2025Activity​​ reportTeamLINKMEDIA

Machine Learning for Multimedia‌​‌ Material.

  Many approaches are​​ used to extract relevant​​​‌ information from multimedia material,‌ ranging from very low-level‌​‌ to higher-level descriptions (classes,​​ captions, ...). That diversity​​​‌ of information is produced‌ by algorithms that have‌​‌ varying degrees of supervision.​​ Lately, fully supervised approaches​​​‌ based on deep learning‌ proved to outperform most‌​‌ older techniques. This is​​ particularly true for the​​​‌ latest developments of Recurrent‌ Neural Networkds (RNN, such‌​‌ as LSTMs) or convolutional​​ neural network (CNNs) for​​​‌ images that reach excellent‌ performance  39. Linkmedia‌​‌ contributes to advancing the​​ state of the art​​​‌ in computing representations for‌ multimedia material by investigating‌​‌ the topics listed below.​​ Some of them go​​​‌ beyond the very processing‌ of multimedia material as‌​‌ they also question the​​ fundamentals of machine learning​​​‌ procedures when applied to‌ multimedia.

• Learning from few‌​‌ samples/weak supervisions. CNNs and​​ RNNs need large collections​​​‌ of carefully annotated data.‌ They are not fitted‌​‌ for analyzing datasets where​​ few examples per category​​​‌ are available or only‌ cheap image-level labels are‌​‌ provided. Linkmedia investigates low-shot,​​ semi-supervised and weakly supervised​​​‌ learning processes: Augmenting scarce‌ training data by automatically‌​‌ propagating labels  42,​​ or transferring what was​​​‌ learned on few very‌ well annotated samples to‌​‌ allow the precise processing​​ of poorly annotated data​​​‌  51. Note that‌ this context also applies‌​‌ to the processing of​​ heritage collections (paintings, illuminated​​​‌ manuscripts, ...) that strongly‌ differ from contemporary natural‌​‌ images. Not only annotations​​ are scarce, but the​​​‌ learning processes must cope‌ with material departing from‌​‌ what standard CNNs deal​​ with, as classes such​​​‌ as "planes", "cars", etc,‌ are irrelevant in this‌​‌ case.
• Ubiquitous Training. NN​​ (CNNs, LSTMs) are mainstream​​​‌ for producing representations suited‌ for high-quality classification. Their‌​‌ training phase is ubiquitous​​ because the same representations​​​‌ can be used for‌ tasks that go beyond‌​‌ classification, such as retrieval,​​ few-shot, meta- and incremental​​​‌ learning, all boiling down‌ to some form of‌​‌ metric learning. We demonstrated​​​‌ that this ubiquitous training​ is relatively simpler  42​‌ yet as powerful as​​ ad-hoc strategies fitting specific​​​‌ tasks  56. We​ study the properties and​‌ the limitations of this​​ ubiquitous training by casting​​​‌ metric learning as a​ classification problem.
• Beyond static​‌ learning. Multimedia collections​​ are by nature continuously​​​‌ growing, and ML processes​ must adapt. It is​‌ not conceivable to re-train​​ a full new model​​​‌ at every change, but​ rather to support continuous​‌ training and/or allowing categories​​ to evolve as the​​​‌ time goes by. New​ classes may be defined​‌ from only very few​​ samples, which links this​​​‌ need for dynamicity to​ the low-shot learning problem​‌ discussed here. Furthermore, active​​ learning strategies determining which​​​‌ is the next sample​ to use to best​‌ improve classification must be​​ considered to alleviate the​​​‌ annotation cost and the​ re-training process  46.​‌ Eventually, the learning process​​ may need to manage​​​‌ an extremely large number​ of classes, up to​‌ millions. In this case,​​ there is a unique​​​‌ opportunity of blending the​ expertise of Linkmedia on​‌ large scale indexing and​​ retrieval with deep learning.​​​‌ Base classes can either​ be "summarized" e.g. as​‌ a multi-modal distribution, or​​ their entire training set​​​‌ can be made accessible​ as an external associative​‌ memory  62.
• Learning​​ and lightweight architectures. Multimedia​​​‌ is everywhere, it can​ be captured and processed​‌ on the mobile devices​​ of users. It is​​​‌ necessary to study the​ design of lightweight ML​‌ architectures for mobile and​​ embedded vision applications. Inspired​​​‌ by  66, we​ study the savings from​‌ quantizing hyper-parameters, pruning connections​​ or other approximations, observing​​​‌ the trade-off between the​ footprint of the learning​‌ and the quality of​​ the inference. Once strategy​​​‌ of choice is progressive​ learning which early aborts​‌ when confident enough  47​​.
• Multimodal embeddings. We​​​‌ pursue pioneering work of​ Linkmedia on multimodal embedding,​‌ i.e., representing multiple modalities​​ or information sources in​​​‌ a single embedded space​  60, 59,​‌ 61. Two main​​ directions are explored: exploiting​​​‌ adversarial architectures (GANs) for​ embedding via translation from​‌ one modality to another,​​ extending initial work in​​​‌  61 to highly heterogeneous​ content; combining and constraining​‌ word and RDF graph​​ embeddings to facilitate entity​​​‌ linking and explanation of​ lexical co-occurrences  36.​‌
• Accountability of ML processes.​​ ML processes achieve excellent​​​‌ results but it is​ mandatory to verify that​‌ accuracy results from having​​ determined an adequate problem​​​‌ representation, and not from​ being abused by artifacts​‌ in the data. Linkmedia​​ designs procedures for at​​​‌ least explaining and possibly​ interpreting and understanding what​‌ the models have learned.​​ We consider heat-maps materializing​​​‌ which input (pixels, words)​ have the most importance​‌ in the decisions  55​​, Taylor decompositions to​​​‌ observe the individual contributions​ of each relevance scores​‌ or estimating LID 23​​ as a surrogate for​​​‌ accounting for the smoothness​ of the space.
• Extracting​‌ information. ML is good​​ at extracting features from​​​‌ multimedia material, facilitating subsequent​ classification, indexing, or mining​‌ procedures. Linkmedia designs extraction​​ processes for identifying parts​​ in the images  52​​​‌, 53, relationships‌ between the various objects‌​‌ that are represented in​​ images  29, learning​​​‌ to localizing objects in‌ images with only weak,‌​‌ image-level supervision  55 or​​ fine-grained semantic information in​​​‌ texts  34. One‌ technique of choice is‌​‌ to rely on generative​​ adversarial networks (GAN) for​​​‌ learning low-level representations. These‌ representations can e.g. be‌​‌ based on the analysis​​ of density  65,​​​‌ shading, albedo, depth, etc.‌
• Learning representations for time‌​‌ evolving multimedia material. Video​​ and audio are time​​​‌ evolving material, and processing‌ them requests to take‌​‌ their time line into​​ account. In  48,​​​‌ 32 we demonstrated how‌ shapelets can be used‌​‌ to transform time series​​ into time-free high-dimensional vectors,​​​‌ preserving however similarities between‌ time series. Representing time‌​‌ series in a metric​​ space improves clustering, retrieval,​​​‌ indexing, metric learning, semi-supervised‌ learning and many other‌​‌ machine learning related tasks.​​ Research directions include adding​​​‌ localization information to the‌ shapelets, fine-tuning them to‌​‌ best fit the task​​ in which they are​​​‌ used as well as‌ designing hierarchical representations.

Adversarial‌​‌ Machine Learning.

  Systems based​​ on ML take more​​​‌ and more decisions on‌ our behalf, and maliciously‌​‌ influencing these decisions by​​ crafting adversarial multimedia material​​​‌ is a potential source‌ of dangers: a small‌​‌ amount of carefully crafted​​ noise imperceptibly added to​​​‌ images corrupts classification and/or‌ recognition. This can naturally‌​‌ impact the insight users​​ get on the multimedia​​​‌ collection they work with,‌ leading to taking erroneous‌​‌ decisions for example.

This​​ adversarial phenomenon is not​​​‌ particular to deep learning,‌ and can be observed‌​‌ even when using other​​ ML approaches  28.​​​‌ Furthermore, it has been‌ demonstrated that adversarial samples‌​‌ generalize very well across​​ classifiers, architectures, training sets.​​​‌ The reasons explaining why‌ such tiny content modifications‌​‌ succeed in producing severe​​ errors are still not​​​‌ well understood.

We are‌ left with little choice:‌​‌ we must gain a​​ better understanding of the​​​‌ weaknesses of ML processes,‌ and in particular of‌​‌ deep learning. We must​​ understand why attacks are​​​‌ possible as well as‌ discover mechanisms protecting ML‌​‌ against adversarial attacks (with​​ a special emphasis on​​​‌ convolutional neural networks). Some‌ initial contributions have started‌​‌ exploring such research directions,​​ mainly focusing on images​​​‌ and computer vision problems.‌ Very little has been‌​‌ done for understanding adversarial​​ ML from a multimedia​​​‌ perspective  33.

Linkmedia‌ is in a unique‌​‌ position to throw at​​ this problem new perspectives,​​​‌ by experimenting with other‌ modalities, used in isolation‌​‌ one another, as well​​ as experimenting with true​​​‌ multimodal inputs. This is‌ very challenging, and far‌​‌ more complicated and interesting​​ than just observing adversarial​​​‌ ML from a computer‌ vision perspective. No one‌​‌ clearly knows what is​​ at stake with adversarial​​​‌ audio samples, adversarial video‌ sequences, adversarial ASR, adversarial‌​‌ NLP, adversarial OCR, all​​ this being often part​​​‌ of a sophisticated multimedia‌ processing pipeline.

Our ambition‌​‌ is to lead the​​ way for initiating investigations​​​‌ where the full diversity‌ of modalities we are‌​‌ used to work with​​​‌ in multimedia are considered​ from a perspective of​‌ adversarial attacks and defenses,​​ both at learning and​​​‌ test time. In addition​ to what is described​‌ above, and in order​​ to trust the multimedia​​​‌ material we analyze and/or​ the algorithms that are​‌ at play, Linkmedia investigates​​ the following topics:

• Beyond​​​‌ classification. Most contributions in​ relation with adversarial ML​‌ focus on classification tasks.​​ We started investigating the​​​‌ impact of adversarial techniques​ on more diverse tasks​‌ such as retrieval  22​​. This problem is​​​‌ related to the very​ nature of euclidean spaces​‌ where distances and neighborhoods​​ can all be altered.​​​‌ Designing defensive mechanisms is​ a natural companion work.​‌
• Detecting false information. We​​ carry-on with earlier pioneering​​​‌ work of Linkmedia on​ false information detection in​‌ social media. Unlike traditional​​ approaches in image forensics​​​‌  37, we build​ on our expertise in​‌ content-based information retrieval to​​ take advantage of the​​​‌ contextual information available in​ databases or on the​‌ web to identify out-of-context​​ use of text or​​​‌ images which contributed to​ creating a false information​‌  49.
• Deep fakes.​​ Progress in deep ML​​​‌ and GANs allow systems​ to generate realistic images​‌ and are able to​​ craft audio and video​​​‌ of existing people saying​ or doing things they​‌ never said or did​​ 45. Gaining in​​​‌ sophistication, these machine learning-based​ "deep fakes" will eventually​‌ be almost indistinguishable from​​ real documents, making their​​​‌ detection/rebutting very hard. Linkmedia​ develops deep learning based​‌ counter-measures to identify such​​ modern forgeries. We also​​​‌ carry on with making​ use of external data​‌ in a provenance filtering​​ perspective  54 in order​​​‌ to debunk such deep​ fakes.
• Distributions, frontiers, smoothness,​‌ outliers. Many factors that​​ can possibly explain the​​​‌ adversarial nature of some​ samples are in relation​‌ with their distribution in​​ space which strongly differs​​​‌ from the distribution of​ natural, genuine, non adversarial​‌ samples. We are investigating​​ the use of various​​​‌ information theoretical tools that​ facilitate observing distributions, how​‌ they differ, how far​​ adversarial samples are from​​​‌ benign manifolds, how smooth​ is the feature space,​‌ etc. In addition, we​​ are designing original adversarial​​​‌ attacks and develop detection​ and curating mechanisms  23​‌.

Multimedia Knowledge Extraction.​​

  Information obtained from collections​​​‌ via computer ran processes​ is not the only​‌ thing that needs to​​ be represented. Humans are​​​‌ in the loop, and​ they gradually improve their​‌ level of understanding of​​ the content and nature​​​‌ of the multimedia collection.​ Discovering knowledge and getting​‌ insight is involving multiple​​ people across a long​​​‌ period of time, and​ what each understands, concludes​‌ and discovers must be​​ recorded and made available​​​‌ to others. Collaboratively inspecting​ collections is crucial. Ontologies​‌ are an often preferred​​ mechanism for modeling what​​​‌ is inside a collection,​ but this is probably​‌ limitative and narrow.

Linkmedia​​ is concerned with making​​​‌ use of existing strategies​ in relation with ontologies​‌ and knowledge bases. In​​ addition, Linkmedia uses mechanisms​​​‌ allowing to materialize the​ knowledge gradually acquired by​‌ humans and that might​​ be subsequently used either​​ by other humans or​​​‌ by computers in order‌ to better and more‌​‌ precisely analyze collections. This​​ line of work is​​​‌ instantiated at the core‌ of the iCODA project‌​‌ Linkmedia coordinates.

We are​​ therefore concerned with:

• Multimedia​​​‌ analysis and ontologies. We‌ develop approaches for linking‌​‌ multimedia content to entities​​ in ontologies for text​​​‌ and images, building on‌ results in multimodal embedding‌​‌ to cast entity linking​​ into a nearest neighbor​​​‌ search problem in a‌ high-dimensional joint embedding of‌​‌ content and entities 59​​. We also investigate​​​‌ the use of ontological‌ knowledge to facilitate information‌​‌ extraction from content 36​​.
• Explainability and accountability​​​‌ in information extraction. In‌ relation with ontologies and‌​‌ entity linking, we develop​​ innovative approaches to explain​​​‌ statistical relations found in‌ data, in particular lexical‌​‌ or entity co-occurrences in​​ textual data, for example​​​‌ using embeddings constrained with‌ translation properties of RDF‌​‌ knowledge or path-based explanation​​ within RDF graphs. We​​​‌ also work on confidence‌ measures in entity linking‌​‌ and information extraction, studying​​ how the notions of​​​‌ confidence and information source‌ can be accounted for‌​‌ in knowledge basis and​​ used in human-centric collaborative​​​‌ exploration of collections.
• Dynamic‌ evolution of models for‌​‌ information extraction. In interactive​​ exploration and information extraction,​​​‌ e.g., on cultural or‌ educational material, knowledge progressively‌​‌ evolves as the process​​ goes on, requiring on-the-fly​​​‌ design of new models‌ for content-based information extractors‌​‌ from very few examples,​​ as well as continuous​​​‌ adaptation of the models.‌ Combining in a seamless‌​‌ way low-shot, active and​​ incremental learning techniques is​​​‌ a key issue that‌ we investigate to enable‌​‌ this dynamic mechanisms on​​ selected applications.

Searching.

  Search engines​ must run similarity searches​‌ very efficiently. High-dimensional indexing​​ techniques therefore play a​​​‌ central role. Yet, recent​ contributions in ML suggest​‌ to revisit indexing in​​ order to adapt to​​​‌ the specific properties of​ modern features describing contents.​‌

• Advanced scalable indexing. High-dimensional​​ indexing is one of​​​‌ the foundations of Linkmedia​. Modern features extracted​‌ from the multimedia material​​ with the most recent​​​‌ ML techniques shall be​ indexed as well. This,​‌ however, poses a series​​ of difficulties due to​​​‌ the dimensionality of these​ features, their possible sparsity,​‌ the complex metrics in​​ use, the task in​​​‌ which they are involved​ (instance search, $k$-nn,​‌ class prototype identification, manifold​​ search  42, time​​​‌ series retrieval, ...). Furthermore,​ truly large datasets require​‌ involving sketching  26,​​ secondary storage and/or distribution​​​‌  25, 24,​ alleviating the explosion of​‌ the number of features​​ to consider due to​​​‌ their local nature or​ other innovative methods  41​‌, all introducing complexities.​​ Last, indexing multimodal embedded​​​‌ spaces poses a new​ series of challenges.
• Improving​‌ quality. Scalable indexing techniques​​ are approximate, and what​​​‌ they return typically includes​ a fair amount of​‌ false positives. Linkmedia works​​ on improving the quality​​​‌ of the results returned​ by indexing techniques. Approaches​‌ taking into account neighborhoods​​  35, manifold structures​​​‌ instead of pure distance​ based similarities  42 must​‌ be extended to cope​​ with advanced indexing in​​​‌ order to enhance quality.​ This includes feature selection​‌ based on intrinsic dimensionality​​ estimation  23.
• Dynamic​​​‌ indexing. Feature collections grow,​ and it is not​‌ an option to fully​​ reindex from scratch an​​​‌ updated collection. This trivially​ applies to the features​‌ directly extracted from the​​ media items, but also​​ to the base class​​​‌ prototypes that can evolve‌ due to the non-static‌​‌ nature of learning processes.​​ Linkmedia will continue investigating​​​‌ what is at stake‌ when designing dynamic indexing‌​‌ strategies.

Navigating.

  Navigating a​​ multimedia collection is very​​​‌ central to its understanding.‌ It differs from searching‌​‌ as navigation is not​​ driven by any specific​​​‌ query. Rather, it is‌ mostly driven by the‌​‌ relationships that various documents​​ have one another. Relationships​​​‌ are supported by the‌ links between documents and/or‌​‌ parts of documents. Links​​ rely on semantic similarity,​​​‌ depicting the fact that‌ two documents share information‌​‌ on the same topic.​​ But other aspects than​​​‌ semantics are also at‌ stake, e.g., time with‌​‌ the dates of creation​​ of the documents or​​​‌ geography with mentions or‌ appearance in documents of‌​‌ some geographical landmarks or​​ with geo-tagged data.

In​​​‌ multimedia collections, links can‌ be either implicit or‌​‌ explicit, the latter being​​ much easier to use​​​‌ for navigation. An example‌ of an implicit link‌​‌ can be the name​​ of someone existing in​​​‌ several different news articles;‌ we, as humans, create‌​‌ a mental link between​​ them. In some cases,​​​‌ the computer misses such‌ configurations, leaving such links‌​‌ implicit. Implicit links are​​ subject to human interpretation,​​​‌ hence they are sometimes‌ hard to identify for‌​‌ any automatic analysis process.​​ Implicit links not being​​​‌ materialized, they can therefore‌ hardly be used for‌​‌ navigation or faceted search.​​ Explicit links can typically​​​‌ be seen as hyperlinks,‌ established either by content‌​‌ providers or, more aligned​​ with Linkmedia, automatically​​​‌ determined from content analysis.‌ Entity linking (linking content‌​‌ to an entity referenced​​ in a knowledge base)​​​‌ is a good example‌ of the creation of‌​‌ explicit links. Semantic similarity​​ links, as investigated in​​​‌ the LIMAH project and‌ as considered in the‌​‌ search and hyperlinking task​​ at MediaEval and TRECVid,​​​‌ are also prototypical links‌ that can be made‌​‌ explicit for navigation. Pursuing​​ work, we investigate two​​​‌ main issues:

• Improving multimodal‌ content-based linking. We exploit‌​‌ achievements in entity linking​​ to go beyond lexical​​​‌ or lexico-visual similarity and‌ to provide semantic links‌​‌ that are easy to​​ interpret for humans; carrying​​​‌ on, we work on‌ link characterization, in search‌​‌ of mechanisms addressing link​​ explainability (i.e., what is​​​‌ the nature of the‌ link), for instance using‌​‌ attention models so as​​ to focus on the​​​‌ common parts of two‌ documents or using natural‌​‌ language generation; a final​​ topic that we address​​​‌ is that of linking‌ textual content to external‌​‌ data sources in the​​ field of journalism, e.g.,​​​‌ leveraging topic models and‌ cue phrases along with‌​‌ a short description of​​ the external sources.
• Dynamicity​​​‌ and user-adaptation. One difficulty‌ for explicit link creation‌​‌ is that links are​​ often suited for one​​​‌ particular usage but not‌ for another, thus requiring‌​‌ creating new links for​​ each intended use; whereas​​​‌ link creation cannot be‌ done online because of‌​‌ its computational cost, the​​ alternative is to generate​​​‌ (almost) all possible links‌ and provide users with‌​‌ selection mechanisms enabling personalization​​​‌ and user-adaptation in the​ exploration process; we design​‌ such strategies and investigate​​ their impact on exploration​​​‌ tasks in search of​ a good trade-off between​‌ performance (few high-quality links)​​ and genericity.

Summarizing.

  Multimedia​​​‌ collections contain far too​ much information to allow​‌ any easy comprehension. It​​ is mandatory to have​​​‌ facilities to aggregate and​ summarize a large body​‌ on information into a​​ compact, concise and meaningful​​​‌ representation facilitating getting insight.​ Current technology suggests that​‌ multimedia content aggregation and​​ story-telling are two complementary​​​‌ ways to provide users​ with such higher-level views.​‌ Yet, very few studies​​ already investigated these issues.​​​‌ Recently, video or image​ captioning  63, 58​‌ have been seen as​​ a way to summarize​​​‌ visual content, opening the​ door to state-of-the-art multi-document​‌ text summarization  38 with​​ text as a pivot​​​‌ modality. Automatic story-telling has​ been addressed for highly​‌ specific types of content,​​ namely TV series  30​​​‌ and news  50,​ 57, but still​‌ need a leap forward​​ to be mostly automated,​​​‌ e.g., using constraint-based approaches​ for summarization  27,​‌ 57.

Furthermore, not​​ only the original multimedia​​​‌ material has to be​ summarized, but the knowledge​‌ acquired from its analysis​​ is also to summarize.​​​‌ It is important to​ be able to produce​‌ high-level views of the​​ relationships between documents, emphasizing​​​‌ some structural distinguishing qualities.​ Graphs establishing such relationships​‌ need to be constructed​​ at various level of​​​‌ granularity, providing some support​ for summarizing structural traits.​‌

Summarizing multimedia information poses​​ several scientific challenges that​​​‌ are:

• Choosing the most​ relevant multimedia aggregation type​‌: Taking a multimedia​​ collection into account, a​​​‌ same piece of information​ can be present in​‌ several modalities. The issue​​ of selecting the most​​​‌ suitable one to express​ a given concept has​‌ thus to be considered​​ together with the way​​​‌ to mix the various​ modalities into an acceptable​‌ production. Standard summarization algorithms​​ have to be revisited​​​‌ so that they can​ handle continuous representation spaces,​‌ allowing them to benefit​​ from the various modalities​​​‌  31.
• Expressing user’s​ preferences: Different users​‌ may appreciate quite different​​ forms of multimedia summaries,​​​‌ and convenient ways to​ express their preferences have​‌ to be proposed. We​​ for example focus on​​​‌ the opportunities offered by​ the constraint-based framework.
• Evaluating​‌ multimedia summaries: Finding​​ criteria to characterize what​​​‌ a good summary is​ remains challenging, e.g., how​‌ to measure the global​​ relevance of a multimodal​​​‌ summary and how to​ compare information between and​‌ across two modalities. We​​ tackle this issue particularly​​​‌ via a collaboration with​ A. Smeaton at DCU,​‌ comparing the automatic measures​​ we will develop to​​​‌ human judgments obtained by​ crowd-sourcing.
• Taking into account​‌ structuring and dynamicity:​​ Typed links between multimedia​​​‌ fragments, and hierarchical topical​ structures of documents obtained​‌ via work previously developed​​ within the team are​​​‌ two types of knowledge​ which have seldom been​‌ considered as long as​​ summarization is concerned. Knowing​​​‌ that the event present​ in a document is​‌ causally related to another​​ event described in another​​ document can however modify​​​‌ the ways summarization algorithms‌ have to consider information.‌​‌ Moreover the question of​​ producing coarse-to-fine grain summaries​​​‌ exploiting the topical structure‌ of documents is still‌​‌ an open issue. Summarizing​​ dynamic collections is also​​​‌ challenging and it is‌ one of the questions‌​‌ we consider.

The Synapses Labcom

The‌​‌ year 2025 is marked​​ by close collaboration with​​​‌ a major French media‌ organization. The Linkmedia Ouest-France‌​‌ team is running Synapses,​​ the first “joint laboratory”​​​‌ with a press organization‌ to develop AI for‌​‌ journalism. Supported by the​​ French National Research Agency​​​‌ (ANR), it comes after‌ thirty years of partnership,‌​‌ and targets the analysis​​ of photo archives, the​​​‌ processing of historical texts‌ and the visualization of‌​‌ complex data. Synapses combines​​ “AI and data sovereignty”​​​‌ to exploit a unique‌ heritage of 105 million‌​‌ documents. This partnership highlights​​ the sharing of scientific​​​‌ knowledge, but also our‌ respective sensitivities to the‌​‌ societal impact of AI​​​‌ in order to work​ on better information for​‌ diverse audiences.

7.1.1 MADHyS​

• Name:
  MULTI-LEVEL AGGREGATIONS FOR​‌ DYNAMIC HYPERGRAPHS STORYLINES
• Keywords:​​
  Data Exploration, Data visualization​​​‌
• Functional Description:
  Visualization of​ large numbers of interdependent​‌ data sets, with relationships​​ that evolve over time.​​​‌ Need to simultaneously visualize​ information at different scales,​‌ both detailed and broad,​​ in order to understand​​​‌ a phenomenon in its​ entirety.
• Contact:
  Nicolas Fouque​‌
• Participants:
  Laurent Amsaleg, Vanessa​​ Pena Araya, Anastasia Bezerianos​​​‌

8.1.1 Revisiting Transferable​​ Adversarial Images: Systemization, Evaluation,​​​‌ and New Insights

Participants:​ Zhengyu Zhao [Xjtu -​‌ Xi'an Jiaotong University],​​ Hanwei Zhang [Institute of​​​‌ Intelligent Software, Guangzhou –​ Saarland University, Saarbrücken],​‌ Renjue Li [School of​​ Artificial Intelligence - Nanjing]​​​‌, Ronan Sicre [M2P2​ - Laboratoire de Mécanique,​‌ Modélisation et Procédés Propres]​​, Laurent Amsaleg,​​​‌ Michael Backes [CISPA -​ Helmholtz Center for Information​‌ Security, Saarbrücken], Qi​​ Li [THU - Tsinghua​​​‌ University, Beijing], Qian​ Wang [Artificial Intelligence Institute​‌ of Wuhan University, Wuhan​​ City], Chao Shen​​​‌ [Xjtu - Xi'an Jiaotong​ University].

Transferable adversarial​‌ images raise critical security​​ concerns for computer vision​​​‌ systems in real-world, blackbox​ attack scenarios. Although many​‌ transfer attacks have been​​ proposed, existing research lacks​​​‌ a systematic and comprehensive​ evaluation. In this paper​‌ 12, we systemize​​ transfer attacks into five​​​‌ categories around the general​ machine learning pipeline and​‌ provide the first comprehensive​​ evaluation, with 23 representative​​​‌ attacks against 11 representative​ defenses, including the recent,​‌ transfer-oriented defense and the​​ real-world Google Cloud Vision.​​​‌ In particular, we identify​ two main problems of​‌ existing evaluations: (1) for​​ attack transferability, lack of​​​‌ intra-category analyses with fair​ hyperparameter settings, and (2)​‌ for attack stealthiness, lack​​ of diverse measures. Our​​​‌ evaluation results validate that​ these problems have indeed​‌ caused misleading conclusions and​​ missing points, and addressing​​​‌ them leads to new,​ consensuschallenging insights, such as​‌ (1) an early attack,​​ DI, even outperforms all​​​‌ similar follow-up ones, (2)​ the state-of-the-art (whitebox) defense,​‌ DiffPure, is even vulnerable​​ to (black-box) transfer attacks,​​​‌ and (3) even under​ the same Lp constraint,​‌ different attacks yield dramatically​​ different stealthiness results regarding​​​‌ diverse imperceptibility metrics, finer-grained​ measures, and a user​‌ study. We hope that​​ our analyses will serve​​​‌ as guidance on properly​ evaluating transferable adversarial images​‌ and advance the design​​ of attacks and defenses.​​​‌

8.1.2 Bregman Conditional Random​ Fields: Sequence Labeling with​‌ Parallelizable Inference Algorithms

Participants:​​ Caio Corro, Mathieu​​​‌ Lacroix [LIPN - Laboratoire​ d'Informatique de Paris-Nord],​‌ Joseph Le Roux [LIPN​​ - Laboratoire d'Informatique de​​​‌ Paris-Nord].

We propose​ a novel discriminative model​‌ for sequence labeling called​​ Bregman conditional random fields​​​‌ (BCRF). Contrary to standard​ linear-chain conditional random fields,​‌ BCRF allows fast parallelizable​​ inference algorithms based on​​​‌ iterative Bregman projections. In​ this paper, we show​‌ how such models can​​ be learned using Fenchel-Young​​ losses, including extension for​​​‌ learning from partial labels‌ 14. Experimentally, our‌​‌ approach delivers comparable results​​ to CRF while being​​​‌ faster, and achieves better‌ results in highly constrained‌​‌ settings compared to mean​​ field, another parallelizable alternative.​​​‌

8.1.3 Few-Shot Domain Adaptation‌ for Named-Entity Recognition via‌​‌ Joint Constrained k-Means and​​ Subspace Selection

Participants: Ayoub​​​‌ Hammal [STL - Sciences‌ et Technologies des Langues‌​‌ - LISN], Benno​​ Uthayasooriyar [LMBA - Laboratoire​​​‌ de Mathématiques de Bretagne‌ Atlantique, SCOR SE, Paris]‌​‌, Caio Corro.​​

Named-entity recognition (NER) is​​​‌ a task that typically‌ requires large annotated datasets,‌​‌ which limits its applicability​​ across domains with varying​​​‌ entity definitions. This paper‌ addresses few-shot NER, aiming‌​‌ to transfer knowledge to​​ new domains with minimal​​​‌ supervision 15. Unlike‌ previous approaches that rely‌​‌ solely on limited annotated​​ data, we propose a​​​‌ weakly supervised algorithm that‌ combines small labeled datasets‌​‌ with large amounts of​​ unlabeled data. Our method​​​‌ extends the kmeans algorithm‌ with label supervision, cluster‌​‌ size constraints and domain-specific​​ discriminative subspace selection. This​​​‌ unified framework achieves state-of-the-art‌ results in fewshot NER‌​‌ on several English datasets.​​

8.1.4 Training LayoutLM from​​​‌ Scratch for Efficient Named-Entity‌ Recognition in the Insurance‌​‌ Domain

Participants: Benno Uthayasooriyar​​ [LMBA - Laboratoire de​​​‌ Mathématiques de Bretagne Atlantique,‌ SCOR SE, Paris],‌​‌ Antoine Ly [SCOR SE,​​ Paris], Franck Vermet​​​‌ [LMBA - Laboratoire de‌ Mathématiques de Bretagne Atlantique]‌​‌, Caio Corro.​​

Generic pre-trained neural networks​​​‌ may struggle to produce‌ good results in specialized‌​‌ domains like finance and​​ insurance. This is due​​​‌ to a domain mismatch‌ between training data and‌​‌ downstream tasks, as in-domain​​ data are often scarce​​​‌ due to privacy constraints.‌ In this work, we‌​‌ compare different pre-training strategies​​ for LAYOUTLM 19.​​​‌ We show that using‌ domain-relevant documents improves results‌​‌ on a named-entity recognition​​ (NER) problem using a​​​‌ novel dataset of anonymized‌ insurance-related financial documents called‌​‌ PAYSLIPS. Moreover, we show​​ that we can achieve​​​‌ competitive results using a‌ smaller and faster model.‌​‌

8.1.5 EuroBERT: Scaling Multilingual​​ Encoders for European Languages​​​‌

Participants: Nicolas Boizard [MICS‌ - Mathématiques et Informatique‌​‌ pour la Complexité et​​ les Systèmes], Hippolyte​​​‌ Gisserot-Boukhlef [MICS - Mathématiques‌ et Informatique pour la‌​‌ Complexité et les Systèmes]​​, Duarte M. Alves​​​‌ [Instituto Superior Técnico],‌ André F T Martins‌​‌ [Instituto Superior Técnico],​​ Ayoub Hammal [Université Paris-Saclay]​​​‌, Caio Corro,‌ Céline Hudelot [MICS -‌​‌ Mathématiques et Informatique pour​​ la Complexité et les​​​‌ Systèmes], Emmanuel Malherbe‌ [Artefact, Paris], Etienne‌​‌ Malaboeuf [CINES], Fanny​​ Jourdan [IRT Saint Exupéry​​​‌ - Institut de Recherche‌ Technologique], Gabriel Hautreux‌​‌ [CINES], João Alves​​ [Unbabel], Kevin El-Haddad​​​‌ [ISIA - Institut Supérieur‌ d'Informatique et d'Automatique],‌​‌ Manuel Faysse [Illuin Technology,​​ Centrale Supelec], Maxime​​​‌ Peyrard [GETALP - Groupe‌ d'Étude en Traduction Automatique/Traitement‌​‌ Automatisé des Langues et​​ de la Parole],​​​‌ Nuno M Guerreiro [MICS‌ - Mathématiques et Informatique‌​‌ pour la Complexité et​​ les Systèmes], Patrick​​​‌ Fernandes [Instituto Superior Técnico]‌, Ricardo Rei [Unbabel]‌​‌, Pierre Colombo [MICS​​​‌ - Mathématiques et Informatique​ pour la Complexité et​‌ les Systèmes].

General-purpose​​ multilingual vector representations, used​​​‌ in retrieval, regression, and​ classification, are traditionally obtained​‌ from bidirectional encoder models.​​ Despite their wide applicability,​​​‌ encoders have been recently​ overshadowed by advances in​‌ generative decoder-only models. However,​​ many innovations driving this​​​‌ progress are not inherently​ tied to decoders. In​‌ this paper, we revisit​​ the development of multilingual​​​‌ encoders through the lens​ of these advances, and​‌ introduce EuroBERT, a family​​ of multilingual encoders covering​​​‌ European and widely spoken​ global languages 13.​‌ Our models outperform existing​​ alternatives across a diverse​​​‌ range of tasks, spanning​ multilingual capabilities, mathematics, and​‌ coding, and natively support​​ sequences of up to​​​‌ 8,192 tokens. We also​ examine the design decisions​‌ behind EuroBERT, offering insights​​ into our dataset composition​​​‌ and training pipeline. We​ publicly release the EuroBERT​‌ models, including intermediate training​​ checkpoints, together with our​​​‌ training framework.

8.1.6 Relaxed​ syntax modeling in Transformers​‌ for future-proof license plate​​ recognition

Participants: Florent Meyer​​​‌ [ANTAI], Laurent Guichard​ [ANTAI], Denis Coquenet​‌ [SHADOC], Guillaume Gravier​​, Yann Soullard [SHADOC]​​​‌, Bertrand Coüasnon [SHADOC]​.

Effective license plate​‌ recognition systems are required​​ to be resilient to​​​‌ constant change, as new​ license plates are released​‌ into traffic daily. While​​ Transformer-based networks excel in​​​‌ their recognition at first​ sight, we observe significant​‌ performance drop over time​​ which proves them unsuitable​​​‌ for tense production environments.​ Indeed, such systems obtain​‌ state-of-the-art results on plates​​ whose syntax is seen​​​‌ during training. Yet, we​ show they perform similarly​‌ to random guessing on​​ future plates where legible​​​‌ characters are wrongly recognized​ due to a shift​‌ in their syntax. After​​ highlighting the flows of​​​‌ positional and contextual information​ in Transformer encoder-decoders, we​‌ identify several causes for​​ their over-reliance on past​​​‌ syntax. Following, we devise​ architectural cut-offs and replacements​‌ which we integrate into​​ SaLT, an attempt at​​​‌ a Syntax-Less Transformer for​ syntax-agnostic modeling of license​‌ plate representations. Experiments on​​ both real and synthetic​​​‌ datasets show that our​ approach reaches top accuracy​‌ on past syntax and​​ most importantly nearly maintains​​​‌ performance on future license​ plates. We further demonstrate​‌ the robustness of our​​ architecture enhancements by way​​​‌ of various ablations 17​.

8.1.7 CroissantLLM: A​‌ Truly Bilingual French-English Language​​ Model

Participants: Manuel Faysse​​​‌ [MICS - Mathématiques et​ Informatique pour la Complexité​‌ et les Systèmes],​​ Patrick Fernandes [Instituto de​​​‌ Telecomunicações, Lisboa, Portugal],​ Nuno M Guerreiro [MICS​‌ - Mathématiques et Informatique​​ pour la Complexité et​​​‌ les Systèmes], Antonio​ Loison [Illuin Technology],​‌ Duarte M. Alves [Instituto​​ Superior Técnico], Caio​​​‌ Corro, Nicolas Boizard​ [MICS - Mathématiques et​‌ Informatique pour la Complexité​​ et les Systèmes],​​​‌ Ricardo Rei [INESC-ID -​ Instituto de Engenharia de​‌ Sistemas e Computadores Investigação​​ e Desenvolvimento em Lisboa]​​​‌, Pedro Raphaël Martins​ [LTSI], Antoni Casademunt​‌ [Imperial College London],​​ François Yvon [MLIA -​​​‌ Machine Learning and Information​ Access]], André Martins​‌ [Instituto de Telecomunicações, Lisboa,​​ Portugal], Gautier Viaud​​ [Illuin Technology], Céline​​​‌ Hudelot [MICS - Mathématiques‌ et Informatique pour la‌​‌ Complexité et les Systèmes]​​, Pierre Colombo [MICS​​​‌ - Mathématiques et Informatique‌ pour la Complexité et‌​‌ les Systèmes].

We​​ introduce CroissantLLM 10,​​​‌ a 1.3B language model‌ pretrained on a set‌​‌ of 3T English and​​ French tokens, to bring​​​‌ to the research and‌ industrial community a high-performance,‌​‌ fully open-sourced bilingual model​​ that runs swiftly on​​​‌ consumer-grade local hardware. To‌ that end, we pioneer‌​‌ the approach of training​​ an intrinsically bilingual model​​​‌ with a 1:1 English-to-French‌ pretraining data ratio, a‌​‌ custom tokenizer, and bilingual​​ finetuning datasets. We release​​​‌ the training dataset, notably‌ containing a French split‌​‌ with manually curated, high-quality,​​ and varied data sources.​​​‌ To assess performance outside‌ of English, we craft‌​‌ a novel benchmark, FrenchBench,​​ consisting of an array​​​‌ of classification and generation‌ tasks, covering various orthogonal‌​‌ aspects of model performance​​ in the French Language.​​​‌ Additionally, rooted in transparency‌ and to foster further‌​‌ Large Language Model research,​​ we release codebases, and​​​‌ dozens of checkpoints across‌ various model sizes, training‌​‌ data distributions, and training​​ steps, as well as​​​‌ fine-tuned Chat models, and‌ strong translation models. We‌​‌ evaluate our model through​​ the FMTI framework, and​​​‌ validate 81 % of‌ the transparency criteria, far‌​‌ beyond the scores of​​ even most open initiatives.​​​‌ This work enriches the‌ NLP landscape, breaking away‌​‌ from previous English-centric work​​ in order to strengthen​​​‌ our understanding of multilinguality‌ in language models.

8.1.8‌​‌ Extraction of Contrastive Rules​​ from Syntactic Treebanks: A​​​‌ Case Study in Romance‌ Languages

Participants: Santiago Herrera‌​‌ [MoDyCo - Modèles, Dynamiques,​​ Corpus], Ioana-Madalina Silai​​​‌ [Université Paris Nanterre -‌ Département Sciences du Langage]‌​‌, Caio Corro,​​ Bruno Guillaume [SEMAGRAMME],​​​‌ Sylvain Kahane [MoDyCo -‌ Modèles, Dynamiques, Corpus].‌​‌

In this paper, we​​ develop a data-driven contrastive​​​‌ framework to extract common‌ and distinctive linguistic descriptions‌​‌ from syntactic treebanks 16​​. The extracted contrastive​​​‌ rules are defined by‌ a statistically significant difference‌​‌ in frequency and precision,​​ and classified as common​​​‌ and distinctive rules across‌ the set of treebanks.‌​‌ We illustrate our method​​ by working on object​​​‌ word order using Universal‌ Dependencies (UD) treebanks in‌​‌ 6 Romance languages: Brazilian​​ Portuguese, Catalan, French, Italian,​​​‌ Romanian and Spanish. We‌ discuss the limitations faced‌​‌ due to inconsistent annotation​​ and the feasibility of​​​‌ conducting contrastive studies using‌ the UD collection.

8.1.9‌​‌ Discrete Latent Structure in​​ Neural Networks

Participants: Vlad​​​‌ Niculae [Informatics Institute Amsterdam]‌, Caio Corro,‌​‌ Nikita Nangia [NYU -​​ New York University, New​​​‌ York], Tsvetomila Mihaylova‌ [IST / Técnico Lisboa‌​‌ - Instituto Superior Técnico,​​ Universidade de Lisboa, Lisboa]​​​‌, André Martins [IST‌ / Técnico Lisboa -‌​‌ Instituto Superior Técnico, Universidade​​ de Lisboa, Unbabel].​​​‌

Many types of data‌ from fields including natural‌​‌ language processing, computer vision,​​ and bioinformatics, are well​​​‌ represented by discrete, compositional‌ structures such as trees,‌​‌ sequences, or matchings. Latent​​ structure models are a​​​‌ powerful tool for learning‌ to extract such representations,‌​‌ offering a way to​​​‌ incorporate structural bias, discover​ insight about the data,​‌ and interpret decisions. However,​​ effective training is challenging,​​​‌ as neural networks are​ typically designed for continuous​‌ computation. This text explores​​ three broad strategies for​​​‌ learning with discrete latent​ structure: continuous relaxation, surrogate​‌ gradients, and probabilistic estimation.​​ Our presentation relies on​​​‌ consistent notations for a​ wide range of models.​‌ As such, we reveal​​ many new connections between​​​‌ latent structure learning strategies,​ showing how most consist​‌ of the same small​​ set of fundamental building​​​‌ blocks, but use them​ differently, leading to substantially​‌ different applicability and properties​​ 11.

8.1.10 Nested​​​‌ Named Entity Recognition as​ Single-Pass Sequence Labeling

Participants:​‌ Alberto Muñoz-Ortiz [Universidade da​​ Coruña], David Vilares​​​‌ [Universidade da Coruña],​ Caio Corro, Carlos​‌ Gómez-Rodríguez [Universidade da Coruña]​​.

In this paper​​​‌ 18 we cast nested​ named entity recognition (NNER)​‌ as a sequence labeling​​ task by leveraging prior​​​‌ work that linearizes constituency​ structures, effectively reducing the​‌ complexity of this structured​​ prediction problem to straightforward​​​‌ token classification. By combining​ these constituency linearizations with​‌ pretrained encoders, our method​​ captures nested entities while​​​‌ performing exactly $n$ tagging​ actions. Our approach achieves​‌ competitive performance compared to​​ less efficient systems, and​​​‌ it can be trained​ using any off-the-shelf sequence​‌ labeling library.

CIFRE PhD:​‌ Machine learning for identification​​ of factors impacting the​​​‌ quality of service of​ urban buses

Participants: Simon​‌ Malinowski, Guillaume Gravier​​, Erwan Vincent.​​​‌

Duration: 3 years, started​ in Feb. 2022Partner: KEOLIS​‌

CIFRE PhD: Introduction​​​‌ of rejection capabilities and​ externalized language models in​‌ deep learning systems for​​ text reading under adverse​​​‌ conditions

Participants: Guillaume Gravier​.

Duration: 3 years,​‌ started in June 2023​​ Partner: ANTAI

Astrid Maturation:​​ TrustedNews

Participants: Guillaume Gravier​​​‌, Morgane Casanova,‌ Laurent Amsaleg.

Duration:‌​‌ 36 months, started Nov.​​ 2025

Labcom‌​‌ Synapses

Participants: Laurent Amsaleg​​, Guillaume Gravier,​​​‌ Pascale Sébillot, Michel‌ Le Nouy [Ouest-France],‌​‌ Morgane Casanova.

Duration:​​ 54 months, started Jan.​​​‌ 2024

ANR AGAPE​

Participants: Laurent Amsaleg,​‌ Thomas Derrien, Pascale​​ Sébillot.

Duration: 48​​​‌ months, started Jan. 2025​

11.1.1​ Scientific events: organisation

11.1.2 Scientific​‌ events: selection

11.1.3 Journal

11.1.4 Research administration​​

• Guillaume Gravier was director​​​‌ of IRISA (UMR 6074)​ till December 2025
• Pascale​‌ Sébillot was deputy director​​ of IRISA till December​​​‌ 2025

11.2.1 Teaching

• Master: Laurent​​​‌ Amsaleg, Bases de données​ avancées, 25h, M2, INSA​‌ Rennes, France
• Master: Guillaume​​ Gravier, Natural Language Processing,​​ 8h, M1, INSA Rennes​​​‌
• Licence: Guillaume Gravier, Natural‌ language processing, 8h, L3,‌​‌ INSA Rennes
• Master: Pascale​​ Sébillot, Natural Language Processing,​​​‌ 4h, M1, INSA Rennes,‌ France
• Master: Pascale Sébillot,‌​‌ Databases, 18h, M1, DIGISPORT​​ graduate school (EUR), France​​​‌
• Licence: Pascale Sébillot, Natural‌ Language Processing, 6h, L3,‌​‌ INSA Rennes, France
• Licence:​​ Caio Corro, Machine Learning,​​​‌ 10h, L3, INSA Rennes,‌ France

11.2.2 Supervision

• PhD‌​‌ in progress: Hugo Thomas,​​ Zero-shot and few-shot relation​​​‌ extraction in press archives.‌ Started Sept. 2022, Guillaume‌​‌ Gravier and Pascale Sébillot​​
• Ph.D. in progress: Thomas​​​‌ Derrien, Liage d'entités multimodal.‌ Started Oct. 2025, Laurent‌​‌ Amsaleg and Pascale Sébillot​​
• Ph.D. in progress: Lilas​​​‌ Pastré, À la conquête‌ de l'Ouest-France. Started Oct.‌​‌ 2025, Laurent Amsaleg and​​ Christian Le Bart (sciencePo)​​​‌ and Olivier Trédan (Univ.‌ Rennes)
• Ph.D. in progress:‌​‌ Ayoub Hammal, Language modeling​​ under distribution shifts. Started​​​‌ Nov. 2024, Caio Corro‌ and Pierre Zweigenbaum (LISN)‌​‌
• Ph.D. in progress: Carolina​​ Jeronimo De Almeida, Machine​​​‌ learning for temporal graphs.‌ Started Sept. 2022, Silvio‌​‌ Guimarães (PUC Minas, Brésil),​​ Guillaume Gravier, Simon Malinowski​​​‌ (équipe MALT)
• Ph.D. finished‌ in Nov. 2025: Benno‌​‌ Uthayasooriyar, Insurance Document Understanding​​ with Transformers based Language​​​‌ Models, Caio Corro, Franck‌ Vermet (Université de Bretagne‌​‌ Occidentale) and Antoine Ly​​ (entreprise SCOR)

11.2.3 Juries​​​‌

• Pascale Sébillot was the‌ president of the PhD.‌​‌ jury of Darun Cao,​​ Univ. Bretagne Sud, Feb.​​​‌ 2025
• Pascale Sébillot was‌ a jury member for‌​‌ the PhD. of Oumaïma​​ El Khettari, Nantes Univ.,​​​‌ Feb. 2025
• Pascale Sébillot‌ was a reviewer for‌​‌ the PhD. of Marco​​ Naguib, Univ. Paris-Saclay, Sept.​​​‌ 2025
• Pascale Sébillot was‌ the president of the‌​‌ PhD. jury of Élise​​ Lincker, Conservatoire national des​​​‌ arts et métiers, Dec.‌ 2025
• Caio Corro was‌​‌ a jury member for​​ the Ph.D. of Junjie​​​‌ Yang, Institut polytechnique de‌ Paris, July 2025
• Caio‌​‌ Corro was a jury​​ member for the Ph.D.​​​‌ of Santiago Herrera, Université‌ de Nanterre, September 2025‌​‌

11.2.4 Specific official responsibilities​​ in science outreach structures​​​‌

• Caio Corro is a‌ member of the Comité‌​‌ de Rédaction of Bulletins​​ de l'AFIA

11.2.5 Participation​​​‌ in Live events

• Laurent‌ Amsaleg ran two webminars‌​‌ inside the premises of​​ Ouest-France, presenting the main​​​‌ concepts of IA and‌ detailling the synapses project.‌​‌

International journals

• 10 article​​​‌M.Manuel Faysse,​ P.Patrick Fernandes,​‌ N.Nuno Guerreiro,​​ A.Antonio Loison,​​​‌ D.Duarte Alves,​ C.Caio Corro,​‌ N.Nicolas Boizard,​​ J.Jaoe Alves,​​​‌ R.Ricardo Rei,​ P. R.Pedro Raphaël​‌ Martins, A.Antoni​​ Casademunt, F.François​​​‌ Yvon, A.André​ Martins, G.Gautier​‌ Viaud, C.Céline​​ Hudelot and P.Pierre​​​‌ Colombo. CroissantLLM: A​ Truly Bilingual French-English Language​‌ Model.Transactions of​​ Machine Learning ResearchMarch​​​‌ 2025HALback to​ text
• 11 articleV.​‌Vlad Niculae, C.​​Caio Corro, N.​​​‌Nikita Nangia, T.​Tsvetomila Mihaylova and A.​‌André Martins. Discrete​​ Latent Structure in Neural​​​‌ Networks.Foundations and​ Trends in Signal Processing​‌192June 2025​​, 99-211HALDOI​​​‌back to text
• 12​ articleZ.Zhengyu Zhao​‌, H.Hanwei Zhang​​, R.Renjue Li​​​‌, R.Ronan Sicre​, L.Laurent Amsaleg​‌, M.Michael Backes​​, Q.Qi Li​​, Q.Qian Wang​​​‌ and C.Chao Shen‌. Revisiting Transferable Adversarial‌​‌ Images: Systemization, Evaluation, and​​ New Insights.IEEE​​​‌ Transactions on Pattern Analysis‌ and Machine IntelligenceSeptember‌​‌ 2025, 1-16HAL​​DOIback to text​​​‌

International peer-reviewed conferences

• 13‌ inproceedingsN.Nicolas Boizard‌​‌, H.Hippolyte Gisserot-Boukhlef​​, D. M.Duarte​​​‌ M. Alves, A.‌ F.André F T‌​‌ Martins, A.Ayoub​​ Hammal, C.Caio​​​‌ Corro, C.Céline‌ Hudelot, E.Emmanuel‌​‌ Malherbe, E.Etienne​​ Malaboeuf, F.Fanny​​​‌ Jourdan, G.Gabriel‌ Hautreux, J.João‌​‌ Alves, K.Kevin​​ El-Haddad, M.Manuel​​​‌ Faysse, M.Maxime‌ Peyrard, N. M.‌​‌Nuno M Guerreiro,​​ P.Patrick Fernandes,​​​‌ R.Ricardo Rei and‌ P.Pierre Colombo.‌​‌ EuroBERT: Scaling Multilingual Encoders​​ for European Languages.​​​‌Second Conference on Language‌ ModelingCOLM 2025 -‌​‌ Second Conference on Language​​ ModelingMontreal, CanadaMarch​​​‌ 2025, 1-28HAL‌back to text
• 14‌​‌ inproceedingsC.Caio Corro​​, M.Mathieu Lacroix​​​‌ and J. L.Joseph‌ Le Roux. Bregman‌​‌ Conditional Random Fields: Sequence​​ Labeling with Parallelizable Inference​​​‌ Algorithms.Proceedings of‌ the 63rd Annual Meeting‌​‌ of the Association for​​ Computational LinguisticsACL 2025​​​‌ - 63rd Annual Meeting‌ of the Association for‌​‌ Computational Linguistics1Vienne,​​ AustriaAssociation for Computational​​​‌ LinguisticsJuly 2025,‌ 29557-29574HALDOIback‌​‌ to text
• 15 inproceedings​​A.Ayoub Hammal,​​​‌ B.Benno Uthayasooriyar and‌ C.Caio Corro.‌​‌ Few-Shot Domain Adaptation for​​ Named-Entity Recognition via Joint​​​‌ Constrained k-Means and Subspace‌ Selection.COLING 2025‌​‌ - 31st International Conference​​ on Computational LinguisticsAbu​​​‌ Dhabi, United Arab Emirates‌2025, 9902–9916HAL‌​‌DOIback to text​​
• 16 inproceedingsS.Santiago​​​‌ Herrera, I.-M.Ioana-Madalina‌ Silai, C.Caio‌​‌ Corro, B.Bruno​​ Guillaume and S.Sylvain​​​‌ Kahane. Extraction of‌ Contrastive Rules from Syntactic‌​‌ Treebanks: A Case Study​​ in Romance Languages.​​​‌QUASY 2025 - Third‌ Workshop on Quantitative Syntax‌​‌Ljubljana, SloveniaAugust 2025​​, 26--38HALback​​​‌ to text
• 17 inproceedings‌F.Florent Meyer,‌​‌ L.Laurent Guichard,​​ D.Denis Coquenet,​​​‌ G.Guillaume Gravier,‌ Y.Yann Soullard and‌​‌ B.Bertrand Coüasnon.​​ Relaxed syntax modeling in​​​‌ Transformers for future-proof license‌ plate recognition.International‌​‌ Conference on Document Analysis​​ and Recognition (ICDAR) 2025​​​‌ICDAR 2025 - International‌ Conference on Document Analysis‌​‌ and RecognitionWuhan, China​​2025, 154-171HAL​​​‌DOIback to text‌
• 18 inproceedingsA.Alberto‌​‌ Muñoz-Ortiz, D.David​​ Vilares, C.Caio​​​‌ Corro and C.Carlos‌ Gómez-Rodríguez. Nested Named‌​‌ Entity Recognition as Single-Pass​​ Sequence Labeling.Findings​​​‌ of the Association for‌ Computational Linguistics: EMNLP 2025‌​‌EMNLP 2025 - Conference​​ on Empirical Methods in​​​‌ Natural Language ProcessingSuzhou,‌ ChinaNovember 2025,‌​‌ 9993-10002HALDOIback​​ to text
• 19 inproceedings​​​‌B.Benno Uthayasooriyar,‌ A.Antoine Ly,‌​‌ F.Franck Vermet and​​ C.Caio Corro.​​​‌ Training LayoutLM from Scratch‌ for Efficient Named-Entity Recognition‌​‌ in the Insurance Domain​​​‌.Proceeedings of the​ COLING 2025 Workshop on​‌ Financial Technology and Natural​​ Language Processing (FinNLP), Financial​​​‌ Narrative Processing (FNP), and​ on Large Language Models​‌ for Finance and Legal​​ (LLMFinLegal)COLING 2025 -​​​‌ 31st International Conference on​ Computational LinguisticsAbu Dabi,​‌ United Arab Emirates2025​​, 1-9HALback​​​‌ to text

Edition (books,​ proceedings, special issue of​‌ a journal)

• 20 proceedings​​18th International Conference on​​​‌ Similarity Search and Applications:​ 18th International Conference, SISAP​‌ 2025, Reykjavik, Iceland, October​​ 1–3, 2025, Proceedings.​​​‌SISAP 2025 - 18th​ International Conference on Similarity​‌ Search and Applications16134​​Lecture Notes in Computer​​​‌ ScienceReykjavik, IcelandSpringer​ Nature Switzerland2025HAL​‌DOI
• 21 periodicalJ.-D.​​Jean-Daniel Kant, G.​​​‌Grégory Bonnet and D.​Dominique Longin, eds.​‌ IA & économie.​​Bulletin de l'Association Française​​​‌ pour l'Intelligence Artificielle129​Association Française pour l'Intelligence​‌ ArtificielleJuly 2025,​​ 1-60HAL