Research

We do methodological and applied NLP research, including (but not limited to): Low-Resource and Historical Languages, (Sampling-based) Decoding, Reproducibility & Benchmarking (of LLMs), NLP-Resources and Evaluation Strategies, Uncertainty Quantification

Publications

A complete list of our publications can be found here.
Read more

Recently accepted publications of our group can be found below:

  • Braun L, Aßenmacher M, Hohenadler M, Kauermann G (2026) Two Lawyers, Three Opinions: Can Output-Based LLM Uncertainty Find the Cases Where Experts Disagree? Accepted at the Third Workshop on Uncertainty-Aware NLP (EMNLP 2026)
  • Kandlinger M, Ma B, Haensch A-C, Aßenmacher M (2026) Do LLMs Give Consistent Opinions? Evaluating Response Reliability Under Varying Likert-Scale Formulations in Survey-Style MCQA Accepted at the 11th Workshop on Natural User-generated Text (EMNLP 2026)
  • Ding Y, Li M, Garces Arias E, Aßenmacher M, Heumann C, Zhang C (2026) Breaking the Likelihood Trap: Variance-Calibrated Modulation for Large Language Model Decoding Accepted at EMNLP 2026 Main Track
  • Zhang C, Wu S, Chen Y, Men Y, Fan G, Jiang Y, Aßenmacher M, Heumann C (2026) Explainable Coarse-to-Fine Text-Centric Oracle Bone Duplicates Discovery Accepted to EMNLP 2026 System Demonstrations
  • Garces Arias E, Sapargali N, Heumann C, Aßenmacher M (2026) The Truncation Blind Spot: How Decoding Strategies Systematically Exclude Human-Like Token Choices Accepted at INLG 2026
  • Gnadt K, Meidinger M, Aßenmacher M (2026) MUCnoHARM@GermEval Shared Task 2026: Retrieval-based In-Context Learning for Defamatory Offences, and Where It Falls Short Accepted at the GermEval Workshop on Harmful Content Detection
  • Leininger C, Veit H, Aßenmacher M, Andreas A (2026) Fairness Beyond Anonymization? Demographic Leakage in German LLM-Generated Résumés Accepted to [BIAS 2026] 6th Workshop on Bias and Fairness in AI