News - Data Science Chair

We created two German-only decoder models, LLäMmlein 120M and 1B, from scratch!

Janna Omeliyanenko just gave a keynote on "Efficient Language Models as Continual Knowledge Bases" at the Workshop on Knowledge Base Construction from Pre-Trained Language Models at the ISWC 2024.

We're looking back on an amazing week of presentations and discussions at both the LWDA and KI conferences, plus all the great workshops! A big thanks to all the participants and helpers who made the conference a success and for the awesome chance to share ideas and knowledge.

In our work "Adapting Sequential Recommender Models to Content Recommendation in Chat Data using Non-Item Page-Models", we adapt sequential recommender models for content recommendation in twitch and multiple other chat systems by modeling messages as non-item pages with pretrained embeddings.

We are proud to announce that our chair will be presenting their research with three papers at the ECML PKDD 2024 conference in Vilnius.

Our paper proposes a novel adapter-based approach to integrate knowledge from knowledge graphs into large language models (LLMs).

We had a succesful NAACL 2024 in Mexico City! We showcased "SuperGLEBer" - the first comprehensive German language benchmark, earned a best paper honorable mention with our hierarchical classifier, and even presented the Roman Empire's take on model hallucinations!

In our paper, we use multi-agent systems to simulate labeled company data, allowing us to automatically detect and explain occupational fraud in otherwise unlabeled data.

We are happy to announce that our paper "GrINd: Grid Interpolation Network for Scattered Observations" got accepted for the Research Track at ECML-PKDD 2024! In our work we investigate if Fourier Interpolation can improve the prediction accuracy of physical systems in a low-resolution data regime on the Dynabench Benchmark.