Hinrich Schütze · page 2
Papers
- CRAFT Your Dataset: Task-Specific Synthetic Dataset Generation Through Corpus Retrieval and Augmentation (2025 · 20 citations)
Fine-tuning LLMs for specialized tasks usually works better than prompting, but building large task-specific training sets requires costly manual annotation and domain expertise. - Active Learning for Coreference Resolution (2012 · 19 citations)
Supervised coreference resolution (CR) relies on costly hand-labeled data. Active learning (AL) can reduce that cost, but earlier attempts at AL for CR had failed. Uncertainty sampling performed no better than random selection, and the authors confirmed this in their own setup. They propose an AL scheme built on the… - How Transliterations Improve Crosslingual Alignment (2025 · 12 citations)
Multilingual pretrained language models (mPLMs) transfer knowledge less effectively between languages written in different scripts. - TransMI: A Framework to Create Strong Baselines from Multilingual Pretrained Language Models for Transliterated Data (2025 · 10 citations)
Writing related languages in one shared script can improve crosslingual transfer. However, existing mPLMs split romanized text into meaningless pieces, so earlier work usually had to pretrain new models or run adaptation training. The authors propose TransMI (Transliterate-Merge-Initialize), a procedure that requires… - Hybrid Human-LLM Corpus Construction and LLM Evaluation for Rare Linguistic Phenomena (2024 · 10 citations)
Construction Grammar holds that some meaning belongs to constructions rather than to individual words. The caused-motion construction (CMC) is a classic example: in it, a verb like "sneeze" takes on a motion sense and an extra argument. The authors ask whether current LLMs grasp this, and how to collect enough data… - Problem Solving Through Human–AI Preference-based Cooperation (2025 · 9 citations)
Generative AI systems, including LLMs, vision-language models and code models, still struggle with complex expert-domain tasks. - How Far Can Bias Go? Tracing Bias from Pre-Training Data to Alignment (2026)
This paper asks where gender-occupation stereotypes in LLMs come from, in particular whether imbalances in pre-training data carry over into what a model generates. - Frequency Matters: Pitch Accents and Information Status (2009)
The paper examines whether a word's information status (given or new) and the frequency of a pitch accent category affect how consistently speakers produce tokens of that accent.
This page in Markdown → https://md.research.elamur.ai/free/en/authors/hinrich-schutze/page/2/