Event List
retrieve:
Return the details about the given Event id.
list:
List all Event objects.
GET /api/v1/events/?format=api&offset=90&ordering=-event__contact
{ "count": 203, "next": "https://memento.epfl.ch/api/v1/events/?format=api&limit=10&offset=100&ordering=-event__contact", "previous": "https://memento.epfl.ch/api/v1/events/?format=api&limit=10&offset=80&ordering=-event__contact", "results": [ { "id": 71754, "title": "Towards Understanding Large Language Models", "slug": "towards-understanding-large-language-models", "event_url": "https://memento.epfl.ch/event/towards-understanding-large-language-models", "visual_url": "https://memento.epfl.ch/image/33080/200x112.jpg", "visual_large_url": "https://memento.epfl.ch/image/33080/720x405.jpg", "visual_maxsize_url": "https://memento.epfl.ch/image/33080/max-size.jpg", "lang": "en", "start_date": "2026-06-12", "end_date": "2026-06-12", "start_time": "15:30:00", "end_time": "17:30:00", "description": "<u>EDIC candidacy exam</u><br>\r\nExam president: Prof. Robert West<br>\r\nThesis advisor: Prof. Nicolas Flammarion<br>\r\nCo-examiner: Prof. Martin Jaggi<br>\r\n<br>\r\n<u>Abstract</u><br>\r\ncoming soon<br>\r\n<br>\r\n<u>Selected papers</u><br>\r\ncoming soon", "image_description": "", "creation_date": "2026-04-29T09:57:35", "last_modification_date": "2026-05-08T11:16:06", "link_label": "", "link_url": "", "canceled": "False", "cancel_reason": "", "place_and_room": "BC 329", "url_place_and_room": "https://plan.epfl.ch/?room==BC%20329", "url_online_room": "", "spoken_languages": [], "speaker": "Mark Rofin", "organizer": "", "contact": "[email protected]", "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": "EDIC candidacy exam", "file": null, "icalendar_url": "https://memento.epfl.ch/event/export/120586/", "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/6/?format=api", "https://memento.epfl.ch/api/v1/mementos/27/?format=api", "https://memento.epfl.ch/api/v1/mementos/48/?format=api" ] }, { "id": 71755, "title": "Controllable Generative Models via Test-Time Adaptation", "slug": "controllable-generative-models-via-test-time-adapt", "event_url": "https://memento.epfl.ch/event/controllable-generative-models-via-test-time-adapt", "visual_url": "https://memento.epfl.ch/image/33081/200x112.jpg", "visual_large_url": "https://memento.epfl.ch/image/33081/720x405.jpg", "visual_maxsize_url": "https://memento.epfl.ch/image/33081/max-size.jpg", "lang": "en", "start_date": "2026-07-07", "end_date": "2026-07-07", "start_time": "09:30:00", "end_time": "11:30:00", "description": "<u>EDIC candidacy exam</u><br>\r\nExam president: Prof. Martin Rajman<br>\r\nThesis advisor: Prof. Pascal Fua<br>\r\nCo-examiner: Prof. Jean-Philippe Thiran<br>\r\n<br>\r\n<u>Abstract</u><br>\r\ncoming soon<br>\r\n<br>\r\n<u>Selected papers</u><br>\r\ncoming soon", "image_description": "", "creation_date": "2026-04-29T10:00:51", "last_modification_date": "2026-05-08T11:17:00", "link_label": "", "link_url": "", "canceled": "False", "cancel_reason": "", "place_and_room": "BC 329", "url_place_and_room": "https://plan.epfl.ch/?room==BC%20329", "url_online_room": "", "spoken_languages": [], "speaker": "Hantao Zhang", "organizer": "", "contact": "[email protected]", "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": "EDIC candidacy exam", "file": null, "icalendar_url": "https://memento.epfl.ch/event/export/120587/", "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/6/?format=api", "https://memento.epfl.ch/api/v1/mementos/27/?format=api", "https://memento.epfl.ch/api/v1/mementos/48/?format=api" ] }, { "id": 71834, "title": "Language-Based Safety and Isolation Across System Composition Boundaries", "slug": "language-based-safety-and-isolation-across-system", "event_url": "https://memento.epfl.ch/event/language-based-safety-and-isolation-across-system", "visual_url": "https://memento.epfl.ch/image/33154/200x112.jpg", "visual_large_url": "https://memento.epfl.ch/image/33154/720x405.jpg", "visual_maxsize_url": "https://memento.epfl.ch/image/33154/max-size.jpg", "lang": "en", "start_date": "2026-06-19", "end_date": "2026-06-19", "start_time": "09:00:00", "end_time": "11:00:00", "description": "<u>DIC candidacy exam</u><br>\r\nExam president: Prof. Anne-Marie Kermarrec<br>\r\nThesis advisor: Prof. Mathias Payer<br>\r\nCo-examiner: Prof. Thomas Bourgeat<br>\r\n<br>\r\n<u>Abstract</u><br>\r\nThe OS kernel was exposed due to vulnerable drivers as one vulnerability could bring down the whole system. Since fixing all driver bugs is nearly impossible, isolation of drivers is necessary to be a fundamental mechanism for reducing the attack surface between system composition boundaries. Compared with hardware-based isolation(Xen, Nooks, etc) and software-based isolation (e.g. SFI), language-based isolation often gains fine-grained safety, such as type constraints and thread-safety properties. <br>\r\nSafeDrive was an early effort to protect Linux C drivers using type checks, memory-safety invariants, and recovery support, but it still had overhead and incomplete safety. RedLeaf, built in Rust completely, reduces runtime checks by relying on Rust's ownership system while enabling efficient communication through validated interfaces. As Rust enters mainline Linux, a new challenge appears: Rust components need to interact with legacy foreign-language code, where safety guarantees can break. Omniglot addresses this by enabling safe, zero-copy interaction between Rust and unmodified untrusted code using carefully designed memory isolation primitives.<br>\r\nGoing forward, my research will continue to focus on the intersection of operating systems and programming languages.<br>\r\n<br>\r\n<u>Selected papers</u><br>\r\n<strong>Paper 1: </strong>SafeDrive: Safe and Recoverable Extensions Using Language-Based Techniques [OSDI'06]\r\n<ul>\r\n\t<li><a href=\"https://dl.acm.org/doi/10.5555/1298455.1298461\">https://dl.acm.org/doi/10.5555/1298455.1298461</a></li>\r\n</ul>\r\n<strong>Paper 2: </strong>RedLeaf: Isolation and Communication in a Safe Operating System [OSDI'20]\r\n\r\n<ul>\r\n\t<li><a href=\"https://www.usenix.org/conference/osdi20/presentation/narayanan-vikram\">https://www.usenix.org/conference/osdi20/presentation/narayanan-vikram</a></li>\r\n</ul>\r\n<strong>Paper 3: </strong>Building Bridges: Safe Interactions with Foreign Languages through Omniglot [OSDI'25]\r\n\r\n<ul>\r\n\t<li><a href=\"https://www.usenix.org/conference/osdi25/presentation/schuermann\">https://www.usenix.org/conference/osdi25/presentation/schuermann</a></li>\r\n</ul>", "image_description": "", "creation_date": "2026-05-08T11:09:55", "last_modification_date": "2026-05-08T11:10:13", "link_label": "", "link_url": "", "canceled": "False", "cancel_reason": "", "place_and_room": "BC 133", "url_place_and_room": "https://plan.epfl.ch/?room==BC%20133", "url_online_room": "", "spoken_languages": [], "speaker": "Yiwen Xu", "organizer": "", "contact": "[email protected]", "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": "EDIC candidacy exam", "file": null, "icalendar_url": "https://memento.epfl.ch/event/export/120700/", "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/6/?format=api", "https://memento.epfl.ch/api/v1/mementos/27/?format=api", "https://memento.epfl.ch/api/v1/mementos/48/?format=api" ] }, { "id": 71835, "title": "Testing the Hypervisor Boundary: Semantic Oracles and Adversarial Input Generation for pKVM", "slug": "testing-the-hypervisor-boundary-semantic-oracles-a", "event_url": "https://memento.epfl.ch/event/testing-the-hypervisor-boundary-semantic-oracles-a", "visual_url": "https://memento.epfl.ch/image/33155/200x112.jpg", "visual_large_url": "https://memento.epfl.ch/image/33155/720x405.jpg", "visual_maxsize_url": "https://memento.epfl.ch/image/33155/max-size.jpg", "lang": "en", "start_date": "2026-06-22", "end_date": "2026-06-22", "start_time": "13:00:00", "end_time": "15:00:00", "description": "<u>EDIC candidacy exam</u><br>\r\nExam president: Prof. Katerina Argyraki<br>\r\nThesis advisor: Prof. Mathias Payer<br>\r\nCo-examiner: Prof. Thomas Bourgeat<br>\r\n<br>\r\n<u>Abstract</u><br>\r\nHypervisors underpin modern confidential computing stacks, yet their exposed interfaces - hypercalls, fault handlers, and device interactions - remain a large and under-tested attack surface. Existing approaches fall short: coverage-driven fuzzers (e.g., HYPERPILL) explore broadly but rely on crash oracles, missing non-crashing vulnerabilities, while specification-based systems (e.g., Ghost) detect semantic violations but constrain input generation and under explore adversarial cases.<br>\r\n<br>\r\nThis thesis focuses on systematic bug finding in pKVM by combining semantic oracles with coverage-guided, adversarial fuzzing. Executable interface specifications are used to flag behavioral inconsistencies, while mutation-based input generation deliberately produces both valid and model-violating inputs to reach boundary conditions and deep execution paths. The approach further incorporates EL2 coverage feedback and concurrent vCPU sequences to expose bugs such as silent state corruption, ownership violations, and TOCTOU races.<br>\r\n<br>\r\nThe goal is to uncover security-critical vulnerabilities that evade both crash-based fuzzing and model-constrained testing, particularly in unspecified or adversarial input classes. More broadly, the work aims to establish a practical methodology for bug-driven testing of production hypervisors.<br>\r\n<br>\r\n<u>Selected papers</u><br>\r\ncoming soon", "image_description": "", "creation_date": "2026-05-08T11:14:42", "last_modification_date": "2026-05-08T11:14:42", "link_label": "", "link_url": "", "canceled": "False", "cancel_reason": "", "place_and_room": "BC 133", "url_place_and_room": "https://plan.epfl.ch/?room==BC%20133", "url_online_room": "", "spoken_languages": [], "speaker": "Sofiia Saltovskaia", "organizer": "", "contact": "[email protected]", "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": "EDIC candidacy exam", "file": null, "icalendar_url": "https://memento.epfl.ch/event/export/120701/", "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/6/?format=api", "https://memento.epfl.ch/api/v1/mementos/27/?format=api", "https://memento.epfl.ch/api/v1/mementos/48/?format=api" ] }, { "id": 71838, "title": "Composable Performance Interfaces with Controlled Resolution", "slug": "composable-performance-interfaces-with-controlled", "event_url": "https://memento.epfl.ch/event/composable-performance-interfaces-with-controlled", "visual_url": "https://memento.epfl.ch/image/33158/200x112.jpg", "visual_large_url": "https://memento.epfl.ch/image/33158/720x405.jpg", "visual_maxsize_url": "https://memento.epfl.ch/image/33158/max-size.jpg", "lang": "en", "start_date": "2026-06-26", "end_date": "2026-06-26", "start_time": "11:15:00", "end_time": "13:15:00", "description": "<u>EDIC candidacy exam</u><br>\r\nExam president: Prof. Nate Foster<br>\r\nThesis advisor: Prof. George Candea<br>\r\nCo-examiner: Prof. Rachid Guerraoui<br>\r\n<br>\r\n<u>Abstract</u><br>\r\ncoming soon<br>\r\n<br>\r\n<u>Selected papers</u><br>\r\ncoming soon", "image_description": "", "creation_date": "2026-05-08T16:49:17", "last_modification_date": "2026-05-08T16:49:17", "link_label": "", "link_url": "", "canceled": "False", "cancel_reason": "", "place_and_room": "INN 326", "url_place_and_room": "https://plan.epfl.ch/?room==INN%20326", "url_online_room": "", "spoken_languages": [], "speaker": "Alexandre Piveteau", "organizer": "", "contact": "[email protected]", "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": "EDIC candidacy exam", "file": null, "icalendar_url": "https://memento.epfl.ch/event/export/120704/", "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/6/?format=api", "https://memento.epfl.ch/api/v1/mementos/27/?format=api", "https://memento.epfl.ch/api/v1/mementos/48/?format=api" ] }, { "id": 71877, "title": "Aligning Models Throughout Training", "slug": "aligning-models-throughout-training", "event_url": "https://memento.epfl.ch/event/aligning-models-throughout-training", "visual_url": "https://memento.epfl.ch/image/33199/200x112.jpg", "visual_large_url": "https://memento.epfl.ch/image/33199/720x405.jpg", "visual_maxsize_url": "https://memento.epfl.ch/image/33199/max-size.jpg", "lang": "en", "start_date": "2026-06-18", "end_date": "2026-06-18", "start_time": "13:00:00", "end_time": "15:00:00", "description": "<u>EDIC candidacy exam</u><br>\r\nExam president: Prof. Martin Jaggi<br>\r\nThesis advisor: Prof. Robert West<br>\r\nCo-examiner: Prof. Nicolas Flammarion<br>\r\n<br>\r\n<u>Abstract</u><br>\r\ncoming soon<br>\r\n<br>\r\n<u>Selected papers</u><br>\r\ncoming soon", "image_description": "", "creation_date": "2026-05-13T12:37:39", "last_modification_date": "2026-05-13T12:37:39", "link_label": "", "link_url": "", "canceled": "False", "cancel_reason": "", "place_and_room": "BC 333", "url_place_and_room": "https://plan.epfl.ch/?room==BC%20333", "url_online_room": "", "spoken_languages": [], "speaker": "Raghav Singhal", "organizer": "", "contact": "[email protected]", "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": "EDIC candidacy exam", "file": null, "icalendar_url": "https://memento.epfl.ch/event/export/120764/", "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/6/?format=api", "https://memento.epfl.ch/api/v1/mementos/27/?format=api", "https://memento.epfl.ch/api/v1/mementos/48/?format=api" ] }, { "id": 71930, "title": "Generative models for black-box optimization of complex objectives", "slug": "generative-models-for-black-box-optimization-of-co", "event_url": "https://memento.epfl.ch/event/generative-models-for-black-box-optimization-of-co", "visual_url": null, "visual_large_url": null, "visual_maxsize_url": null, "lang": "en", "start_date": "2026-07-09", "end_date": "2026-07-09", "start_time": "10:00:00", "end_time": "12:00:00", "description": "<u>EDIC candidacy exam</u><br>\r\nExam president: Prof. Nicolas Flammarion<br>\r\nThesis advisor: Prof. Pascal Fua<br>\r\nCo-examiner: Prof. Alexander Mathis<br>\r\n<br>\r\n<u>Abstract</u><br>\r\nMany high-impact problems in science and engineering reduce to optimizing a complex objective, where each evaluation is costly and often unreliable: physical experiments fail, simulators diverge, and stochastic systems return noisy feedback. Classical black-box optimizers struggle in this regime, where evaluation budgets are tight and the feasible region is hard to characterize a priori. Our work investigates how the statistical expressivity of modern generative samplers can be leveraged in black-box optimization. We propose that these samplers enable efficient navigation of feasible sets and optimization of geometrically complex objectives under limited, unreliable observations. <br>\r\n<br>\r\nTwo complementary lines of work have already been studied. The main one, SPARROW, demonstrated that a sampler-driven optimizer can substantially reduce the number of function calls needed to maximize complex and unreliable objectives. A parallel line established that samplers can be steered to satisfy hard constraints without sacrificing the statistical diversity of the generated samples. Because SPARROW imposes minimal assumptions on the underlying sampler, these constrained samplers can drive it directly, opening a path to constrained black-box optimization.<br>\r\n<br>\r\nBuilding on these results, two directions are planned. The first is theoretical: a deeper mathematical study of sampler-driven optimization, drawing on the broader literature on sample-efficient black-box optimization, in order to derive stronger guarantees and more efficient algorithms. The second is applied: applying the methods to concrete domains and exploiting the structure those domains expose to sharpen performance beyond what a pure black-box treatment can achieve.<br>\r\n<br>\r\nThe applications of low-budget black-box optimization are diverse, including engineering design (e.g. aerodynamic shape optimization, mechanical components), scientific discovery (e.g. protein and molecule design under synthesizability and stability constraints), and the safety and security of opaque ML systems (e.g. query-efficient adversarial probing, red-teaming of black-box models).<br>\r\n<br>\r\n<u>Selected papers</u><br>\r\n ", "image_description": "", "creation_date": "2026-05-22T12:53:32", "last_modification_date": "2026-05-22T13:09:04", "link_label": "", "link_url": "", "canceled": "False", "cancel_reason": "", "place_and_room": "BC 333", "url_place_and_room": "https://plan.epfl.ch/?room==BC%20333", "url_online_room": "", "spoken_languages": [], "speaker": "Edouard Dufour", "organizer": "", "contact": "[email protected]", "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": "EDIC candidacy exam", "file": null, "icalendar_url": "https://memento.epfl.ch/event/export/120837/", "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/48/?format=api", "https://memento.epfl.ch/api/v1/mementos/6/?format=api", "https://memento.epfl.ch/api/v1/mementos/27/?format=api" ] }, { "id": 71931, "title": "Access Bit Scanning for Page Replacement Policies", "slug": "access-bit-scanning-for-page-replacement-policies", "event_url": "https://memento.epfl.ch/event/access-bit-scanning-for-page-replacement-policies", "visual_url": null, "visual_large_url": null, "visual_maxsize_url": null, "lang": "en", "start_date": "2026-06-22", "end_date": "2026-06-22", "start_time": "10:00:00", "end_time": "12:00:00", "description": "<u>EDIC candidacy exam</u><br>\r\nExam president: Prof. Thomas Bourgeat<br>\r\nThesis advisor: Prof. Babak Falsafi<br>\r\nCo-examiner: Prof. Sanidhya Kashyap<br>\r\n<br>\r\n<u>Abstract</u><br>\r\nModern workloads with memory footprints ranging from hundreds of gigabytes to several terabytes have become widespread over the last decade. With memory accounting for nearly 50% of server costs in modern data centres, provisioning large DRAM capacities significantly increases total cost of ownership (TCO). Tiered memory architectures combine memories with different capacities, latencies, and cost characteristics to improve cost efficiency while satisfying application performance requirements. Application performance in these architectures is highly sensitive to page placement decisions. Consequently, making informed placement decisions requires precise tracking of data access patterns and accurate identification of hot pages.<br>\r\n<br>\r\nExisting operating systems commonly rely on hardware-maintained access bits in page table entries to estimate memory access behaviour and identify hot data regions. They periodically scan the application's entire virtual address space to construct workload access profiles. However, linear scanning of page table access bits does not scale for modern applications with terabyte-scale memory footprints. Since scanning overhead scales with the memory footprint, a single full scan can take several seconds for workloads with large memory footprints. Reducing scanning frequency lowers monitoring overhead but also decreases profiling accuracy and responsiveness to changing access patterns, resulting in a fundamental tradeoff between profiling accuracy and monitoring overhead. Therefore, we need scalable, low-overhead mechanisms for access monitoring in next-generation memory management systems.<br>\r\n<br>\r\nIn this talk, I will discuss recent approaches to scalable memory access profiling in large-memory systems. Specifically, I will present works that use techniques such as adaptive sampling, region-based monitoring, and hierarchical profiling of page table trees to reduce and bound scanning overhead. <br>\r\n<br>\r\n<u>Selected papers</u><br>\r\ncoming soon", "image_description": "", "creation_date": "2026-05-22T13:08:29", "last_modification_date": "2026-05-22T13:18:11", "link_label": "", "link_url": "", "canceled": "False", "cancel_reason": "", "place_and_room": "BC 010", "url_place_and_room": "https://plan.epfl.ch/?room==BC%20010", "url_online_room": "", "spoken_languages": [], "speaker": "<br>\r\nArunkrishna Annai Madalam Sivasubramanian ", "organizer": "", "contact": "[email protected]", "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": "EDIC candidacy exam", "file": null, "icalendar_url": "https://memento.epfl.ch/event/export/120838/", "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/48/?format=api", "https://memento.epfl.ch/api/v1/mementos/6/?format=api", "https://memento.epfl.ch/api/v1/mementos/27/?format=api" ] }, { "id": 71932, "title": "Failure-Aware LLM Serving for Agentic Workloads", "slug": "failure-aware-llm-serving-for-agentic-workloads", "event_url": "https://memento.epfl.ch/event/failure-aware-llm-serving-for-agentic-workloads", "visual_url": null, "visual_large_url": null, "visual_maxsize_url": null, "lang": "en", "start_date": "2026-06-17", "end_date": "2026-06-17", "start_time": "09:00:00", "end_time": "11:00:00", "description": "<u>EDIC candidacy exam</u><br>\r\nExam president: Prof. Anastasia Ailamaki<br>\r\nThesis advisor: Prof. Anne-Marie Kermarrec<br>\r\nCo-examiner: Prof. Robert West<br>\r\n<br>\r\n<u>Abstract</u><br>\r\n<br>\r\n<br>\r\n<u>Selected papers</u>\r\n<ul>\r\n\t<li>ReAct (<a href=\"https://arxiv.org/pdf/2210.03629\" target=\"_blank\">https://arxiv.org/pdf/2210.03629</a>)</li>\r\n\t<li>Agentix (<a href=\"https://www.usenix.org/system/files/nsdi26-luo.pdf\" target=\"_blank\">https://www.usenix.org/system/files/nsdi26-luo.pdf</a>)</li>\r\n\t<li>Why do multi-agent systems fail? (<a href=\"https://arxiv.org/pdf/2503.13657\" target=\"_blank\">https://arxiv.org/pdf/2503.13657</a>)</li>\r\n</ul>", "image_description": "", "creation_date": "2026-05-22T13:11:09", "last_modification_date": "2026-05-22T13:12:05", "link_label": "", "link_url": "", "canceled": "False", "cancel_reason": "", "place_and_room": "BC 133", "url_place_and_room": "https://plan.epfl.ch/?room==BC%20133", "url_online_room": "", "spoken_languages": [], "speaker": "Palak", "organizer": "", "contact": "[email protected]", "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": "EDIC candidacy exam", "file": null, "icalendar_url": "https://memento.epfl.ch/event/export/120839/", "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/48/?format=api", "https://memento.epfl.ch/api/v1/mementos/6/?format=api", "https://memento.epfl.ch/api/v1/mementos/27/?format=api" ] }, { "id": 71933, "title": "Controlling and Guiding Generative Models for Three-Dimensional Shape Optimization", "slug": "controlling-and-guiding-generative-models-for-thre", "event_url": "https://memento.epfl.ch/event/controlling-and-guiding-generative-models-for-thre", "visual_url": null, "visual_large_url": null, "visual_maxsize_url": null, "lang": "en", "start_date": "2026-07-07", "end_date": "2026-07-07", "start_time": "11:00:00", "end_time": "13:00:00", "description": "<u>EDIC candidacy exam</u><br>\r\nExam president: Prof. Martin Rajman<br>\r\nThesis advisor: Prof. Pascal Fua<br>\r\nCo-examiner: Prof. Mark Pauly<br>\r\n<br>\r\n<u>Abstract</u><br>\r\nOptimizing 3D shapes within the latent spaces of deep generative models is fundamental to computer assisted engineering, yet remains prone to a critical failure mode we term manifold drift: the tendency of gradient-based optimization to move latent vectors away from the manifold of valid shapes. This problem is exacerbated in state-of-the-art 3D shape generative models that operate in increasingly high-dimensional latent spaces where valid shapes occupy a vanishingly small fraction of the full space. Existing mitigation strategies, including latent regularization and flow-matching approaches, either sacrifice expressiveness, demand a difficult trade-off between objective guidance and generative fidelity that remains prone to manifold drift, or are computationally infeasible to scale to modern, large-capacity 3D shape models. We introduce a novel optimizer-corrector framework that alternates between gradient steps for objective minimization and guided flow matching to drive the latent state back to the valid shape manifold. By decoupling objective minimization from flow-based correction, optimizing freely and correcting strictly, this alternating design avoids inherent trade-offs, preserving geometric validity without sacrificing expressiveness while remaining computationally feasible on modern 3D shape models. We demonstrate its effectiveness across generative priors of varying complexity, from simple vector latent spaces to large-scale architectures across a variety of downstream optimization tasks, including aerodynamic drag reduction and object compliance optimization.<br>\r\n<br>\r\n<u>Selected papers</u><br>\r\n ", "image_description": "", "creation_date": "2026-05-22T13:14:54", "last_modification_date": "2026-05-22T13:14:54", "link_label": "", "link_url": "", "canceled": "False", "cancel_reason": "", "place_and_room": "BC 333", "url_place_and_room": "https://plan.epfl.ch/?room==BC%20333", "url_online_room": "", "spoken_languages": [], "speaker": "Emilien Seiler", "organizer": "", "contact": "[email protected]", "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": "EDIC candidacy exam", "file": null, "icalendar_url": "https://memento.epfl.ch/event/export/120840/", "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/48/?format=api", "https://memento.epfl.ch/api/v1/mementos/6/?format=api", "https://memento.epfl.ch/api/v1/mementos/27/?format=api" ] } ] }
