniil Gurgurov, Yusser Al Ghussin, Simon Ostermann
FineWeb-CLaR: Culture, Language, and Region Annotations for Benchmark-Aligned Corpus Auditing
Yusser Al Ghussin, Eva Gavaller, Cristina España-Bonet, Josef van Genabith, Simon Ostermann
When Tokenization is Secretly Output Supervision
Tanja Baeumel, Josef van Genabith, Simon Ostermann
Findings:
Formatting Confounds in AI Text Detection: A Clever Hans Effect
Koel Dutta Chowdhury, Cristina España-Bonet, Josef van Genabith
Can Large Language Models Still Explain Themselves? Investigating the Impact of Quantization on Self-Explanations
Qianli Wang, Nils Feldhus, Pepa Atanasova, Fedor Splitt, Simon Ostermann, Sebastian Möller, Vera Schmitt
Macro: Enhancing Multilingual Counterfactual Explanations through Alignment-as-Preference Optimization
Yilong Wang, Qianli Wang, Bohao Chu, Yihong Liu, Jing Yang, Simon Ostermann
Separating Syntax from Language: A Mechanistic Account of Translation in Multilingual LLMs
Mikhail Sonkin, Tanja Baeumel, Daniil Gurgurov, Josef van Genabith, Simon Ostermann