Event List
retrieve:
Return the details about the given Event id.
list:
List all Event objects.
GET /api/v1/events/?format=api&offset=140&ordering=start_time
{ "count": 258, "next": "https://memento.epfl.ch/api/v1/events/?format=api&limit=10&offset=150&ordering=start_time", "previous": "https://memento.epfl.ch/api/v1/events/?format=api&limit=10&offset=130&ordering=start_time", "results": [ { "id": 73127, "title": "AI Center Seminar – AI Fundamentals Series – MATS Fellows – \"Towards Robust Model Organisms for Eliciting Latent Knowledge\"", "slug": "ai-center-seminar-ai-fundamentals-series-mats-fell", "event_url": "https://memento.epfl.ch/event/ai-center-seminar-ai-fundamentals-series-mats-fell", "visual_url": "https://memento.epfl.ch/image/34315/200x112.jpg", "visual_large_url": "https://memento.epfl.ch/image/34315/720x405.jpg", "visual_maxsize_url": "https://memento.epfl.ch/image/34315/max-size.jpg", "lang": "en", "start_date": "2026-11-17", "end_date": "2026-11-17", "start_time": "14:30:00", "end_time": "15:30:00", "description": "<p>This talk is jointly organized by the <strong>EPFL AI Center, the ML4D Lab, and the</strong> <strong>MATS Research Program</strong> as part of the <strong>AI Fundamentals Seminar Series</strong>.<br>\r\n<br>\r\n<strong>Title</strong><br>\r\nTowards Robust Model Organisms for Eliciting Latent Knowledge<br>\r\n<br>\r\n<strong>Abstract</strong><br>\r\nA key problem in AI safety is eliciting latent knowledge (ELK), extracting information from a model that is misaligned and more knowledgeable than its overseer. For example, a misaligned model used to scan code for vulnerabilities may want to withhold its findings, i.e., sandbag, in order to exploit them later.<br>\r\nTo study ELK empirically, we need model organisms that mirror this setting. Existing model organisms fall short in several respects. In this work, we train model organisms to withhold solutions in a vulnerability detection task.<br>\r\nWe combine transcript distillation with counter-prompting and tamper-resistance training to instill three properties that we argue are critical for a useful ELK testbed:\r\n</p><ol start=\"1\">\r\n\t<li><strong>Selectivity:</strong> The model sandbags when it thinks the overseer doesn't know the answer and complies otherwise.</li>\r\n\t<li><strong>Robustness:</strong> The model resists elicitation.</li>\r\n\t<li><strong>Retention:</strong> The model retains a knowledge advantage over the overseer, as opposed to losing the relevant capability.</li>\r\n</ol>\r\nWe train a sequence of model organisms and show that our strongest model substantially outperforms prior models on robustness while keeping the other two properties intact.<br>\r\nWe hope that our model organisms provide a useful ELK testbed for the community, and that our methodology sparks the development of increasingly challenging model organisms.<br>\r\n<br>\r\n<strong>Speakers & Bios</strong>\r\n\r\n<ul>\r\n\t<li><strong>Euodia Dodd — Imperial College London</strong><br>\r\n\tEuodia is a PhD student in the AI Security and Privacy Lab at Imperial College London supervised by Prof. Yves-Alexandre de Montjoye, and a MATS 10.1 fellow researching AI Control. Her research focuses on detecting and controlling AI systems that behave adversarially, with research spanning model organisms, backdoors and quantifying privacy leakage in AI models. Before her PhD, Euodia completed an MPhil at Cambridge and spent two years building production AI systems at Goldman Sachs.<br>\r\n\t </li>\r\n\t<li><strong>Lennart Finke — MATS Fellow</strong><br>\r\n\tLennart Finke's work centers on the rigorous evaluation of large language models, with an emphasis on constructing difficult, contamination-resistant benchmarks that measure capabilities at the frontier of expert human knowledge. He contributed to Humanity's Last Exam, a multi-modal benchmark of expert-authored academic questions designed to resist saturation and internet lookup, and to a Korean professional benchmark suite that cleans existing datasets and tests licensure-level domain expertise. This places his interests squarely in AI capability measurement and evaluation methodology, relevant to AI safety.<br>\r\n\t </li>\r\n\t<li><strong>Hardik Bhatnagar — Tübingen AI Center</strong> - Hardik is a PhD student at the Tübingen AI Center, advised by Matthias Bethge and Maksym Andriushchenko. He studies how to rigorously measure what LLMs can and can't do, especially as autonomous agents over long horizons, so we can deploy them safely and reliably. He co-led PostTrainBench, which measures whether LLM agents can autonomously post-train LLMs, and <em>A Sober Look at LM Reasoning</em>, which showed that many reported reasoning gains shrink under standardized evaluation. Most recently, at MATS with Mary Phuong (Google DeepMind), he trained model organisms for sandbagging: models that deliberately underperform on evaluations.</li>\r\n</ul>", "image_description": "", "creation_date": "2026-11-09T14:27:49", "last_modification_date": "2026-11-09T14:31:06", "link_label": "Zoom Link", "link_url": "https://epfl.zoom.us/j/67428504046?pwd=bpqZcJmCalrvsTlGVbCi097IO2ryIP.1", "canceled": "False", "cancel_reason": "", "place_and_room": "ELE 117", "url_place_and_room": "https://plan.epfl.ch/?room==ELE%20117", "url_online_room": "https://epfl.zoom.us/j/67428504046?pwd=bpqZcJmCalrvsTlGVbCi097IO2ryIP.1", "spoken_languages": [ "https://memento.epfl.ch/api/v1/spoken_languages/2/?format=api" ], "speaker": "Euodia Dodd, Lennart Finke, Hardik Bhatnagar", "organizer": "", "contact": "Nicolas Machado & Fares Fawzi", "is_internal": "False", "theme": "", "vulgarization": { "id": 1, "fr_label": "Tout public", "en_label": "General public" }, "registration": { "id": 3, "fr_label": "Entrée libre", "en_label": "Free" }, "keywords": "", "file": null, "icalendar_url": "https://memento.epfl.ch/event/export/122656/", "category": { "id": 1, "code": "CONF", "fr_label": "Conférences - Séminaires", "en_label": "Conferences - Seminars", "activated": true }, "academic_calendar_category": null, "domains": [], "mementos": [ "https://memento.epfl.ch/api/v1/mementos/1/?format=api", "https://memento.epfl.ch/api/v1/mementos/6/?format=api", "https://memento.epfl.ch/api/v1/mementos/416/?format=api" ] }, { "id": 72315, "title": "EPFL Investor Day 2026 - Discover & Network", "slug": "epfl-investor-day-2026-discover-network", "event_url": "https://memento.epfl.ch/event/epfl-investor-day-2026-discover-network", "visual_url": "https://memento.epfl.ch/image/33578/200x112.jpg", "visual_large_url": "https://memento.epfl.ch/image/33578/720x405.jpg", "visual_maxsize_url": "https://memento.epfl.ch/image/33578/max-size.jpg", "lang": "en", "start_date": "2026-11-25", "end_date": "2026-11-25", "start_time": "15:00:00", "end_time": "19:00:00", "description": "<p>EPFL Investor Day opens its doors to the wider innovation community for an afternoon of inspiring presentations, keynotes, thought-provoking discussions, emerging technologies and networking opportunities.<br>\r\n<br>\r\nWhether you're an entrepreneur, investor, corporate leader, researcher, student or simply passionate about innovation, join us to discover the ideas, technologies and people shaping tomorrow's world.<br>\r\n<br>\r\n<strong>Programme Highlights<br>\r\n<br>\r\n15:15 - Doors open</strong><br>\r\n<br>\r\n<strong>15:15 - European Spinout Report: Swiss Premiere</strong><br>\r\n<br>\r\nFollowing its official launch at Slush on 18 November, the <strong>European Spinouts Report 2026</strong> will be presented exclusively in Switzerland at EPFL Investor Day.<br>\r\nDiscover the latest insights into Europe's university spinout ecosystem, the role of research institutions in driving deeptech innovation, and Switzerland's position in this rapidly evolving landscape. <a href=\"https://dealroom.co/european-spinouts-report-2026/\">Learn more about the report</a><br>\r\n<br>\r\n<strong>16:00 - Inspiring Keynotes & Panel Discussions</strong><br>\r\n<br>\r\nAn international keynote will explore how bold ideas and breakthrough innovation can help tackle some of the world's greatest challenges.<br>\r\nTwo expert panels will address key questions for Switzerland's innovation ecosystem:\r\n</p><ul>\r\n\t<li><strong>The AI Race Is On: What Strategic Positions Should Swiss Startups Take?</strong> - Exploring opportunities, challenges and strategies for Swiss startups in the global AI landscape.</li>\r\n\t<li><strong>Scaling Globally, Deciding Locally</strong> - Discussing how startups can access international capital and markets while maintaining strategic independence.</li>\r\n</ul>\r\n<strong>17:30 - Startups in Creation: Pitch Session</strong><br>\r\n<br>\r\nGet a first glimpse of the next generation of entrepreneurs! Ten promising early-stage ventures will each have two minutes to present their ideas and ambitions.<br>\r\n<br>\r\n<strong>18:00 - Networking Apéritif</strong><br>\r\n<br>\r\nEnd the day by connecting with founders, investors, industry leaders, researchers and key players from the innovation ecosystem.<br>\r\n<br>\r\n<strong>Startup Village - Open throughout the afternoon</strong><br>\r\n<br>\r\nExplore the Startup Village at any time during the afternoon! Discover breakthrough technologies, meet innovative startups and connect directly with the entrepreneurs bringing their ideas to market.<br>\r\n<br>\r\n<strong>Join Us!</strong>\r\n\r\n<ul>\r\n\t<li>Participation is free, but registration is required.</li>\r\n\t<li>Register here: <a data-rich-text-autolink=\"\" href=\"https://www.b2match.com/e/investor-day-2026\">https://www.b2match.com/e/investor-day-2026</a></li>\r\n\t<li><strong>Registration code: plenaryID26</strong></li>\r\n</ul>\r\n<br>\r\nEPFL Investor Day 2026 is organized in partnership with <a data-test-app-aware-link=\"\" href=\"https://www.linkedin.com/company/ubs/\" tabindex=\"0\" target=\"_self\">UBS</a> and supported by <a data-test-app-aware-link=\"\" href=\"https://www.linkedin.com/company/swisscom-ventures/\" tabindex=\"0\" target=\"_self\">Swisscom Ventures</a> and <a data-test-app-aware-link=\"\" href=\"https://www.linkedin.com/company/etat-de-vaud/\" tabindex=\"0\" target=\"_self\">Canton of Vaud</a>.", "image_description": "", "creation_date": "2026-08-07T16:24:50", "last_modification_date": "2026-10-16T13:32:25", "link_label": "Agenda", "link_url": "https://www.b2match.com/e/investor-day-2026/components/70770", "canceled": "False", "cancel_reason": "", "place_and_room": "SwissTech Convention Center", "url_place_and_room": "https://plan.epfl.ch/?room==STCC%200%209407", "url_online_room": "", "spoken_languages": [ "https://memento.epfl.ch/api/v1/spoken_languages/2/?format=api" ], "speaker": "", "organizer": "Vice Presidency for Innovation and Impact", "contact": "<a href=\"mailto:[email protected]\">Aurelie Schick</a>", "is_internal": "False", "theme": "", "vulgarization": { "id": 2, "fr_label": "Public averti", "en_label": "Informed public" }, "registration": { "id": 1, "fr_label": "Sur inscription", "en_label": "Registration required" }, "keywords": "", "file": null, "icalendar_url": "https://memento.epfl.ch/event/export/121390/", "category": { "id": 1, "code": "CONF", "fr_label": "Conférences - Séminaires", "en_label": "Conferences - Seminars", "activated": true }, "academic_calendar_category": null, "domains": [], "mementos": [ "https://memento.epfl.ch/api/v1/mementos/1/?format=api", "https://memento.epfl.ch/api/v1/mementos/290/?format=api", "https://memento.epfl.ch/api/v1/mementos/422/?format=api", "https://memento.epfl.ch/api/v1/mementos/119/?format=api", "https://memento.epfl.ch/api/v1/mementos/6/?format=api", "https://memento.epfl.ch/api/v1/mementos/416/?format=api", "https://memento.epfl.ch/api/v1/mementos/3/?format=api", "https://memento.epfl.ch/api/v1/mementos/5/?format=api", "https://memento.epfl.ch/api/v1/mementos/8/?format=api" ] }, { "id": 72400, "title": "Training \"Data/Code Publication\"", "slug": "training-datacode-publication-5", "event_url": "https://memento.epfl.ch/event/training-datacode-publication-5", "visual_url": "https://memento.epfl.ch/image/33651/200x112.jpg", "visual_large_url": "https://memento.epfl.ch/image/33651/720x405.jpg", "visual_maxsize_url": "https://memento.epfl.ch/image/33651/max-size.jpg", "lang": "en", "start_date": "2026-11-26", "end_date": "2026-11-26", "start_time": "15:00:00", "end_time": "16:30:00", "description": "<div class=\"all_text\"><strong>While research funders and publishers mandate the dissemination of data and code underlying the published results, little guidance is provided on how to do it. </strong><br>\r\n<br>\r\nThis workshop addresses the selection of suitable data repositories, explores the role of data journals, and examines data publishing beyond the basic requirements of funders. The session includes a hands-on demonstration of Zenodo, the preferred data repository among EPFL researchers. We will also explore how Zenodo integrates with GitHub repositories, ensuring all necessary publishing metadata is available to disseminate your academic work. Additionally, we will cover the appropriate licensing for your datasets and address data anonymization concerns.<br>\r\n<br>\r\nAt the end of this workshop, which will be held <strong>in presence only</strong>, you will be able to:\r\n<ul>\r\n\t<li>Choose the right repository for your code/data;</li>\r\n\t<li>Publish your code/data in the EPFL Zenodo community.</li>\r\n</ul>\r\nWhile not mandatory, basic knowledge of Research Data Management is recommended: for an introductory training, please register for the workshop <a href=\"https://bookwhen.com/fr/epfl_library/e/ev-s9esg-20260903150000\">ABC of Research Data Management</a>.<br>\r\n<br>\r\n<a href=\"https://bookwhen.com/fr/epfl_library/e/ev-s0c7o-20261126150000\">Registration</a><br>\r\n<br>\r\nMore information about <a href=\"https://www.epfl.ch/campus/library/training/\">EPFL Library Training offer</a></div>", "image_description": "CC-BY-NC-SA EPFL Library", "creation_date": "2026-08-20T10:31:54", "last_modification_date": "2026-08-20T15:54:06", "link_label": "Registration", "link_url": "https://bookwhen.com/fr/epfl_library/e/ev-s0c7o-20261126150000", "canceled": "False", "cancel_reason": "", "place_and_room": "INM 011", "url_place_and_room": "https://plan.epfl.ch/?room==INM%20011", "url_online_room": "", "spoken_languages": [ "https://memento.epfl.ch/api/v1/spoken_languages/2/?format=api" ], "speaker": "EPFL Library Research Data Team", "organizer": "EPFL Library", "contact": "[email protected]", "is_internal": "False", "theme": "", "vulgarization": { "id": 1, "fr_label": "Tout public", "en_label": "General public" }, "registration": { "id": 1, "fr_label": "Sur inscription", "en_label": "Registration required" }, "keywords": "", "file": null, "icalendar_url": "https://memento.epfl.ch/event/export/121524/", "category": { "id": 15, "code": "FORM", "fr_label": "Formations internes", "en_label": "Internal trainings", "activated": true }, "academic_calendar_category": null, "domains": [], "mementos": [ "https://memento.epfl.ch/api/v1/mementos/1/?format=api", "https://memento.epfl.ch/api/v1/mementos/3/?format=api", "https://memento.epfl.ch/api/v1/mementos/5/?format=api", "https://memento.epfl.ch/api/v1/mementos/6/?format=api", "https://memento.epfl.ch/api/v1/mementos/291/?format=api", "https://memento.epfl.ch/api/v1/mementos/8/?format=api", "https://memento.epfl.ch/api/v1/mementos/269/?format=api", "https://memento.epfl.ch/api/v1/mementos/275/?format=api" ] }, { "id": 72745, "title": "Generative Artificial Intelligence for Accelerating Materials Discovery", "slug": "generative-artificial-intelligence-for-accelerat-2", "event_url": "https://memento.epfl.ch/event/generative-artificial-intelligence-for-accelerat-2", "visual_url": "https://memento.epfl.ch/image/33960/200x112.jpg", "visual_large_url": "https://memento.epfl.ch/image/33960/720x405.jpg", "visual_maxsize_url": "https://memento.epfl.ch/image/33960/max-size.jpg", "lang": "en", "start_date": "2026-11-27", "end_date": "2026-11-27", "start_time": "15:00:00", "end_time": null, "description": "<p>Thesis Director: Prof. Ph. Schwaller,<br>\r\nChemistry and Chemical Engineering doctoral program<br>\r\nThesis Nr. 11943<br>\r\n<br>\r\nTo take part in the public defense, please contact directly the speaker</p>", "image_description": "", "creation_date": "2026-10-01T19:43:58", "last_modification_date": "2026-10-01T19:43:59", "link_label": "", "link_url": "", "canceled": "False", "cancel_reason": "", "place_and_room": "CM 1 4", "url_place_and_room": "https://plan.epfl.ch/?room==CM%201%204", "url_online_room": "https://epfl.zoom.us/j/66541265867?pwd=7w9G38oGbpRC8F0eWZyRE9yhD8pvgW.1", "spoken_languages": [], "speaker": "<a href=\"mailto:[email protected]\"><strong>Junwu CHEN</strong></a>", "organizer": "", "contact": "<a href=\"mailto:[email protected]\"><strong>Junwu CHEN</strong></a>", "is_internal": "False", "theme": "", "vulgarization": { "id": 1, "fr_label": "Tout public", "en_label": "General public" }, "registration": { "id": 3, "fr_label": "Entrée libre", "en_label": "Free" }, "keywords": "EDCH", "file": null, "icalendar_url": "https://memento.epfl.ch/event/export/122032/", "category": { "id": 12, "code": "SOUTE", "fr_label": "Soutenances de thèses", "en_label": "Thesis defenses", "activated": true }, "academic_calendar_category": null, "domains": [], "mementos": [ "https://memento.epfl.ch/api/v1/mementos/1/?format=api", "https://memento.epfl.ch/api/v1/mementos/5/?format=api", "https://memento.epfl.ch/api/v1/mementos/6/?format=api" ] }, { "id": 73114, "title": "Proximal optimal transport divergences for generative modeling and sampling", "slug": "proximal-optimal-transport-divergences-for-generat", "event_url": "https://memento.epfl.ch/event/proximal-optimal-transport-divergences-for-generat", "visual_url": "https://memento.epfl.ch/image/34303/200x112.jpg", "visual_large_url": "https://memento.epfl.ch/image/34303/720x405.jpg", "visual_maxsize_url": "https://memento.epfl.ch/image/34303/max-size.jpg", "lang": "en", "start_date": "2026-11-11", "end_date": "2026-11-11", "start_time": "15:00:00", "end_time": "16:00:00", "description": "<p>The motivation for this work comes from generative modeling and sampling problems where the target distributions are either supported on low-dimensional structure and/or known only through empirical samples. Through proximal regularization we construct new information theoretic divergences which combine in a flexible manner the desirable properties of relative entropy and optimal transport (Wasserstein distances). Using these divergences, associated Wasserstein gradient flows, and a mixture of implicit and explicit schemes we build efficient sampling and generative modeling algorithms.</p>", "image_description": "", "creation_date": "2026-11-06T10:50:13", "last_modification_date": "2026-11-06T14:26:44", "link_label": "", "link_url": "", "canceled": "False", "cancel_reason": "", "place_and_room": "GA 3 21", "url_place_and_room": "https://plan.epfl.ch/?room==GA%203%2021", "url_online_room": "", "spoken_languages": [ "https://memento.epfl.ch/api/v1/spoken_languages/2/?format=api" ], "speaker": "Professor Luc Rey-Bellet, UMass Amherst", "organizer": "STOAN unit", "contact": "[email protected]", "is_internal": "True", "theme": "", "vulgarization": { "id": 2, "fr_label": "Public averti", "en_label": "Informed public" }, "registration": { "id": 3, "fr_label": "Entrée libre", "en_label": "Free" }, "keywords": "proximal; transport; divergences; STOAN", "file": null, "icalendar_url": "https://memento.epfl.ch/event/export/122636/", "category": { "id": 1, "code": "CONF", "fr_label": "Conférences - Séminaires", "en_label": "Conferences - Seminars", "activated": true }, "academic_calendar_category": null, "domains": [], "mementos": [ "https://memento.epfl.ch/api/v1/mementos/1/?format=api", "https://memento.epfl.ch/api/v1/mementos/5/?format=api", "https://memento.epfl.ch/api/v1/mementos/6/?format=api", "https://memento.epfl.ch/api/v1/mementos/7/?format=api" ] }, { "id": 73115, "title": "Potent optogenetic suppression of neurotransmitter release in the mouse neocortex in vivo", "slug": "potent-optogenetic-suppression-of-neurotransmitter", "event_url": "https://memento.epfl.ch/event/potent-optogenetic-suppression-of-neurotransmitter", "visual_url": "https://memento.epfl.ch/image/34302/200x112.jpg", "visual_large_url": "https://memento.epfl.ch/image/34302/720x405.jpg", "visual_maxsize_url": "https://memento.epfl.ch/image/34302/max-size.jpg", "lang": "en", "start_date": "2026-11-20", "end_date": "2026-11-20", "start_time": "15:00:00", "end_time": null, "description": "<p>Thesis Director: Prof. C. Petersen,<br>\r\nNeuroscience doctoral program<br>\r\nThesis Nr. 11614<br>\r\n<br>\r\nTo take part in the public defense, please contact directly the speaker</p>", "image_description": "", "creation_date": "2026-11-06T10:57:00", "last_modification_date": "2026-11-06T10:57:00", "link_label": "", "link_url": "", "canceled": "False", "cancel_reason": "", "place_and_room": "AI 1153", "url_place_and_room": "https://plan.epfl.ch/?room==AI%201153", "url_online_room": "", "spoken_languages": [], "speaker": "<strong>Meriam MALEKZADEH</strong>", "organizer": "", "contact": "<strong>Meriam MALEKZADEH</strong>", "is_internal": "False", "theme": "", "vulgarization": { "id": 1, "fr_label": "Tout public", "en_label": "General public" }, "registration": { "id": 3, "fr_label": "Entrée libre", "en_label": "Free" }, "keywords": "EDNE", "file": null, "icalendar_url": "https://memento.epfl.ch/event/export/122638/", "category": { "id": 12, "code": "SOUTE", "fr_label": "Soutenances de thèses", "en_label": "Thesis defenses", "activated": true }, "academic_calendar_category": null, "domains": [], "mementos": [ "https://memento.epfl.ch/api/v1/mementos/1/?format=api", "https://memento.epfl.ch/api/v1/mementos/6/?format=api", "https://memento.epfl.ch/api/v1/mementos/9/?format=api" ] }, { "id": 73116, "title": "Hypocoercivity without changing the scalar product", "slug": "hypocoercivity-without-changing-the-scalar-product", "event_url": "https://memento.epfl.ch/event/hypocoercivity-without-changing-the-scalar-product", "visual_url": "https://memento.epfl.ch/image/34304/200x112.jpg", "visual_large_url": "https://memento.epfl.ch/image/34304/720x405.jpg", "visual_maxsize_url": "https://memento.epfl.ch/image/34304/max-size.jpg", "lang": "en", "start_date": "2026-12-02", "end_date": "2026-12-02", "start_time": "15:00:00", "end_time": "16:00:00", "description": "<p>Hypocoercive methods allow to prove the longtime convergence of the law of degenerate stochastic processes such as Langevin dynamics or random time HMC. From a technical viewpoint, these approaches usually involve a change of scalar product (as in the H^1 approach à la Villani, or the L^2 approach of Hérau and Dolbeault/Mouhot/Schmeiser). I will discuss two approaches which do not require such a change of scalar product, and henceforth provide the current sharpest quantitative estimates on the resolvent of the generator, which makes it possible for instance to bound the variance of estimators based on time averages along a realization of the process. The first approach, originally due to Armstrong/Mourrat, was adapted to dynamics on the whole space by Cao/Lu/Wang, and is based on space-time Poincare inequalities. I will review some progress made with Giovanni Brigati on better understanding the algebraic structures behind this method, and possibly mention extensions to weak/weighted Poincaré inequalities (in collaboration with Lihan Wang and Andi Wang). The second approach, developed with Etienne Bernard, Max Fathi and Antoine Levitt, is based on Schur complements and provides direct bounds on the resolvent using the specific saddle-point like structure of the generator under an appropriate orthogonal decomposition of L^2.<br>\r\n </p>", "image_description": "", "creation_date": "2026-11-06T11:24:23", "last_modification_date": "2026-11-06T11:35:26", "link_label": "", "link_url": "", "canceled": "False", "cancel_reason": "", "place_and_room": "GA 3 21", "url_place_and_room": "https://plan.epfl.ch/?room==GA%203%2021", "url_online_room": "", "spoken_languages": [ "https://memento.epfl.ch/api/v1/spoken_languages/2/?format=api" ], "speaker": "Professor Gabriel Stoltz, CERMICS, Ecole des Ponts & MATHERIALS, Inria Paris", "organizer": "STOAN", "contact": "[email protected]", "is_internal": "True", "theme": "", "vulgarization": { "id": 2, "fr_label": "Public averti", "en_label": "Informed public" }, "registration": { "id": 3, "fr_label": "Entrée libre", "en_label": "Free" }, "keywords": "Hypocoercive methods, orthogonal decomposition", "file": null, "icalendar_url": "https://memento.epfl.ch/event/export/122640/", "category": { "id": 1, "code": "CONF", "fr_label": "Conférences - Séminaires", "en_label": "Conferences - Seminars", "activated": true }, "academic_calendar_category": null, "domains": [], "mementos": [ "https://memento.epfl.ch/api/v1/mementos/5/?format=api", "https://memento.epfl.ch/api/v1/mementos/7/?format=api" ] }, { "id": 73131, "title": "Interface-Engineering and Alloying of Aluminium Nanolaminates: Structure-Property-Stability Relationships", "slug": "interface-engineering-and-alloying-of-aluminium--2", "event_url": "https://memento.epfl.ch/event/interface-engineering-and-alloying-of-aluminium--2", "visual_url": "https://memento.epfl.ch/image/34319/200x112.jpg", "visual_large_url": "https://memento.epfl.ch/image/34319/720x405.jpg", "visual_maxsize_url": "https://memento.epfl.ch/image/34319/max-size.jpg", "lang": "en", "start_date": "2026-12-11", "end_date": "2026-12-11", "start_time": "15:00:00", "end_time": null, "description": "<p>Thesis Directors: Prof. J. Michler, Dr J. J. Schwiedrzik<br>\r\nMaterials Science and Engineering doctoral program<br>\r\nThesis Nr. 12237<br>\r\n<br>\r\nTo take part in the public defense, please contact directly the speaker</p>", "image_description": "", "creation_date": "2026-11-09T16:06:42", "last_modification_date": "2026-11-09T16:08:16", "link_label": "", "link_url": "", "canceled": "False", "cancel_reason": "", "place_and_room": "EMPA", "url_place_and_room": "", "url_online_room": "", "spoken_languages": [], "speaker": "<a href=\"mailto:[email protected]\"><strong>Hendrik Constantin JANSEN</strong></a>", "organizer": "", "contact": "<a href=\"mailto:[email protected]\"><strong>Hendrik Constantin JANSEN</strong></a>", "is_internal": "False", "theme": "", "vulgarization": { "id": 1, "fr_label": "Tout public", "en_label": "General public" }, "registration": { "id": 3, "fr_label": "Entrée libre", "en_label": "Free" }, "keywords": "EDMX", "file": null, "icalendar_url": "https://memento.epfl.ch/event/export/122664/", "category": { "id": 12, "code": "SOUTE", "fr_label": "Soutenances de thèses", "en_label": "Thesis defenses", "activated": true }, "academic_calendar_category": null, "domains": [], "mementos": [ "https://memento.epfl.ch/api/v1/mementos/1/?format=api", "https://memento.epfl.ch/api/v1/mementos/6/?format=api" ] }, { "id": 72508, "title": "How attention learns structure form data?", "slug": "how-attention-learns-structure-form-data-2", "event_url": "https://memento.epfl.ch/event/how-attention-learns-structure-form-data-2", "visual_url": "https://memento.epfl.ch/image/33754/200x112.jpg", "visual_large_url": "https://memento.epfl.ch/image/33754/720x405.jpg", "visual_maxsize_url": "https://memento.epfl.ch/image/33754/max-size.jpg", "lang": "en", "start_date": "2026-11-13", "end_date": "2026-11-13", "start_time": "15:15:00", "end_time": "16:45:00", "description": "<p>Transformer architectures have demonstrated remarkable empirical success, yet their ability to extract statistical structure from data remains only partially understood. In this talk, I will present recent work shedding light on attention mechanisms through a statistical and operator-theoretic lens.<br>\r\nBased on large-prompt asymptotics, one can show that softmax attention converges to a linear operator acting on the input-token distribution under Gaussian assumptions. This regime enables a precise analysis of both the outputs and the training dynamics using concentration arguments, revealing a surprising bridge between nonlinear softmax attention and tractable linear models.<br>\r\nThese results suggest that attention can be understood as a flexible statistical operator that adapts to the underlying data distribution, providing a unifying framework to study its role in representation learning and in-context inference. I will conclude by discussing the connections between transformer architectures and principal component analysis, using this setting to illustrate the power of concentration methods in high-dimensional statistics.<br>\r\n </p>", "image_description": "", "creation_date": "2026-09-03T14:50:19", "last_modification_date": "2026-11-05T14:45:02", "link_label": "", "link_url": "https://www.epfl.ch/schools/sb/research/math/statistics-seminars/", "canceled": "False", "cancel_reason": "", "place_and_room": "CM 1 517", "url_place_and_room": "https://plan.epfl.ch/?room==CM%201%20517", "url_online_room": "", "spoken_languages": [ "https://memento.epfl.ch/api/v1/spoken_languages/2/?format=api" ], "speaker": "Claire Boyer, Université Paris-Saclay", "organizer": "Myrto Limnios", "contact": "Maroussia Schaffner", "is_internal": "False", "theme": "", "vulgarization": { "id": 2, "fr_label": "Public averti", "en_label": "Informed public" }, "registration": { "id": 3, "fr_label": "Entrée libre", "en_label": "Free" }, "keywords": "", "file": null, "icalendar_url": "https://memento.epfl.ch/event/export/121686/", "category": { "id": 1, "code": "CONF", "fr_label": "Conférences - Séminaires", "en_label": "Conferences - Seminars", "activated": true }, "academic_calendar_category": null, "domains": [], "mementos": [ "https://memento.epfl.ch/api/v1/mementos/288/?format=api", "https://memento.epfl.ch/api/v1/mementos/1/?format=api", "https://memento.epfl.ch/api/v1/mementos/5/?format=api", "https://memento.epfl.ch/api/v1/mementos/7/?format=api", "https://memento.epfl.ch/api/v1/mementos/6/?format=api" ] }, { "id": 72511, "title": "TBA", "slug": "tba-136", "event_url": "https://memento.epfl.ch/event/tba-136", "visual_url": "https://memento.epfl.ch/image/33758/200x112.jpg", "visual_large_url": "https://memento.epfl.ch/image/33758/720x405.jpg", "visual_maxsize_url": "https://memento.epfl.ch/image/33758/max-size.jpg", "lang": "en", "start_date": "2026-11-27", "end_date": "2026-11-27", "start_time": "15:15:00", "end_time": "16:45:00", "description": "<p>tba</p>", "image_description": "", "creation_date": "2026-09-03T15:11:47", "last_modification_date": "2026-09-03T15:13:14", "link_label": "", "link_url": "https://www.epfl.ch/schools/sb/research/math/statistics-seminars/", "canceled": "False", "cancel_reason": "", "place_and_room": "CM 1 517", "url_place_and_room": "https://plan.epfl.ch/?room==CM%201%20517", "url_online_room": "", "spoken_languages": [ "https://memento.epfl.ch/api/v1/spoken_languages/2/?format=api" ], "speaker": "Torben Martinussen, University of Copenhagen", "organizer": "Myrto Limnios", "contact": "Maroussia Schaffner", "is_internal": "False", "theme": "", "vulgarization": { "id": 2, "fr_label": "Public averti", "en_label": "Informed public" }, "registration": { "id": 3, "fr_label": "Entrée libre", "en_label": "Free" }, "keywords": "", "file": null, "icalendar_url": "https://memento.epfl.ch/event/export/121692/", "category": { "id": 1, "code": "CONF", "fr_label": "Conférences - Séminaires", "en_label": "Conferences - Seminars", "activated": true }, "academic_calendar_category": null, "domains": [], "mementos": [ "https://memento.epfl.ch/api/v1/mementos/288/?format=api", "https://memento.epfl.ch/api/v1/mementos/1/?format=api", "https://memento.epfl.ch/api/v1/mementos/5/?format=api", "https://memento.epfl.ch/api/v1/mementos/6/?format=api" ] } ] }
