Elena Burceanu (@ilarele) header
Elena Burceanu (@ilarele) profile picture

Interested in Generalization problems under Distribution Shifts in our awesome Theoretical Research team at Bitdefender https://teapaw.org/t.co/pqMUlODZc1

Bucharest, Romania
Joined January 2010
Photo by Elena Burceanu (@ilarele)Photo by Elena Burceanu (@ilarele)Photo by Elena Burceanu (@ilarele)Photo by Elena Burceanu (@ilarele)Photo by Elena Burceanu (@ilarele)Photo by Elena Burceanu (@ilarele)Photo by Elena Burceanu (@ilarele)Photo by Elena Burceanu (@ilarele)Photo by Elena Burceanu (@ilarele)Photo by Elena Burceanu (@ilarele)
Elena Burceanu retweeted
In our new NeurIPS paper, we show how to use epiplexity to select and generate data for out-of-distribution generalization! Paper: arxiv.org/abs/2608.11746 Joint work with Andres Potapczynski, @ShikaiQiu @edwardfhughes and @andrewgwils ๐Ÿงต
19
51
7
341
29,817
Elena Burceanu retweeted
1/9 Current optimizers are static during training. What if the optimizer could think and react while the network trains? We let an agent watch the training dynamics and rewrite the optimizer's source code as training goes: a reactive optimizer. In every language and image setting we tried that has no tuned recipe, it matches or beats the best recipe one can one-shot, and it guards against loss spikes and overfitting by itself. In the near future, optimization decisions may belong to agents, not humans. ๐Ÿ“ xingjianbai.com/reactive-optโ€ฆ ๐Ÿ’ป Try it on your task: github.com/xingjian-bai/opal ๐Ÿงต
79
208
72
1,605
224,778
Elena Burceanu retweeted
We formally announced Mirendil less than 4 months ago. Today, I'm excited to share an important milestone for: Pretraining at scale without a pretraining team. Mirendil's mission is to solve Recursive Self-Improvement (RSI) and use it to democratize frontier AI development, ultimately accelerating science and technology. At Mirendil, when we think about RSI, we think holistically about the company as an intelligent entity that self-improves toward its goal. From that perspective, the most powerful notion of acceleration is reducing the time and resources (number of human experts, compute, and other costs) needed to achieve ambitious goals. Given that pretraining is perhaps the most important yet most challenging part of frontier AI development, we decided to run an ambitious controlled experiment: Starting fully from scratch, can we significantly reduce the time, number of human experts, and compute resources needed to go all the way to successfully pretraining a strong coding model at scale? Mirendil-Base is the result of this experiment. Our first pretrained model was built by one researcher in 3.5 months, with less than $10M in compute. This includes all the research and small-scale experiments it took to get there, as well as the final training run. The recipe holds up as the models get bigger. For reference, Mirendil-Base is in the 100B-500B parameter range, Mirendil-Base-Large in the 500B-2T range, and Mirendil-Base-XL in the 2T-10T range. We're beginning to see a world where a single researcher, empowered by AI, can take on work that once required an entire research organization. As these systems improve, the bottleneck shifts from the size of a team to the ambition and judgment of the people directing them. This has profound implications for who gets to build AI and, ultimately, who gets to push the frontiers of science and technology forward. Weโ€™ll share independent evaluations, post-training benchmarks and API-access sooner than you think so stay tuned!
172
165
130
1,616
311,351
Elena Burceanu retweeted
Reading all those recent beautiful essays by Terry Tao, Kevin Buzzard and @wtgowers, I could not resist, but write up my deeply personal perspective on how I perceive and see things: docs.google.com/document/d/eโ€ฆ
73
287
91
1,730
522,124
Elena Burceanu retweeted
๐Ÿ“ข As #CoLLAs2026 comes to a close in Bucharest ๐Ÿ‡ท๐Ÿ‡ด, we want to send a massive THANK YOU to the team who made it happen! ๐Ÿ™Œ To our incredible Local Organizers, Elena Burceanu (Bitdefender, Politehnica University of Bucharest) & Traian Rebedea (NVIDIA, Politehnica University of Bucharest), and our dedicated General Chairs, Sarath Chandar (Mila - Quebec AI Institute, Polytechnique Montrรฉal) & Razvan Pascanu (Google DeepMind, Mila), thank you all so much for an unforgettable event! See you at CoLLAs 2027 in Montreal ๐Ÿ‡จ๐Ÿ‡ฆ! ๐ŸŒโค๏ธ
1
2
1
18
958
Elena Burceanu retweeted
๐Ÿ“ข Weโ€™ve just wrapped up Poster Session II at #CoLLAs2026! ๐Ÿš€ A huge thank you to everyone who presented and attended. We loved the insightful discussions and we hope everyone sparked some productive new connections! ๐ŸŒ๐Ÿง 
2
5
431
Elena Burceanu retweeted
๐Ÿ“ข Donโ€™t prevent forgetting: FIX it! ๐Ÿง  At #CoLLAs2026, we explored a refreshing take on Continual Learning in our second keynote of Day 1 in Bucharest. Instead of just fighting catastrophic forgetting, Joost van de Weijer showed us how drift compensation can correct representations after the fact. ๐Ÿ”„ ๐Ÿ› ๏ธ Plus: Fascinating insights into Model Merging to effectively combine independently trained task models into one! ๐ŸŽ™๏ธ Huge thanks to Joost for the very insightful talk!
2
7
369
Elena Burceanu retweeted
๐Ÿ“ข In our first keynote, Michael Bowling challenged the core foundations of Continual RL! Are MDPs, optimal policies & ergodicity assumptions holding us back? ๐Ÿค” We explored alternative foundations & new algorithms for ever-changing worlds. ๐ŸŒโœจ ๐ŸŽ™๏ธ Huge thanks for the inspiring talk!
2
5
317
Elena Burceanu retweeted
Join us July 1st at 10 am CET for our third talk where Zeynep Akata speaks on โ€œExplainable and Adaptive Multimodal AIโ€. ๐Ÿ”— Learn more: lifelong-ml.cc/seminar ๐Ÿ’ป Zoom link: polymtl-ca.zoom.us/j/8406671โ€ฆ
3
3
513
Elena Burceanu retweeted
Only a few more days until the CoLLAs Early-Career Spotlight Program deadline on June 1st!
The CoLLAs Early-Career Spotlight Program is back! Showcasing early-stage scholars with an invited 30-min in-person talk & blog feature. Submit your motivation letter, outline, & CV to [email protected] by June 1, 2026 (AoE). lifelong-ml.cc/Conferences/2โ€ฆ
3
3
828
Elena Burceanu retweeted
Have Too Many Tabs Open? G O V E R T I C A L !
Too many @GoogleChrome tabs open? Try vertical tabs, rolling out now. Just right-click any Chrome window and select โ€œShow Tabs Verticallyโ€ to move your tabs to the side of the browser window, making it easier to read page titles and manage tab groups.
100
57
24
1,363
316,573
Elena Burceanu retweeted
Just finished this riveting, fact-based, storytelling book on how AI rose in the past 15+ years to where it is today, by @scmallaby, featuring @demishassabis and @GoogleDeepMind. Join Sebastian and me for a Ground Truths live podcast tomorrow 12N PT erictopol.substack.com
8
56
5
398
34,131
Elena Burceanu retweeted
Introducing Project Glasswing: an urgent initiative to help secure the worldโ€™s most critical software. Itโ€™s powered by our newest frontier model, Claude Mythos Preview, which can find software vulnerabilities better than all but the most skilled humans. anthropic.com/glasswing
1,924
6,497
4,419
43,330
32,034,979
Elena Burceanu retweeted
Continual learning is the future of AI and @CoLLAs_Conf is the best venue to publish your state-of-the-art research in designing adaptive machine learning systems! Abstract deadline in 10 days and the conference is in Romania this year!
โฐ The CoLLAs abstract deadline is only 10 days away! We invite researchers to explore all facets of ML adaptation, from incorporating new capabilities during continuous training to efficiently removing outdated or harmful data. - ๐—”๐—ฏ๐˜€๐˜๐—ฟ๐—ฎ๐—ฐ๐˜ ๐——๐—ฒ๐—ฎ๐—ฑ๐—น๐—ถ๐—ป๐—ฒ: April 10, 2026 - ๐—ฆ๐˜‚๐—ฏ๐—บ๐—ถ๐˜€๐˜€๐—ถ๐—ผ๐—ป ๐——๐—ฒ๐—ฎ๐—ฑ๐—น๐—ถ๐—ป๐—ฒ: April 15, 2026 - ๐—–๐—ผ๐—ป๐—ณ๐—ฒ๐—ฟ๐—ฒ๐—ป๐—ฐ๐—ฒ ๐——๐—ฎ๐˜๐—ฒ๐˜€: Sep 14โ€“17, 2026 ๐Ÿ“š Accepted papers will be published in the Proceedings of Machine Learning Research (PMLR). ๐Ÿ”— ๐—™๐—ผ๐—ฟ ๐—ณ๐˜‚๐—น๐—น ๐—ฑ๐—ฒ๐˜๐—ฎ๐—ถ๐—น๐˜€ ๐—ผ๐—ป ๐˜๐—ต๐—ฒ ๐—–๐—ฎ๐—น๐—น ๐—ณ๐—ผ๐—ฟ ๐—ฃ๐—ฎ๐—ฝ๐—ฒ๐—ฟ๐˜€: lifelong-ml.cc/Conferences/2โ€ฆ
4
6
34
6,569
CoLLAs 2026 is happening in Bucharest, Romania ๐Ÿ“๐Ÿ‡ท๐Ÿ‡ด, and the abstract deadline is only 10 days away ๐Ÿ“ฃโฐ. Consider submitting if you are working on lifelong learning or adaptive ML ๐Ÿ”ฅ .
โฐ The CoLLAs abstract deadline is only 10 days away! We invite researchers to explore all facets of ML adaptation, from incorporating new capabilities during continuous training to efficiently removing outdated or harmful data. - ๐—”๐—ฏ๐˜€๐˜๐—ฟ๐—ฎ๐—ฐ๐˜ ๐——๐—ฒ๐—ฎ๐—ฑ๐—น๐—ถ๐—ป๐—ฒ: April 10, 2026 - ๐—ฆ๐˜‚๐—ฏ๐—บ๐—ถ๐˜€๐˜€๐—ถ๐—ผ๐—ป ๐——๐—ฒ๐—ฎ๐—ฑ๐—น๐—ถ๐—ป๐—ฒ: April 15, 2026 - ๐—–๐—ผ๐—ป๐—ณ๐—ฒ๐—ฟ๐—ฒ๐—ป๐—ฐ๐—ฒ ๐——๐—ฎ๐˜๐—ฒ๐˜€: Sep 14โ€“17, 2026 ๐Ÿ“š Accepted papers will be published in the Proceedings of Machine Learning Research (PMLR). ๐Ÿ”— ๐—™๐—ผ๐—ฟ ๐—ณ๐˜‚๐—น๐—น ๐—ฑ๐—ฒ๐˜๐—ฎ๐—ถ๐—น๐˜€ ๐—ผ๐—ป ๐˜๐—ต๐—ฒ ๐—–๐—ฎ๐—น๐—น ๐—ณ๐—ผ๐—ฟ ๐—ฃ๐—ฎ๐—ฝ๐—ฒ๐—ฟ๐˜€: lifelong-ml.cc/Conferences/2โ€ฆ
101
Elena Burceanu retweeted
Yann LeCun ๐Ÿค Saining Xie insane crossover of the 2 biggest visual representation researchers in the AI field โ€œBeyond Language Modeling: An Exploration of Multimodal Pretrainingโ€ Right now, most multimodal models are basically a language model with a vision adapter bolted on, so they can describe images, but they donโ€™t really think in images or video. This paper shows what happens when you do it the hard way: train one model from scratch on text, images, and video with a unified setup. They key idea is if you give the model a good visual internal format and it can use vision for both understanding and generating. Additionally, multimodal data can improve language instead of distracting it, and mixture-of-experts lets you scale visionโ€™s huge data intake without bloating everything else. This paves the way towards changing the vision paradigm from โ€œcaptioning add-onโ€ model to native multimodal foundation model.
20
144
11
923
81,284
Elena Burceanu retweeted
What Happens After We Solve Continual Learning? - Stephanie Chan youtube.com/watch?v=Ri6HNMOmโ€ฆ As we look ahead to CoLLAs 2026, weโ€™re revisiting last yearโ€™s standout keynote sessions. Call for papers: ๐Ÿ”— Abstract deadline: 10th April ๐Ÿ“… Conference: Sep 14โ€“17, 2026 in Bucharest, Romania
2
8
1,008
Elena Burceanu retweeted
Preparing for CoLLAs 2026, weโ€™re revisiting last yearโ€™s standout keynotes: Non Stationarity Online Learning and Loss of Plasticity - Andras Gyorgy youtube.com/watch?v=AtbstohOโ€ฆ Join us in Bucharest: ๐Ÿ”— Abstract deadline: 10th April ๐Ÿ“… Conference: Sep 14โ€“17 2026
2
3
569
Elena Burceanu retweeted
Preparing for CoLLAs 2026, weโ€™re revisiting last yearโ€™s standout keynotes: Leveraging Traces for Continual Reinforcement Learning - Martha White youtube.com/watch?v=Y6awtEKWโ€ฆ Join us in Bucharest: ๐Ÿ”— Abstract deadline: 10th April ๐Ÿ“… Conference: Sep 14โ€“17, 2026
2
5
734
Elena Burceanu retweeted
Preparing for CoLLAs 2026, weโ€™re revisiting last yearโ€™s standout keynotes: Computational Building Blocks for Complex Cognition - Cristina Savin youtube.com/watch?v=7gZCqrqmโ€ฆ Join us in Bucharest: ๐Ÿ”— Abstract deadline: 10th April ๐Ÿ“… Conference: Sep 14โ€“17, 2026 in Bucharest, Romania
3
4
802