Maria Maistro

Joakim Edin, Andreas Geert Motzfeldt, Casper L. Christensen et al.

Deep neural network predictions are notoriously difficult to interpret. Feature attribution methods aim to explain these predictions by identifying the contribution of each input feature. Faithfulness, often evaluated using the area over the perturbation curve (AOPC), reflects feature attributions' accuracy in describing the internal mechanisms of deep neural networks. However, many studies rely on AOPC to compare faithfulness across different models, which we show can lead to false conclusions about models' faithfulness. Specifically, we find that AOPC is sensitive to variations in the model, resulting in unreliable cross-model comparisons. Moreover, AOPC scores are difficult to interpret in isolation without knowing the model-specific lower and upper limits. To address these issues, we propose a normalization approach, Normalized AOPC (NAOPC), enabling consistent cross-model evaluations and more meaningful interpretation of individual scores. Our experiments demonstrate that this normalization can radically change AOPC results, questioning the conclusions of earlier studies and offering a more robust framework for assessing feature attribution faithfulness. Our code is available at https://github.com/JoakimEdin/naopc.

Theresia Veronika Rampisela, Maria Maistro, Tuukka Ruotsalo et al.

Individual user fairness is commonly understood as treating similar users similarly. In Recommender Systems (RSs), several evaluation measures exist for quantifying individual user fairness. These measures evaluate fairness via either: (i) the disparity in RS effectiveness scores regardless of user similarity, or (ii) the disparity in items recommended to similar users regardless of item relevance. Both disparity in recommendation effectiveness and user similarity are very important in fairness, yet no existing individual user fairness measure simultaneously accounts for both. In brief, current user fairness evaluation measures implement a largely incomplete definition of fairness. To fill this gap, we present Pairwise User unFairness (PUF), a novel evaluation measure of individual user fairness that considers both effectiveness disparity and user similarity. PUF is the only measure that can express this important distinction. We empirically validate that PUF does this consistently across 4 datasets and 7 rankers, and robustly when varying user similarity or effectiveness. In contrast, all other measures are either almost insensitive to effectiveness disparity or completely insensitive to user similarity. We contribute the first RS evaluation measure to reliably capture both user similarity and effectiveness in individual user fairness. Our code: https://github.com/theresiavr/PUF-individual-user-fairness-recsys.

Joakim Edin, Alexander Junge, Jakob D. Havtorn et al.

Medical coding is the task of assigning medical codes to clinical free-text documentation. Healthcare professionals manually assign such codes to track patient diagnoses and treatments. Automated medical coding can considerably alleviate this administrative burden. In this paper, we reproduce, compare, and analyze state-of-the-art automated medical coding machine learning models. We show that several models underperform due to weak configurations, poorly sampled train-test splits, and insufficient evaluation. In previous work, the macro F1 score has been calculated sub-optimally, and our correction doubles it. We contribute a revised model comparison using stratified sampling and identical experimental setups, including hyperparameters and decision boundary tuning. We analyze prediction errors to validate and falsify assumptions of previous works. The analysis confirms that all models struggle with rare codes, while long documents only have a negligible impact. Finally, we present the first comprehensive results on the newly released MIMIC-IV dataset using the reproduced models. We release our code, model parameters, and new MIMIC-III and MIMIC-IV training and evaluation pipelines to accommodate fair future comparisons.

Sara Vera Marjanović, Haeun Yu, Pepa Atanasova et al.

Knowledge-intensive language understanding tasks require Language Models (LMs) to integrate relevant context, mitigating their inherent weaknesses, such as incomplete or outdated knowledge. However, conflicting knowledge can be present in the LM's parameters, termed intra-memory conflict, which can affect a model's propensity to accept contextual knowledge. To study the effect of intra-memory conflict on an LM's ability to accept relevant context, we utilize two knowledge conflict measures and a novel dataset containing inherently conflicting data, DynamicQA. This dataset includes facts with a temporal dynamic nature where facts can change over time and disputable dynamic facts, which can change depending on the viewpoint. DynamicQA is the first to include real-world knowledge conflicts and provide context to study the link between the different types of knowledge conflicts. We also evaluate several measures on their ability to reflect the presence of intra-memory conflict: semantic entropy and a novel coherent persuasion score. With our extensive experiments, we verify that LMs exhibit a greater degree of intra-memory conflict with dynamic facts compared to facts that have a single truth value. Furthermore, we reveal that facts with intra-memory conflict are harder to update with context, suggesting that retrieval-augmented generation will struggle with the most commonly adapted facts.

Theresia Veronika Rampisela, Maria Maistro, Tuukka Ruotsalo et al.

The development of Artificial Intelligence (AI), including AI in Science (AIS), should be done following the principles of responsible AI. Progress in responsible AI is often quantified through evaluation metrics, yet there has been less work on assessing the robustness and reliability of the metrics themselves. We reflect on prior work that examines the robustness of fairness metrics for recommender systems as a type of AI application and summarise their key takeaways into a set of non-exhaustive guidelines for developing reliable metrics of responsible AI. Our guidelines apply to a broad spectrum of AI applications, including AIS.

Theresia Veronika Rampisela, Maria Maistro, Tuukka Ruotsalo et al.

Fairness in recommender systems (RSs) is commonly categorised into group fairness and individual fairness. However, there is no established scientific understanding of the relationship between the two fairness types, as prior work on both types has used different evaluation measures or evaluation objectives for each fairness type, thereby not allowing for a proper comparison of the two. As a result, it is currently not known how increasing one type of fairness may affect the other. To fill this gap, we study the relationship of group and individual fairness through a comprehensive comparison of evaluation measures that can be used for both fairness types. Our experiments with 8 runs across 3 datasets show that recommendations that are highly fair for groups can be very unfair for individuals. Our finding is novel and useful for RS practitioners aiming to improve the fairness of their systems. Our code is available at: https://github.com/theresiavr/stairway-to-fairness.

Joakim Edin, Róbert Csordás, Tuukka Ruotsalo et al.

Ensuring faithful interpretability in large language models is imperative for trustworthy and reliable AI. A key obstacle is self-repair, a phenomenon where networks compensate for reduced signal in one component by amplifying others, masking the true importance of the ablated component. While prior work attributes self-repair to layer normalization and back-up components that compensate for ablated components, we identify a novel form occurring within the attention mechanism, where softmax redistribution conceals the influence of important attention scores. This leads traditional ablation and gradient-based methods to underestimate the significance of all components contributing to these attention scores. We introduce Gradient Interaction Modifications (GIM), a technique that accounts for self-repair during backpropagation. Extensive experiments across multiple large language models (Gemma 2B/9B, LLAMA 1B/3B/8B, Qwen 1.5B/3B) and diverse tasks demonstrate that GIM significantly improves faithfulness over existing circuit identification and feature attribution methods. Our work is a significant step toward better understanding the inner mechanisms of LLMs, which is crucial for improving them and ensuring their safety. Our code is available at https://github.com/JoakimEdin/gim.

Pietro Tropeano, Maria Maistro, Tuukka Ruotsalo et al.

Language Model (LM) pruning compresses the model by removing weights, nodes, or other parts of its architecture. Typically, pruning focuses on the resulting efficiency gains at the cost of effectiveness. However, when looking at how individual data points are affected by pruning, it turns out that a particular subset of data points always bears most of the brunt (in terms of reduced accuracy) when pruning, but this effect goes unnoticed when reporting the mean accuracy of all data points. These data points are called PIEs and have been studied in image processing, but not in NLP. In a study of various NLP datasets, pruning methods, and levels of compression, we find that PIEs impact inference quality considerably, regardless of class frequency, and that BERT is more prone to this than BiLSTM. We also find that PIEs contain a high amount of data points that have the largest influence on how well the model generalises to unseen data. This means that when pruning, with seemingly moderate loss to accuracy across all data points, we in fact hurt tremendously those data points that matter the most. We trace what makes PIEs both hard and impactful to inference to their overall longer and more semantically complex text. These findings are novel and contribute to understanding how LMs are affected by pruning. The code is available at: https://github.com/pietrotrope/AsEasyAsPIE

Lovisa Hagström, Sara Vera Marjanović, Haeun Yu et al.

Retrieval-augmented generation (RAG) helps address the limitations of parametric knowledge embedded within a language model (LM). In real world settings, retrieved information can vary in complexity, yet most investigations of LM utilisation of context has been limited to synthetic text. We introduce DRUID (Dataset of Retrieved Unreliable, Insufficient and Difficult-to-understand contexts) with real-world queries and contexts manually annotated for stance. The dataset is based on the prototypical task of automated claim verification, for which automated retrieval of real-world evidence is crucial. We compare DRUID to synthetic datasets (CounterFact, ConflictQA) and find that artificial datasets often fail to represent the complexity and diversity of realistically retrieved context. We show that synthetic datasets exaggerate context characteristics rare in real retrieved data, which leads to inflated context utilisation results, as measured by our novel ACU score. Moreover, while previous work has mainly focused on singleton context characteristics to explain context utilisation, correlations between singleton context properties and ACU on DRUID are surprisingly small compared to other properties related to context source. Overall, our work underscores the need for real-world aligned context utilisation studies to represent and improve performance in real-world RAG settings.

Mihaela Rotar, Theresia Veronika Rampisela, Maria Maistro

Large Language Models (LLMs) can infer sensitive attributes such as gender or age from indirect cues like names and pronouns, potentially biasing recommendations. While several debiasing methods exist, they require access to the LLMs' weights, are computationally costly, and cannot be used by lay users. To address this gap, we investigate implicit biases in LLM Recommenders (LLMRecs) and explore whether prompt-based strategies can serve as a lightweight and easy-to-use debiasing approach. We contribute three bias-aware prompting strategies for LLMRecs. To our knowledge, this is the first study on prompt-based debiasing approaches in LLMRecs that focuses on group fairness for users. Our experiments with 3 LLMs, 4 prompt templates, 9 sensitive attribute values, and 2 datasets show that our proposed debiasing approach, which instructs an LLM to be fair, can improve fairness by up to 74% while retaining comparable effectiveness, but might overpromote specific demographic groups in some cases.

Tianyi Hu, Andrea Morales-Garzón, Jingyi Zheng et al.

In cross-cultural recipe adaptation, the goal is not only to ensure cultural appropriateness and retain the original dish's essence, but also to provide diverse options for various dietary needs and preferences. Retrieval Augmented Generation (RAG) is a promising approach, combining the retrieval of real recipes from the target cuisine for cultural adaptability with large language models (LLMs) for relevance. However, it remains unclear whether RAG can generate diverse adaptation results. Our analysis shows that RAG tends to overly rely on a limited portion of the context across generations, failing to produce diverse outputs even when provided with varied contextual inputs. This reveals a key limitation of RAG in creative tasks with multiple valid answers: it fails to leverage contextual diversity for generating varied responses. To address this issue, we propose CARRIAGE, a plug-and-play RAG framework for cross-cultural recipe adaptation that enhances diversity in both retrieval and context organization. To our knowledge, this is the first RAG framework that explicitly aims to generate highly diverse outputs to accommodate multiple user preferences. Our experiments show that CARRIAGE achieves Pareto efficiency in terms of diversity and quality of recipe adaptation compared to closed-book LLMs.

Joakim Edin, Maria Maistro, Lars Maaløe et al.

Electronic healthcare records are vital for patient safety as they document conditions, plans, and procedures in both free text and medical codes. Language models have significantly enhanced the processing of such records, streamlining workflows and reducing manual data entry, thereby saving healthcare providers significant resources. However, the black-box nature of these models often leaves healthcare professionals hesitant to trust them. State-of-the-art explainability methods increase model transparency but rely on human-annotated evidence spans, which are costly. In this study, we propose an approach to produce plausible and faithful explanations without needing such annotations. We demonstrate on the automated medical coding task that adversarial robustness training improves explanation plausibility and introduce AttInGrad, a new explanation method superior to previous ones. By combining both contributions in a fully unsupervised setup, we produce explanations of comparable quality, or better, to that of a supervised approach. We release our code and model weights.

Timo Breuer, Nicola Ferro, Maria Maistro et al.

In this work we introduce repro_eval - a tool for reactive reproducibility studies of system-oriented information retrieval (IR) experiments. The corresponding Python package provides IR researchers with measures for different levels of reproduction when evaluating their systems' outputs. By offering an easily extensible interface, we hope to stimulate common practices when conducting a reproducibility study of system-oriented IR experiments.

Lucas Chaves Lima, Dustin Brandon Wright, Isabelle Augenstein et al.

In this paper, we describe our participation in the TREC Health Misinformation Track 2020. We submitted $11$ runs to the Total Recall Task and 13 runs to the Ad Hoc task. Our approach consists of 3 steps: (1) we create an initial run with BM25 and RM3; (2) we estimate credibility and misinformation scores for the documents in the initial run; (3) we merge the relevance, credibility and misinformation scores to re-rank documents in the initial run. To estimate credibility scores, we implement a classifier which exploits features based on the content and the popularity of a document. To compute the misinformation score, we apply a stance detection approach with a pretrained Transformer language model. Finally, we use different approaches to merge scores: weighted average, the distance among score vectors and rank fusion.

Dongsheng Wang, Casper Hansen, Lucas Chaves Lima et al.

The state of the art in learning meaningful semantic representations of words is the Transformer model and its attention mechanisms. Simply put, the attention mechanisms learn to attend to specific parts of the input dispensing recurrence and convolutions. While some of the learned attention heads have been found to play linguistically interpretable roles, they can be redundant or prone to errors. We propose a method to guide the attention heads towards roles identified in prior work as important. We do this by defining role-specific masks to constrain the heads to attend to specific parts of the input, such that different heads are designed to play different roles. Experiments on text classification and machine translation using 7 different datasets show that our method outperforms competitive attention-based, CNN, and RNN baselines.

Lucas Chaves Lima, Casper Hansen, Christian Hansen et al.

This report describes the participation of two Danish universities, University of Copenhagen and Aalborg University, in the international search engine competition on COVID-19 (the 2020 TREC-COVID Challenge) organised by the U.S. National Institute of Standards and Technology (NIST) and its Text Retrieval Conference (TREC) division. The aim of the competition was to find the best search engine strategy for retrieving precise biomedical scientific information on COVID-19 from the largest, at that point in time, dataset of curated scientific literature on COVID-19 -- the COVID-19 Open Research Dataset (CORD-19). CORD-19 was the result of a call to action to the tech community by the U.S. White House in March 2020, and was shortly thereafter posted on Kaggle as an AI competition by the Allen Institute for AI, the Chan Zuckerberg Initiative, Georgetown University's Center for Security and Emerging Technology, Microsoft, and the National Library of Medicine at the US National Institutes of Health. CORD-19 contained over 200,000 scholarly articles (of which more than 100,000 were with full text) about COVID-19, SARS-CoV-2, and related coronaviruses, gathered from curated biomedical sources. The TREC-COVID challenge asked for the best way to (a) retrieve accurate and precise scientific information, in response to some queries formulated by biomedical experts, and (b) rank this information decreasingly by its relevance to the query. In this document, we describe the TREC-COVID competition setup, our participation to it, and our resulting reflections and lessons learned about the state-of-art technology when faced with the acute task of retrieving precise scientific information from a rapidly growing corpus of literature, in response to highly specialised queries, in the middle of a pandemic.

Timo Breuer, Nicola Ferro, Norbert Fuhr et al.

Replicability and reproducibility of experimental results are primary concerns in all the areas of science and IR is not an exception. Besides the problem of moving the field towards more reproducible experimental practices and protocols, we also face a severe methodological issue: we do not have any means to assess when reproduced is reproduced. Moreover, we lack any reproducibility-oriented dataset, which would allow us to develop such methods. To address these issues, we compare several measures to objectively quantify to what extent we have replicated or reproduced a system-oriented IR experiment. These measures operate at different levels of granularity, from the fine-grained comparison of ranked lists, to the more general comparison of the obtained effects and significant differences. Moreover, we also develop a reproducibility-oriented dataset, which allows us to validate our measures and which can also be used to develop future measures.