Seminars - page 2
The DIG team holds a seminar about every two weeks with speakers either from the team, or invited.
You can add the seminars to your calendar with this ics file, and get emails about future seminars by subscribing to our mailing-list.
If you would like to present your work at our seminar, please contact Nils.
Upcoming Seminars
Toward Adaptive Intelligence
Tuesday, October 13, 2026 11:45, 4A301
Nilesh Verma (University of Waikato)
Real-world data arrives as an endless stream whose distribution shifts over time. Models that work today may degrade tomorrow, and the human effort required to repeatedly re-tune them does not scale. This talk presents a line of work on adaptive intelligence, where systems select, configure, and revise their own learning pipelines as the data changes. I begin with AutoML for non-stationary data streams, covering online pipeline search, automated drift and outlier handling, and meta-learning for data streams. I then turn to tabular foundation models and ask what happens when in-context learners encounter concept drift. Along the way, I introduce TuiML, an open-source MCP-native ML runtime that enables agents and researchers to run, benchmark, and empirically study machine learning workflows.
From Observations to Explanations: Exploring Methodologies for Identifying Relevant Actual Causes in Complex Systems
Tuesday, October 27, 2026 11:45, 4A301
Samuel Reyd
Complex systems such as multi-agent and cyber-physical systems exhibit emergent behaviors that are hard to understand and explain. Causal models offer a principled framework for this, but building one for a complex system is already difficult: the relevant scope, abstraction scale, and variables are not given in advance. Moreover, such models capture general causation, i.e., which factors influence a class of outcomes, whereas a user asking “why did this happen?” needs an actual cause: the specific facts responsible for a specific event. Actual causation is conceptually and computationally demanding; finding actual causes is intractable in general, and not all causes are equally informative. Producing a useful explanation therefore requires three steps: modeling the system causally, identifying its actual causes, and filtering them for relevance. This model-identify-filter pipeline forms the backbone of the thesis. This thesis first reviews the challenges of causal modeling in complex adaptive systems and examines the choice of abstraction scale through causal emergence, showing that emergence detection depends heavily on the sampling distribution used. It then addresses the identification of actual causes, proposing a data-based method for Markovian multivariate systems and approximate search algorithms with adjustable precision that recover the full set of actual causes in general SCMs, released as an open-source Python module. Finally, it presents a general, tunable method for selecting relevant causes according to user and context, and develops a normality criterion to produce more relevant explanations. Overall, the thesis connects formal causal models, scalable computation, and human-centred explanation into a coherent framework for context-aware explanation in complex systems.
Past Seminars
Automatic Evaluation of Human-Written and Machine-Generated Text
Tuesday, January 13, 2026 11:45, 4A301
Yanzhu Guo
With the rapid expansion of digital content, automatic evaluation of textual information has become increasingly crucial. My research addresses the challenge of evaluating and enhancing the quality of both human-written and machine-generated texts. Beginning with human-written texts, we develop an argument extraction system to evaluate the substantiation of scientific peer reviews, providing insights into the quality of peer review in NLP conferences in recent years. Additionally, we assess the factuality of abstractive summarization datasets and propose a data refinement approach that enhances model performance while reducing computational demands. For machine-generated texts, we focus on the underexplored aspects of diversity and naturalness. We introduce a suite of metrics for measuring linguistic diversity and conduct a systematic evaluation across state-of-the-art LLMs, exploring how development and deployment choices influence their output diversity. We also investigate the impact of training LLMs on synthetic data produced by earlier models, demonstrating that recursive training loops lead to diminishing diversity. Finally, we explore the naturalness of multilingual LLMs, uncovering an English-centric bias and proposing an alignment method to mitigate it. These contributions advance evaluation methodologies for natural language generation and provide insights into the interaction between evaluation metrics, dataset quality and language model performance.
Contextual knowledge representation for neurosymbolic Artificial Intelligence reasoning
Thursday, December 11, 2025 12:15, 4A301
Simon Coumes
The field of Knowledge Representation and Reasoning is concerned with the representation of information about reality in a form that is both human-readable and machine-processable. It has been a part of artificial intelligence since its inception, and has produced many important formalisms and systems. One key aspect of knowledge is the context in which it is expressed. This has been identified early on in the field and matches with our common experience: understanding a statement or judging its validity often require to know in what context it was meant. Historically, there has been some work aiming at producing logics implementing a general notion of context. None of them saw a lot of adoption, in part because they lack either sufficient expressive power or because they were not sufficiently usable.
SPECTRA: Faster Large Language Model Inference with Optimized Internal and External Speculation
Tuesday, December 09, 2025 11:45, 1D19
Le-Minh Nguyen (Japan Advanced Institute of Science and Technology)
Inference with modern Large Language Models (LLMs) is both computationally intensive and time-consuming. While speculative decoding has emerged as a promising solution, existing approaches face key limitations. Training-based methods require the development of a draft model, which is often difficult to obtain and lacks generalizability. On the other hand, training-free methods provide only modest speedup improvements. In this work, we introduce SPECTRA — a novel framework designed to accelerate LLM inference without requiring any additional training or modifications to the original LLM. SPECTRA incorporates two new techniques that efficiently leverage both internal and external speculation, each independently outperforming corresponding state-of-the-art (SOTA) methods. When combined, these techniques deliver up to a 4.08× speedup across a variety of benchmarks and LLM architectures, significantly surpassing existing training-free approaches. The implementation of SPECTRA is publicly available. Biography: Le-Minh Nguyen is currently a Professor of the School of Information Science and the director of the Interpretable AI Center at JAIST. He leads the Machine Learning and Natural Language Understanding Laboratory at JAIST. He is currently taking his sabbatical at Imperial College London, UK (Until April 2026). His research interests include machine learning & deep learning, natural language processing, legal text processing, and explainable AI. He serves as an action editor of TACL (a leading journal in NLP), a board member of VLSP (Vietnamese language and speech processing), and an editorial board member of AI &Law, Journal of Natural Language Processing (Cambridge). He is a steering committee of Juris-informatics (Jurisin) in Japan – a research area that studies legal issues from informatics.
Mining Expressive Cross-Table Dependencies in Relational Databases
Tuesday, December 02, 2025 11:45, 4A125
François Amat
This thesis addresses the gap between what relational database schemas declare and the richer set of cross-table rules that actually govern real-world data. It introduces MATILDA, the first deterministic system capable of mining expressive first-order tuple-generating dependencies (FO-TGDs) with multi-atom heads, existential witnesses, and recursion directly from arbitrary relational databases, using principled, database-native definitions of support and confidence. MATILDA uncovers hidden business rules, workflow constraints, and multi-relation regularities that schemas alone cannot capture, while ensuring reproducible results through canonicalized search and tractable pruning guided by a constraint graph. To understand when simpler formalisms suffice, the thesis also presents MAHILDA, a relational Horn-rule baseline equipped with disjoint semantics to prevent self-justifying recursion. Overall, the work shows that expressive rule mining on realistic databases is both feasible and insightful, enabling more systematic, explainable, and schema-grounded analyses of complex relational data.
Entity Linking and Relation Extraction for Historical Italian Texts: Challenges and Potential Solutions
Tuesday, October 28, 2025 11:45, 4A125
Cristian Santini (University of Macerata)
Entity Linking and Relation Extraction enable the automatic identification of named entities mentioned in texts, along with their relationships, by connecting them to external knowledge graphs such as Wikidata. While these techniques work well on modern documents, applying them to historical texts presents significant challenges due to the diachronic evolution of language and limited resources for training computational models. This seminar presents recent work on developing methods and datasets for processing historical Italian texts. It will discuss the creation of a new benchmark dataset extracted from digital scholarly editions covering two centuries of Italian literary and political writing. The talk will then present approaches that enhance entity disambiguation by incorporating temporal and contextual information from external Wikidata. Finally, it will detail a method for automatically constructing knowledge graphs from historical correspondence that combines multiple language models in sequence, demonstrating how these technologies can facilitate the exploration and understanding of historical archives without requiring extensive manual annotation or model training.
FLORA: Unsupervised Knowledge Graph Alignment by Fuzzy Logic
Tuesday, October 21, 2025 11:45, 4A301
Yiwen Peng & Fabian Suchanek
Knowledge graph alignment is the task of matching equivalent entities (that is, instances and classes) and relations across two knowledge graphs. Most existing methods focus on pure entity-level alignment, computing the similarity of entities in some embedding space. They lack interpretable reasoning and need training data to work. To solve these issues, we introduce FLORA, a simple yet effective method that (1) is unsupervised, i.e., does not require training data, (2) provides a holistic alignment for entities and relations iteratively, (3) is based on fuzzy logic and thus delivers interpretable results, (4) provably converges, (5) allows dangling entities, i.e., entities without a counterpart in the other KG, and (6) achieves state-of-the-art results on major benchmarks.
Data- and knowledge-driven approaches for step-by-step guidance to differential diagnosis
Tuesday, October 07, 2025 11:45, 4A301
Adrien Coulet (INRIA)
Diagnosis guidelines provide recommendations based on expert consensus that cover the majority of the population, but often overlook patients with uncommon conditions or multiple morbidities. We will present and compare two alternative approaches that provide a step-by-step guidance to the differential diagnosis of anemia and lupus. The first approach relies on reinforcement learning and observational data. The second on large langage models and domain knowledge.
Meaning Representations and Reasoning in the Age of Large Language Models
Tuesday, September 30, 2025 11:45, 3A301
Zacchary Sadeddine
This thesis explores how to make large language models (LLMs) more reliable and transparent in their reasoning. It first examines around fifteen societal issues related to these models, such as disinformation or user overreliance, and then investigates symbolic structures from linguistics and how they can be used to improve the performance and transparency of LLMs. It presents VANESSA, a reasoning neuro-symbolic system that combines the power of neural models with the rigor of symbolic reasoning, achieving performance comparable to LLMs while remaining transparent. Finally, it addresses the problem of verifying LLM outputs by introducing a step-by-step verification benchmark, paving the way for more interpretable, controllable and trustworthy artificial intelligence systems.
Robust Knowledge Graph Cleaning
Tuesday, May 27, 2025 11:45, 4A301
Maximilian Egger
Data quality is needed to properly and reliably use the information represented in the dataset. The increasing volume of data renders data preparation and cleaning increasingly difficult. Additionally, more diverse types of data structures for databases, like graphs, get used and need to be handled differently. This leads to the necessity of robust methods to increase data integrity, scalable approaches for finding and fixing errors, and local-oriented algorithms that can be used to pinpoint attention where needed.
Synthesis & Augmentation of Tabular Data In the Age of Foundation Models
Tuesday, May 13, 2025 11:45, 4A301
Nikola Simidjievski
Foundation models - large pre-trained performant models - have shown remarkable success in applications that predominately focus on vision, language, and sound data. On the other hand, tabular data - one of the most prevalent data modalities in many critical domains of business, science, and healthcare - has seen limited benefits from these advances. Tabular data poses unique challenges that relate to heterogeneity, dimensionality, and scarcity as well as lack of explicit symmetries, implicit structures and incomplete prior knowledge – all of which have limiting effects on how we construct, train and apply/transfer large models for tabular data.