Andreas Waldis

Tilman Beck, Andreas Waldis, Iryna Gurevych

Stance detection deals with identifying an author's stance towards a target. Most existing stance detection models are limited because they do not consider relevant contextual information which allows for inferring the stance correctly. Complementary context can be found in knowledge bases but integrating the context into pretrained language models is non-trivial due to the graph structure of standard knowledge bases. To overcome this, we explore an approach to integrate contextual information as text which allows for integrating contextual information from heterogeneous sources, such as structured knowledge sources and by prompting large language models. Our approach can outperform competitive baselines on a large and diverse stance detection benchmark in a cross-target setup, i.e. for targets unseen during training. We demonstrate that it is more robust to noisy context and can regularize for unwanted correlations between labels and target-specific vocabulary. Finally, it is independent of the pretrained language model in use.

Andreas Waldis, Yufang Hou, Iryna Gurevych

The advent of pre-trained Language Models (LMs) has markedly advanced natural language processing, but their efficacy in out-of-distribution (OOD) scenarios remains a significant challenge. Computational argumentation (CA), modeling human argumentation processes, is a field notably impacted by these challenges because complex annotation schemes and high annotation costs naturally lead to resources barely covering the multiplicity of available text sources and topics. Due to this data scarcity, generalization to data from uncovered covariant distributions is a common challenge for CA tasks like stance detection or argument classification. This work systematically assesses LMs' capabilities for such OOD scenarios. While previous work targets specific OOD types like topic shifts or OOD uniformly, we address three prevalent OOD scenarios in CA: topic shift, domain shift, and language shift. Our findings challenge the previously asserted general superiority of in-context learning (ICL) for OOD. We find that the efficacy of such learning paradigms varies with the type of OOD. Specifically, while ICL excels for domain shifts, prompt-based fine-tuning surpasses for topic shifts. To sum up, we navigate the heterogeneity of OOD scenarios in CA and empirically underscore the potential of base-sized LMs in overcoming these challenges.

Andreas Waldis, Joel Birrer, Anne Lauscher et al.

Gender-fair language, an evolving German linguistic variation, fosters inclusion by addressing all genders or using neutral forms. Nevertheless, there is a significant lack of resources to assess the impact of this linguistic shift on classification using language models (LMs), which are probably not trained on such variations. To address this gap, we present Lou, the first dataset featuring high-quality reformulations for German text classification covering seven tasks, like stance detection and toxicity classification. Evaluating 16 mono- and multi-lingual LMs on Lou shows that gender-fair language substantially impacts predictions by flipping labels, reducing certainty, and altering attention patterns. However, existing evaluations remain valid, as LM rankings of original and reformulated instances do not significantly differ. While we offer initial insights on the effect on German text classification, the findings likely apply to other languages, as consistent patterns were observed in multi-lingual and English LMs.

Neele Falk, Andreas Waldis, Iryna Gurevych

Argument retrieval is the task of finding relevant arguments for a given query. While existing approaches rely solely on the semantic alignment of queries and arguments, this first shared task on perspective argument retrieval incorporates perspectives during retrieval, accounting for latent influences in argumentation. We present a novel multilingual dataset covering demographic and socio-cultural (socio) variables, such as age, gender, and political attitude, representing minority and majority groups in society. We distinguish between three scenarios to explore how retrieval systems consider explicitly (in both query and corpus) and implicitly (only in query) formulated perspectives. This paper provides an overview of this shared task and summarizes the results of the six submitted systems. We find substantial challenges in incorporating perspectivism, especially when aiming for personalization based solely on the text of arguments without explicitly providing socio profiles. Moreover, retrieval systems tend to be biased towards the majority group but partially mitigate bias for the female gender. While we bootstrap perspective argument retrieval, further research is essential to optimize retrieval systems to facilitate personalization and reduce polarization.

Benjamin Schiller, Johannes Daxenberger, Andreas Waldis et al.

The task of Argument Mining, that is extracting and classifying argument components for a specific topic from large document sources, is an inherently difficult task for machine learning models and humans alike, as large Argument Mining datasets are rare and recognition of argument components requires expert knowledge. The task becomes even more difficult if it also involves stance detection of retrieved arguments. In this work, we investigate the effect of Argument Mining dataset composition in few- and zero-shot settings. Our findings show that, while fine-tuning is mandatory to achieve acceptable model performance, using carefully composed training samples and reducing the training sample size by up to almost 90% can still yield 95% of the maximum performance. This gain is consistent across three Argument Mining tasks on three different datasets. We also publish a new dataset for future benchmarking.

Stefan Erben, Andreas Waldis

The long-standing problem of spam and fraudulent messages in the comment sections of Instagram pages in the financial sector claims new victims every day. Instagram's current spam filter proves inadequate, and existing research approaches are primarily confined to theoretical concepts. Practical implementations with evaluated results are missing. To solve this problem, we propose ScamSpot, a comprehensive system that includes a browser extension, a fine-tuned BERT model and a REST API. This approach ensures public accessibility of our results for Instagram users using the Chrome browser. Furthermore, we conduct a data annotation study, shedding light on the reasons and causes of the problem and evaluate the system through user feedback and comparison with existing models. ScamSpot is an open-source project and is publicly available at https://scamspot.github.io/.

Andreas Waldis, Leshem Choshen, Yufang Hou et al.

Instructions trigger a production-centered mechanism in language models. Through a cognitively inspired lens that separates language processing and production, we reveal this mechanism as an asymmetry between the two stages by probing task-specific information layer-wise across five binary judgment tasks. Specifically, we measure how instruction tokens shape information both when sample tokens, the input under evaluation, are processed and when output tokens are produced. Across prompting variations, task-specific information in sample tokens remains largely stable and correlates only weakly with behavior, whereas the same information in output tokens varies substantially and correlates strongly with behavior. Attention-based interventions confirm this pattern causally: blocking instruction flow to all subsequent tokens reduces both behavior and information in output tokens, whereas blocking it only to sample tokens has minimal effect on either. The asymmetry generalizes across model families and tasks, and becomes sharper with model scale and instruction-tuning, both of which disproportionately affect the production stage. Our findings suggest that understanding model capabilities requires jointly assessing internals and behavior, while decomposing the internal perspective by token position to distinguish the processing of input tokens from the production of output tokens.

Andreas Waldis, Yotam Perlitz, Leshem Choshen et al. · ibm-research

We introduce Holmes, a new benchmark designed to assess language models (LMs) linguistic competence - their unconscious understanding of linguistic phenomena. Specifically, we use classifier-based probing to examine LMs' internal representations regarding distinct linguistic phenomena (e.g., part-of-speech tagging). As a result, we meet recent calls to disentangle LMs' linguistic competence from other cognitive abilities, such as following instructions in prompting-based evaluations. Composing Holmes, we review over 270 probing studies and include more than 200 datasets to assess syntax, morphology, semantics, reasoning, and discourse phenomena. Analyzing over 50 LMs reveals that, aligned with known trends, their linguistic competence correlates with model size. However, surprisingly, model architecture and instruction tuning also significantly influence performance, particularly in morphology and syntax. Finally, we propose FlashHolmes, a streamlined version that reduces the computation load while maintaining high-ranking precision.

Andreas Waldis, Yufang Hou, Iryna Gurevych

Pre-trained language models (LMs) perform well in In-Topic setups, where training and testing data come from the same topics. However, they face challenges in Cross-Topic scenarios where testing data is derived from distinct topics -- such as Gun Control. This study analyzes various LMs with three probing-based experiments to shed light on the reasons behind the In- vs. Cross-Topic generalization gap. Thereby, we demonstrate, for the first time, that generalization gaps and the robustness of the embedding space vary significantly across LMs. Additionally, we assess larger LMs and underscore the relevance of our analysis for recent models. Overall, diverse pre-training objectives, architectural regularization, or data deduplication contribute to more robust LMs and diminish generalization gaps. Our research contributes to a deeper understanding and comparison of language models across different generalization scenarios.

Andreas Waldis, Vagrant Gautam, Anne Lauscher et al.

We introduce aligned probing, a novel interpretability framework that aligns the behavior of language models (LMs), based on their outputs, and their internal representations (internals). Using this framework, we examine over 20 OLMo, Llama, and Mistral models, bridging behavioral and internal perspectives for toxicity for the first time. Our results show that LMs strongly encode information about the toxicity level of inputs and subsequent outputs, particularly in lower layers. Focusing on how unique LMs differ offers both correlative and causal evidence that they generate less toxic output when strongly encoding information about the input toxicity. We also highlight the heterogeneity of toxicity, as model behavior and internals vary across unique attributes such as Threat. Finally, four case studies analyzing detoxification, multi-prompt evaluations, model quantization, and pre-training dynamics underline the practical impact of aligned probing with further concrete insights. Our findings contribute to a more holistic understanding of LMs, both within and beyond the context of toxicity.

Yutaro Sigris, Andreas Waldis

Merging methods combine the weights of multiple language models (LMs) to leverage their capacities, such as for domain adaptation. While existing studies investigate merged models from a solely behavioral perspective, we offer the first comprehensive view by assessing and connecting their behavior and internals. We present a novel evaluation pipeline that first merges multiple parent LMs, and then evaluates the merged models in comparison to the initial ones based on their behavior on downstream tasks, like MMLU, and the internal encoded linguistic competence. We showcase this pipeline by assessing the merging of instruction fine-tuned with math- and code-adapted LMs from the Qwen2.5 family. Our results show that merging methods impacts behavior and internals differently. While the performance of merged models is typically between that of the two parent models, their encoded information about linguistic phenomena, particularly in morphology and syntax, can surpass the parent models. Moreover, we find weak ranking correlation between this behavior and internal evaluation. With our pipeline and initial results, we emphasize the need for more comprehensive evaluations of model merging methods to gain a faithful understanding of their capabilities and reliability, beyond potential superficial behavioral advances.

Andreas Waldis, Luca Mazzola

Entity Recognition (ER) within a text is a fundamental exercise in Natural Language Processing, enabling further depending tasks such as Knowledge Extraction, Text Summarisation, or Keyphrase Extraction. An entity consists of single words or of a consecutive sequence of terms, constituting the basic building blocks for communication. Mainstream ER approaches are mainly limited to flat structures, concentrating on the outermost entities while ignoring the inner ones. This paper introduces a partly-layered network architecture that deals with the complexity of overlapping and nested cases. The proposed architecture consists of two parts: (1) a shared Sequence Layer and (2) a stacked component with multiple Tagging Layers. The adoption of such an architecture has the advantage of preventing overfit to a specific word-length, thus maintaining performance for longer entities despite their lower frequency. To verify the proposed architecture's effectiveness, we train and evaluate this architecture to recognise two kinds of entities - Concepts (CR) and Named Entities (NER). Our approach achieves state-of-the-art NER performances, while it outperforms previous CR approaches. Considering these promising results, we see the possibility to evolve the architecture for other cases such as the extraction of events or the detection of argumentative components.