News

Updates, announcements, and highlights from the lab.

• May–June 2026

  Three MSc students defend their theses

  Congratulations to three of the lab's MSc students on successfully defending their theses:

  • Justyna Micota — “Automatic Anonymization of Icelandic Medical Data: Named Entity Recognition for Unseen Text with Fine Tuning on Synthetic Corpora” (28 May 2026)
  • Thelma Rut Jóhannsdóttir — “Evaluating Overt and Covert Semantic Poisoning using Minimal Factual Sentence Injection” (29 May 2026)
  • Einar Örn Gissurarson — “Hugrún: Runtime-Governed Therapeutic Interaction for LLM-Based CBT Systems” (1 June 2026)

• May 2026

  New paper: The Icelandic Language Biobank

  “The Icelandic Language Biobank: Data Collection through a Clinical Analysis Platform” (Iris Nowenstein, Naizeth Núñez Macías, Gunnar Thor Örnólfsson, Stefán Ólafsson, Bryndís Bergþórsdóttir, Iðunn Kristínardóttir, and Hinrik Hafsteinsson) was presented at the RaPID-6 workshop (@MENTAL.ai) at LREC 2026.

• 2026

  HUMI Lab website launched

  Our new website is online, bringing together the lab's research areas, people, and publications in one place.

• 2025

  Paper on Icelandic legal-text summarization presented

  “Aligning Language Models for Icelandic Legal Text Summarization” (Þórir Hrafn Harðarson, Hrafn Loftsson, and Stefán Ólafsson) was presented at the Joint 25th Nordic Conference on Computational Linguistics (NoDaLiDa) and 11th Conference on Natural Language Processing, in Tallinn, Estonia. The work uses preference-based alignment to adapt large language models for summarizing Icelandic legal texts, improving the legal accuracy of generated summaries over standard fine-tuning while highlighting the difficulty of producing high-quality Icelandic in a specialized, low-resource domain. Read the paper.