Nejdl Wolfgang

Aparup Khatua, Wolfgang Nejdl

Social media deliberations allow to explore refugee-related is-sues. AI-based studies have investigated refugee issues mostly around a specific event and considered unimodal approaches. Contrarily, we have employed a multimodal architecture for probing the refugee journeys from their home to host nations. We draw insights from Arnold van Gennep's anthropological work 'Les Rites de Passage', which systematically analyzed an individual's transition from one group or society to another. Based on Gennep's separation-transition-incorporation framework, we have identified four phases of refugee journeys: Arrival of Refugees, Temporal stay at Asylums, Rehabilitation, and Integration of Refugees into the host nation. We collected 0.23 million multimodal tweets from April 2020 to March 2021 for testing this proposed frame-work. We find that a combination of transformer-based language models and state-of-the-art image recognition models, such as fusion of BERT+LSTM and InceptionV4, can out-perform unimodal models. Subsequently, to test the practical implication of our proposed model in real-time, we have considered 0.01 million multimodal tweets related to the 2022 Ukrainian refugee crisis. An F1-score of 71.88 % for this 2022 crisis confirms the generalizability of our proposed framework.

William Xion, Wolfgang Nejdl

Dense retrieval is a promising approach for acquiring relevant context or world knowledge in open-domain natural language processing tasks and is now widely used in information retrieval applications. However, recent reports claim a broad preference for text generated by large language models (LLMs). This bias is called "source bias", and it has been hypothesized that lower perplexity contributes to this effect. In this study, we revisit this claim by conducting a controlled evaluation to trace the emergence of such preferences across training stages and data sources. Using parallel human- and LLM-generated counterparts of the SciFact and Natural Questions (NQ320K) datasets, we compare unsupervised checkpoints with models fine-tuned using in-domain human text, in-domain LLM-generated text, and MS MARCO. Our results show the following: 1) Unsupervised retrievers do not exhibit a uniform pro-LLM preference. The direction and magnitude depend on the dataset. 2) Across the settings tested, supervised fine-tuning on MS MARCO consistently shifts the rankings toward LLM-generated text. 3) In-domain fine-tuning produces dataset-specific and inconsistent shifts in preference. 4) Fine-tuning on LLM-generated corpora induces a pronounced pro-LLM bias. Finally, a retriever-centric perplexity probe involving the reattachment of a language modeling head to the fine-tuned dense retriever encoder indicates agreement with relevance near chance, thereby weakening the explanatory power of perplexity. Our study demonstrates that source bias is a training-induced phenomenon rather than an inherent property of dense retrievers.

Alexander Blezinger, Wolfgang Nejdl, Ming Tang

Foundation models pretrained on large-scale histopathology data have found great success in various fields of computational pathology, but their impact on regressive biomarker prediction remains underexplored. In this work, we systematically evaluate histopathological foundation models for regression-based tasks, demonstrated through the prediction of homologous recombination deficiency (HRD) score - a critical biomarker for personalized cancer treatment. Within multiple instance learning frameworks, we extract patch-level features from whole slide images (WSI) using five state-of-the-art foundation models, and evaluate their impact compared to contrastive learning-based features. Models are trained to predict continuous HRD scores based on these extracted features across breast, endometrial, and lung cancer cohorts from two public medical data collections. Extensive experiments demonstrate that models trained on foundation model features consistently outperform the baseline in terms of predictive accuracy and generalization capabilities while exhibiting systematic differences among the foundation models. Additionally, we propose a distribution-based upsampling strategy to mitigate target imbalance in these datasets, significantly improving the recall and balanced accuracy for underrepresented but clinically important patient populations. Furthermore, we investigate the impact of different sampling strategies and instance bagsizes by ablation studies. Our results highlight the benefits of large-scale histopathological pretraining for more precise and transferable regressive biomarker prediction, showcasing its potential to advance AI-driven precision oncology.