Thesis Director: Prof. G. Carleo,
\r\nPhysics doctoral program
\r\nThesis Nr. 12057
\r\n
\r\nTo take part in the public defense, please contact directly the speaker
The Mary Ann Mansigh Conversation series focuses of non-strictly technical topics of cultural interest for the simulation and modelling community. The format reflects the informative and informal nature of these sessions, with talks introducing the subject followed by a conversation between the speakers and the audience.
\r\nThe lecture series \"Mary Ann Mansigh Conversation series\" is co-organized by CECAM (https://www.cecam.org) and MARVEL (http://nccr-marvel.ch/) at EPFL.
\r\n
\r\nTuesday October 27, 2026, 15:00-18:00 CET
\r\n
\r\nArtificial intelligence is changing how algorithms are conceived, how software is written, and how its outputs are verified. This conversation takes stock of where the field of simulation and modelling stands today and opens a discussion on future directions and best practices — how to safeguard reliability and scientific creativity, how to adapt notions of authorship and ownership for AI-assisted code, and how to preserve and ideally enhance the rich commercial and academic software ecosystem that has driven the growth of computational science in research and development.
\r\n
\r\nWe invite submissions for contributed talks. If you would like to propose a contribution, please submit your abstract using this form. Deadline October 13, 2026.
\r\n
\r\nThe conversation will take place at EPFL, in room BCH 2103 and online.
\r\n
\r\nZoom link https://epfl.zoom.us/j/68443642155?pwd=ZPjdZw2KmHzeQ9Pu376iLb0OyMXCV8.1
\r\n
\r\nTentative program
\r\n15:00 Welcome and introduction
\r\n15:10 Georg Kresse, University of Vienna, Austria
\r\n15:30 Miguel Marques Ruhr, University Bochum, Germany
\r\n15:50 Michele Ceriotti, EPFL, Switzerland
\r\n16:10 Short break
\r\n16:20 Contributed speaker 1
\r\n16:35 Contributed speaker 2
\r\n16:50 Contributed speaker 3
\r\n17:05 Discussion
\r\n17:30 Apero for in person participants
\r\n
\r\nPrevious CECAM and MARVEL lectures can be found at
\r\nhttps://www.materialscloud.org/learn/sections/Btmngu/marvel-events
\r\nhttps://www.cecam.org/lectures
We discuss a class of graph embeddings into Minkowski space R^{2,2} = C^{1,1}, called t-surfaces, which arise in the study of the planar dimer model. A t-surface consists of a (perfect) t-embedding together with its associated origami map. Perfect t-embeddings were recently introduced as a key tool for proving that the gradient of the dimer height function converges to that of the Gaussian Free Field in a canonically associated metric, under suitable technical assumptions. After introducing the notion of a t-embedding, we will describe a construction of perfect t-embeddings for regular hexagons of the hexagonal lattice. In which the corresponding t-surfaces converge to space-like maximal surfaces in Minkowski space R^{2,1}. As a consequence, these constructions yield a new proof of convergence of fluctuations of the dimer height function to the Gaussian Free Field in the conformal structure induced by the limiting maximal surface.
","image_description":"","creation_date":"2026-10-01T16:38:56","last_modification_date":"2026-10-01T16:38:56","link_label":"","link_url":"","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=json"],"speaker":"Marianna Russkikh","organizer":"Prof. Martin Hairer","contact":"Juliana Velasquez","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/122018/","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/7/?format=json","https://memento.epfl.ch/api/v1/mementos/403/?format=json"]},{"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":"Thesis Director: Prof. Ph. Schwaller,
\r\nChemistry and Chemical Engineering doctoral program
\r\nThesis Nr. 11943
\r\n
\r\nTo take part in the public defense, please contact directly the speaker
The talk is jointly organized by the EPFL AI Center and the MLO Lab as part of the AI fundamentals seminar series.
\r\n
\r\nHosting Professor: Prof. Martin Jaggi
\r\n
\r\nTitle
\r\nSpeedrunning SWE-bench, or the $1,000 coding agent
\r\n
\r\nAbstract
\r\nSpeedrunning has become a popular engineering challenge in machine learning: achieve the best performance within a fixed compute budget. We use it as a scientific tool to study how models acquire specialized skills and how far those skills transfer. We speedrun SWE-bench by training coding agents from scratch on demonstrations from stronger models. With about $1,000 of training compute, we reach a 25% solve rate, outperforming GPT-4o. Compared to models with extensive pre-training, our speedrun models solve largely the same SWE-bench problems, and generalize about as well to new bugs and codebases. What, then, does pre-training buy for SWE-bench? Our experiments suggest a one-student hypothesis: pre-training helps a student imitate its teacher, but at matched imitation loss, pre-training choices barely matter. We stress-test this hypothesis by comparing a model trained on the web against a “vintage\" one trained only pre-1931 documents. Up to the SWE-bench scores we reach, pre-training is interchangeable with specialization data. We release nanoswe as an open project for studying specialization through speedrunning.
\r\n
\r\nBio
\r\nRicardo Olmedo is a final-year PhD student at the Max Planck Institute for Intelligent Systems in Tübingen and a Google PhD Fellow. Advised by Moritz Hardt and Bernhard Schölkopf, he works on language model benchmarking and specialization. His research has earned three oral presentations at ICLR and NeurIPS. Most recently, he spent spring and summer 2026 visiting Sanmi Koyejo at Stanford University.
\r\nEPFL Investor Day 2026 brings together leading investors and ambitious deeptech startups from EPFL and across Western Switzerland.
\r\n
\r\nTaking place on Wednesday, 25 November 2026 at the SwissTech Convention Center, the event is structured in two parts:\r\n
tba
","image_description":"","creation_date":"2026-09-03T14:50:19","last_modification_date":"2026-09-03T15:07:40","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=json"],"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/1/?format=json","https://memento.epfl.ch/api/v1/mementos/5/?format=json","https://memento.epfl.ch/api/v1/mementos/6/?format=json","https://memento.epfl.ch/api/v1/mementos/7/?format=json","https://memento.epfl.ch/api/v1/mementos/288/?format=json"]}]}