https://klaudiath.github.io/blog/Recent content in Blog on Klaudia-Doris ThellmannHugo -- 0.155.3en-usFri, 15 May 2026 00:00:00 +0000https://klaudiath.github.io/blog/acl-2026-tqe-paper/Fri, 15 May 2026 00:00:00 +0000https://klaudiath.github.io/blog/acl-2026-tqe-paper/<p>In July 2026, I will present our TQE paper at ACL. This draft is a placeholder for the final conference post.</p>https://klaudiath.github.io/blog/lrec-2026-eu20-benchmark-qa/Fri, 15 May 2026 00:00:00 +0000https://klaudiath.github.io/blog/lrec-2026-eu20-benchmark-qa/<p>In May 2026, I presented our work <strong>&ldquo;Diagnosing Translated Benchmarks: An Automated Quality Assurance Study of the EU20 Benchmark Suite&rdquo;</strong> at LREC in Palma, Mallorca. What struck me most across the conference was a recurring theme: multilingual evaluation is becoming less about simply translating English benchmarks, and more about diagnosing whether our evaluation data is valid in the first place.</p> <img src="images/lrec_poster.png" alt="Presenting the paper at LREC" class="blog-image-medium"> <p>This post collects the main idea of our paper, the most relevant work I saw at the conference, and a few personal takeaways for researchers working on multilingual LLM evaluation.</p>