Modern Agentic AI systems plan, reason, and act across multiple steps, creating execution patterns that are difficult to interpret. Existing observability platforms track prompt I/O and operational metrics but require manual inspection of traces to reconstruct structure and reasoning. We present AgentGraph, which converts execution logs into interactive knowledge graphs and actionable insights. Nodes represent agents, tasks, tools, data inputs/outputs, and humans, while typed edges capture relations such as inputs consumed, tasks delegated or sequenced, tools required or used, outputs produced and delivered, and interventions from agents or humans. Each graph element links to its exact trace span, ensuring verifiability. Building on this representation, AgentGraph enables two analyses: qualitative trace-grounded failure detection and optimisation recommendations, and quantitative robustness evaluation via perturbation testing and causal attribution.

Kenza Benkirane, Jackie Kay, Maria Perez-Ortiz. Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers). 2025.

With the advancement and utility of Artificial Intelligence (AI), personalising education to a global population could be a cornerstone of new educational systems in the future. This work presents the PEEKC dataset and the TrueLearn Python library, which contains a dataset and a series of online learner state models that are essential to facilitate research on learner engagement modelling. TrueLearn family of models was designed following the "open learner" concept, using humanly-intuitive user representations. This family of scalable, online models also help end-users visualise the learner models, which may in the future facilitate user interaction with their models/recommenders. The extensive documentation and coding examples make the library highly accessible to both machine learning developers and educational data mining and learning analytics practitioners. The experiments show the utility of both the dataset and the library with predictive performance significantly exceeding comparative baseline models. The dataset contains a large amount of AI-related educational videos, which are of interest for building and validating AI-specific educational recommenders.

In an effort to reduce pesticide use, agronomists and computer scientists have joined forces to develop site-specific weed detection and classification systems. These systems aim to recognize and locate weed species within a crop field, using precision equipment to apply required herbicides timely and only where needed, with the objective of reducing the sprayable surface required to eliminate the given weed and protect the crop, with both economic and environmental benefits. Yet, with climate change on the rise, common weeds are expected to undergo some changes to adapt to their environment, possibly with new or invasive weeds spreading to areas where they did not exist before. These changes (often morphological) as well as new invasions need to be taken into account by future classifiers and detection algorithms to ensure system robustness and adaptation to new habitats/climate dynamics. This paper proposes a set of experiments evaluating the use of transfer learning and zero-shot learning for weed classification using our novel TomatoWeeds dataset. Residual networks of variable depth, pretrained on the Imagenet and/or DeepWeeds datasets were evaluated. A ResNet50 pretrained on both datasets and fine-tuned on the TomatoWeeds dataset performed best, returning a holdout set accuracy of 77.8%, showing the advantageous use of transfer learning in this domain. Zero-shot learning, using both embeddings of images and morphological and habitat text-based descriptions, is implemented to test the ability of machine learning pipelines of recognising unseen classes at test time (which may arise e.g. due to changing climate dynamics), a learning task in which the field (and our experiments) are still far from satisfactory results. Further research could benefit from larger weed-specific datasets for transfer learning as well as deeper network architectures to improve model performance. The projection-based ZSL could also benefit from larger datasets and new zero-shot learning architectures in hope that unseen classes are accurately projected.

Prior research has shown how ‘content preview tools’ improve speed and accuracy of user relevance judgements across different information retrieval tasks. This paper describes a novel user interface tool, the Content Flow Bar, designed to allow users to quickly identify relevant fragments within informational videos to facilitate browsing, through a cognitively augmented form of navigation. It achieves this by providing semantic “snippets” that enable the user to rapidly scan through video content. The tool provides visually-appealing pop-ups that appear in a time series bar at the bottom of each video, allowing to see in advance and at a glance how topics evolve in the content. We conducted a user study to evaluate how the tool changes the users search experience in video retrieval, as well as how it supports exploration and information seeking. The user questionnaire revealed that participants found the Content Flow Bar helpful and enjoyable for finding relevant information in videos. The interaction logs of the user study, where participants interacted with the tool for completing two informational tasks, showed that it holds promise for enhancing discoverability of content both across and within videos. This discovered potential could leverage a new generation of navigation tools in search and information retrieval.

Increasing popularity of high-dynamic-range (HDR) image and video content brings the need for metrics that could predict the severity of image impairments as seen on displays of different brightness levels and dynamic range. Such metrics should be trained and validated on a sufficiently large subjective image quality dataset to ensure robust performance. As the existing HDR quality datasets are limited in size, we created a Unified Photometric Image Quality dataset (UPIQ) with over 4000 images by realigning and merging existing HDR and standard-dynamic-range (SDR) datasets. The realigned quality scores share the same unified quality scale across all datasets. Such realignment was achieved by collecting additional cross-dataset quality comparisons and re-scaling data with a psychometric scaling method. Images in the proposed dataset are represented in absolute photometric and colorimetric units, corresponding to light emitted from a display. We use the new dataset to retrain existing HDR metrics and show that the dataset is sufficiently large for training deep architectures. We show the utility of the dataset on brightness aware image compression.

This paper presents an empirical study regarding training probabilistic neural networks using training objectives derived from PAC-Bayes bounds. In the context of probabilistic neural networks, the output of training is a probability distribution over network weights. We present two training objectives, used here for the first time in connection with training neural networks. These two training objectives are derived from tight PAC-Bayes bounds. We also re-implement a previously used training objective based on a classical PAC-Bayes bound, to compare the properties of the predictors learned using the different training objectives. We compute risk certificates for the learnt predictors, based on part of the data used to learn the predictors. We further experiment with different types of priors on the weights (both data-free and data-dependent priors) and neural network architectures. Our experiments on MNIST and CIFAR-10 show that our training methods produce competitive test set errors and non-vacuous risk bounds with much tighter values than previous results in the literature, showing promise not only to guide the learning algorithm through bounding the risk but also for model selection. These observations suggest that the methods studied here might be good candidates for self-certified learning, in the sense of using the whole data set for learning a predictor and certifying its risk on any unseen data (from the same distribution as the training data) potentially without the need for holding out test data.

One of the most ambitious use cases of computer-assisted learning is to build a recommendation system for lifelong learning. Most recommender algorithms exploit similarities between content and users, overseeing the necessity to leverage sensible learning trajectories for the learner. Lifelong learning thus presents unique challenges, requiring scalable and transparent models that can account for learner knowledge and content novelty simultaneously, while also retaining accurate learners representations for long periods of time. We attempt to build a novel educational recommender, that relies on an integrative approach combining multiple drivers of learners engagement. Our first step towards this goal is TrueLearn, which models content novelty and background knowledge of learners and achieves promising performance while retaining a human interpretable learner model.

The recent advances in computer-assisted learning systems and the availability of open educational resources today promise a pathway to providing cost-efficient high-quality education to large masses of learners. One of the most ambitious use cases of computer-assisted learning is to build a lifelong learning recommendation system. Unlike short-term courses, lifelong learning presents unique challenges, requiring sophisticated recommendation models that account for a wide range of factors such as background knowledge of learners or novelty of the material while effectively maintaining knowledge states of masses of learners for significantly longer periods of time (ideally, a lifetime). This work presents the foundations towards building a dynamic, scalable and transparent recommendation system for education, modelling learner's knowledge from implicit data in the form of engagement with open educational resources. We i) use a text ontology based on Wikipedia to automatically extract knowledge components of educational resources and, ii) propose a set of online Bayesian strategies inspired by the well-known areas of item response theory and knowledge tracing. Our proposal, TrueLearn, focuses on recommendations for which the learner has enough background knowledge (so they are able to understand and learn from the material), and the material has enough novelty that would help the learner improve their knowledge about the subject and keep them engaged. We further construct a large open educational video lectures dataset and test the performance of the proposed algorithms, which show clear promise towards building an effective educational recommendation system.

Pairwise comparison data arise in many domains with subjective assessment experiments, for example in image and video quality assessment. In these experiments observers are asked to express a preference between two conditions. However, many pairwise comparison protocols require a large number of comparisons to infer accurate scores, which may be unfeasible when each comparison is time-consuming (e.g. videos) or expensive (e.g. medical imaging). This motivates the use of an active sampling algorithm that chooses only the most informative pairs for comparison. In this paper we propose ASAP, an active sampling algorithm based on approximate message passing and expected information gain maximization. Unlike most existing methods, which rely on partial updates of the posterior distribution, we are able to perform full updates and therefore much improve the accuracy of the inferred scores. The algorithm relies on three techniques for reducing computational cost: inference based on approximate message passing, selective evaluations of the information gain, and selecting pairs in a batch that forms a minimum spanning tree of the inverse of information gain. We demonstrate, with real and synthetic data, that ASAP offers the highest accuracy of inferred scores compared to the existing methods. We also provide an open-source GPU implementation of ASAP for large-scale experiments.

Capturing and effectively utilising user states and goals is becoming a timely challenge for successfully leveraging intelligent and usercentric systems in differentweb search and data mining applications. Examples of such systems are conversational agents, intelligent assistants, educational and contextual information retrieval systems, recommender/match-making systems and advertising systems, all of which rely on identifying the user state in order to provide the most relevant information and assist users in achieving their goals. There has been, however, limited work towards building such state-aware intelligent learning mechanisms. Hence, devising information systems that can keep track of the user's state has been listed as one of the grand challenges to be tackled in the next few years [1]. It is thus timely to organize a workshop that re-visits the problem of designing and evaluating state-aware and user-centric systems, ensuring that the community (spanning academic and industrial backgrounds) works together to tackle these challenges.

The goal of psychometric scaling is the quantification of perceptual experiences, understanding the relationship between an external stimulus, the internal representation and the response. In this paper, we propose a probabilistic framework to fuse the outcome of different psychophysical experimental protocols, namely rating and pairwise comparisons experiments. Such a method can be used for merging existing datasets of subjective nature and for experiments in which both measurements are collected. We analyze and compare the outcomes of both types of experimental protocols in terms of time and accuracy in a set of simulations and experiments with benchmark and real-world image quality assessment datasets, showing the necessity of scaling and the advantages of each protocol and mixing. Although most of our examples focus on image quality assessment, our findings generalize to any other subjective quality-of-experience task.

Data augmentation is rapidly gaining attention in machine learning. Synthetic data can be generated by simple transformations or through the data distribution. In the latter case, the main challenge is to estimate the label associated to new synthetic patterns. This paper studies the effect of generating synthetic data by convex combination of patterns and the use of these as unsupervised information in a semi-supervised learning framework with support vector machines, avoiding thus the need to label synthetic examples. We perform experiments on a total of 53 binary classification datasets. Our results show that this type of data over-sampling supports the well-known cluster assumption in semi-supervised learning, showing outstanding results for small high-dimensional datasets and imbalanced learning problems.

Encoding images in a visually lossless manner helps to achieve the best trade-off between image compression performance and quality and so that compression artifacts are invisible to the majority of users. Visually lossless encoding can often be achieved by manually adjusting compression quality parameters of existing lossy compression methods, such as JPEG or WebP. But the required compression quality parameter can also be determined automatically using visibility metrics. However, creating an accurate visibility metric is challenging because of the complexity of the human visual system and the effort needed to collect the required data. In this paper, we investigate how to train an accurate visibility metric for visually lossless compression from a relatively small dataset. Our experiments show that prediction error can be reduced by 40% compared with the state-of-theart, and that our proposed method can save between 25%-75% of storage space compared with the default quality parameter used in commercial software. We demonstrate how the visibility metric can be used for visually lossless image compression and for benchmarking image compression encoders.

Ordinal regression, also named ordinal classification, studies classification problems where there exist a natural order between class labels. This structured order of the labels is crucial in all steps of the learning process in order to take full advantage of the data. ORCA (Ordinal Regression and Classification Algorithms) is a Matlab/Octave framework that implements and integrates different ordinal classification algorithms and specifically designed performance metrics. The framework simplifies the task of experimental comparison to a great extent, allowing the user to: (i) describe experiments by simple configuration files; (ii) automatically run different data partitions; (iii) parallelize the executions; (iv) generate a variety of performance reports and (v) include new algorithms by using its intuitive interface. Source code, binaries, documentation, descriptions and links to data sets and tutorials (including examples of educational purpose) are available at https://github.com/ayrna/orca.

In this paper, we propose a trained perceptually transform for quality assessment of high dynamic range (HDR) images and video. The transform is used to convert absolute luminance values found in HDR images into perceptually uniform units, which can be used with any standard-dynamic-range metric. The new transform is derived by fitting the parameters of a previously proposed perceptual encoding function to 4 different HDR subjective quality assessment datasets using Bayesian optimization. The new transform combined with a simple peak signal-to-noise ratio measure achieves better prediction performance in cross-dataset validation than existing transforms. We provide Matlab code for our metric11https://github.com/ynyCL/T-PT-metric.

Weather and atmospheric patterns are often persistent. The simplest weather forecasting method is the so-called persistence model, which assumes that the future state of a system will be similar (or equal) to the present state. Machine learning (ML) models are widely used in different weather forecasting applications, but they need to be compared to the persistence model to analyse whether they provide a competitive solution to the problem at hand. In this paper, we devise a new model for predicting low-visibility in airports using the concepts of mixture of experts. Visibility level is coded as two different ordered categorical variables: cloud height and runway visual height. The underlying system in this application is stagnant approximately in 90% of the cases, and standard ML models fail to improve on the performance of the persistence model. Because of this, instead of trying to simply beat the persistence model using ML, we use this persistence as a baseline and learn an ordinal neural network model that refines its results by focusing on learning weather fluctuations. The results show that the proposal outperforms persistence and other ordinal autoregressive models, especially for longer time horizon predictions and for the runway visual height variable.

TID2013 is a subjective image quality assessment dataset with a wide range of distortion types and over 3000 images. The dataset has proven to be a challenging test for objective quality metrics. The dataset mean opinion scores were obtained by collecting pairwise comparison judgments using the Swiss tournament system, and averaging votes of observers. However, this approach differs from the usual analysis of multiple pairwise comparisons, which involves psychometric scaling of the comparison data using either Thurstone or Bradley-Terry models. In this paper we investigate how quality scores change when they are computed using such psychometric scaling instead of averaging vote counts. In order to properly scale TID2013 quality scores, we conduct four additional experiments of two different types, which we found necessary to produce a common quality scale: comparisons with reference images, and cross-content comparisons. We demonstrate on a fifth validation experiment that the two additional types of comparisons are necessary and in conjunction with psychometric scaling improve the consistency of quality scores, especially across images depicting different contents.

Ordinal input variables are common in many supervised and unsupervised machine learning problems. We focus on ordinal classification problems, where the target variable is also categorical and ordinal. In order to represent categorical input variables for measuring distances or applying continuous mapping functions, they have to be transformed to numeric values. This paper evaluates five different methods to do so. Two of them are commonly applied by practitioners, the first one based on binarising the ordinal input variable using standard indicator variables (NomBin), and the second one based on directly mapping each category to a consecutive natural number (Num). Furthermore, three novel proposals are evaluated in this paper: 1) an ordinal binarisation based on considering the order of the input variable (OrdBin), 2) the analysis of pairwise distances between input patterns to recover the latent variable generating the ordinal one (NumLVR), and 3) the refinement of the standard numeric transformation by recovering the distance between sets of patterns of consecutive categories (NumCDR). A thorough empirical evaluation is made, considering 12 datasets, 5 performance metrics and 4 classifiers (2 of them of nominal nature and 2 of ordinal nature). The results show that the Nom-Bin representation method leads to the worst results, and that both Num and NumCDR methods obtain very good performance, although NumCDR results are consistently better for almost all performance metrics and classifiers considered.

Melanoma is a type of cancer that usually occurs on the skin. Early detection is crucial for ensuring five-year survival (which varies between 15% and 99% depending on the melanoma stage). Melanoma severity is typically diagnosed by invasive methods (e.g. a biopsy). In this paper, we propose an alternative system combining image analysis and machine learning for detecting melanoma presence and severity. The 86 features selected consider the shape, colour, pigment network and texture of the melanoma. As opposed to previous studies that have focused on distinguishing melanoma and non-melanoma images, our work considers a finer-grain classification problem using five categories: benign lesions and 4 different stages of melanoma. The dataset presents two main characteristics that are approached by specific machine learning methods: 1) the classes representing melanoma severity follow a natural order, and 2) the dataset is imbalanced, where benign lesions clearly outnumber melanoma ones. Different nominal and ordinal classifiers are considered, one of them being based on an ordinal cascade decomposition method. The cascade method is shown to obtain good performance for all classes, while respecting and exploiting the order information. Moreover, we explore the alternative of applying a class balancing technique, presenting good synergy with the ordinal and nominal methods.

Ordinal regression problems are those machine learning problems where the objective is to classify patterns using a categorical scale which shows a natural order between the labels. Many real-world applications present this labelling structure and that has increased the number of methods and algorithms developed over the last years in this field. Although ordinal regression can be faced using standard nominal classification techniques, there are several algorithms which can specifically benefit from the ordering information. Therefore, this paper is aimed at reviewing the state of the art on these techniques and proposing a taxonomy based on how the models are constructed to take the order into account. Furthermore, a thorough experimental study is proposed to check if the use of the order information improves the performance of the models obtained, considering some of the approaches within the taxonomy. The results confirm that ordering information benefits ordinal models improving their accuracy and the closeness of the predictions to actual targets in the ordinal scale.

The imbalanced nature of some real-world data is one of the current challenges for machine learning researchers. One common approach oversamples the minority class through convex combination of its patterns. We explore the general idea of synthetic oversampling in the feature space induced by a kernel function (as opposed to input space). If the kernel function matches the underlying problem, the classes will be linearly separable and synthetically generated patterns will lie on the minority class region. Since the feature space is not directly accessible, we use the empirical feature space (EFS) (a Euclidean space isomorphic to the feature space) for oversampling purposes. The proposed method is framed in the context of support vector machines, where the imbalanced data sets can pose a serious hindrance. The idea is investigated in three scenarios: 1) oversampling in the full and reduced-rank EFSs; 2) a kernel learning technique maximizing the data class separation to study the influence of the feature space structure (implicitly defined by the kernel function); and 3) a unified framework for preferential oversampling that spans some of the previous approaches in the literature. We support our investigation with extensive experiments over 50 imbalanced data sets.

The classification of patterns into naturally ordered labels is referred to as ordinal regression or ordinal classification. Usually, this classification setting is by nature highly imbalanced, because there are classes in the problem that are a priori more probable than others. Although standard over-sampling methods can improve the classification of minority classes in ordinal classification, they tend to introduce severe errors in terms of the ordinal label scale, given that they do not take the ordering into account. A specific ordinal over-sampling method is developed in this paper for the first time in order to improve the performance of machine learning classifiers. The method proposed includes ordinal information by approaching over-sampling from a graph-based perspective. The results presented in this paper show the good synergy of a popular ordinal regression method (a reformulation of support vector machines) with the graph-based proposed algorithms, and the possibility of improving both the classification and the ordering of minority classes. A cost-sensitive version of the ordinal regression method is also introduced and compared with the over-sampling proposals, showing in general lower performance for minority classes.

The classification of patterns into naturally ordered labels is referred to as ordinal regression. This paper proposes an ensemble methodology specifically adapted to this type of problem, which is based on computing different classification tasks through the formulation of different order hypotheses. Every single model is trained in order to distinguish between one given class (k) and all the remaining ones, while grouping them in those classes with a rank lower than k , and those with a rank higher than k. Therefore, it can be considered as a reformulation of the well-known one-versus-all scheme. The base algorithm for the ensemble could be any threshold (or even probabilistic) method, such as the ones selected in this paper: kernel discriminant analysis, support vector machines and logistic regression (LR) (all reformulated to deal with ordinal regression problems). The method is seen to be competitive when compared with other state-of-the-art methodologies (both ordinal and nominal), by using six measures and a total of 15 ordinal datasets. Furthermore, an additional set of experiments is used to study the potential scalability and interpretability of the proposed method when using LR as base methodology for the ensemble.

This paper proposes a novel algorithm for ordinal classification based on combining ensemble techniques and discriminant analysis. The proposal is applied to a real application of liver transplantation, where the objective is to predict survival rates of the graft. Ordinal classification is used for this problem because the classes are defined by the following temporal order: 1) failure of the graft before the first 15 days after transplantation, 2) failure between 15 days and 3 months, 3) failure between 3 months and one year, and 4) no failure presented (taking into account that the patient follow-up is up to one year after the transplantation). When compared to other state-of-the-art classifiers like AdaBoost, EBC(SVM) or KDLOR, the proposed algorithm is shown to be competitive. The models obtained could allow medical experts to predict survival rates without knowing exactly the number of days the transplanted organ survived.

In this paper we apply and test a recent ordinal algorithm for classification (Kernel Discriminant Learning Ordinal Regression, KDLOR), in order to recognize a group of geographically close spatial units with a similar prevalence pattern significantly high (or low), which are called hot-spots (or cold-spots). Different spatial analysis techniques have been used for studying geographical distribution of a specific illness in mental health-care because it could be useful to organize the spatial distribution of health-care services. Ordinal classification is used in this problem because the classes are: spatial unit with depression, spatial unit which could present depression and spatial unit where there is not depression. It is shown that the proposed method is capable of preserving the rank of data classes in a projected data space for this database. In comparison to other standard methods like C4.5, SVMRank, Adaboost, and MLP nominal classifiers, the proposed KDLOR algorithm is shown to be competitive.