Saltar al contenido

Detalles

Scientific Paper Reading Club: Beyond Transformers & Multimodal Frontiers
This is part of the scientific paper reading club.
This Saturday, we are expanding our horizons beyond standard LLMs! Following our recent discussions, we will be diving into points 3 and 5 from our reading list presentation: Alternative architectures (SSM/RNN) and Multimodal foundation models.
To make it even more exciting, we will be connecting these topics to some of the freshest breakthroughs presented at the recent ICLR conference ([https://iclr.cc/](https://www.google.com/url?sa=E&q=https%3A%2F%2Ficlr.cc%2F)).
What to expect:
Our presenter, Maxim Musin, will guide us through these two major frontiers in modern AI:

  1. Alternative Architectures (SSM/RNN): Transformers are incredibly powerful, but their computational costs remain a massive bottleneck for long contexts. We will discuss why State Space Models (SSMs) like Mamba are emerging as the next big leap. We will look at fresh ICLR[1] insights, such as "Mamba-3: Improved Sequence Modeling using State Space Principles" [https://arxiv.org/abs/2603.15569](https://www.google.com/url?sa=E&q=https%3A%2F%2Farxiv.org%2Fabs%2F2603.15569) and "MambaSL: Exploring Single-Layer Mamba for Time Series Classification" [https://arxiv.org/abs/2604.15174](https://www.google.com/url?sa=E&q=https%3A%2F%2Farxiv.org%2Fabs%2F2604.15174)
  2. Multimodal Foundation Models: We will explore how AI is learning to "think" natively across different continuous spaces and complex interactions. We will touch on standout ICLR papers like "Massively Multimodal Foundation Models: A Framework for Capturing Interactions with Specialized Mixture-of-Experts" [https://arxiv.org/abs/2509.25678](https://www.google.com/url?sa=E&q=https%3A%2F%2Farxiv.org%2Fabs%2F2509.25678) and "Theory of Space: Can Foundation Models Construct Spatial Beliefs through Active Exploration?" [https://arxiv.org/abs/2602.07055](https://www.google.com/url?sa=E&q=https%3A%2F%2Farxiv.org%2Fabs%2F2602.07055)

You don't need a heavy math background to join—Maxim will break down the intuition behind these complex architectures so that the core concepts are easy to follow for everyone.
Meeting Details:
🗣 Presenter: Maxim Musin
💬 Language of the meeting: English
🎟 Entrance donation: 10 eur
📍 Location: Maui coworking
(Note: For some reason, Meetup maps sometimes wrongly point to other addresses. Here is the exact address: [https://maps.app.goo.gl/fjx9t4zG8GnEmXLr9?g_st=ac](https://www.google.com/url?sa=E&q=https%3A%2F%2Fmaps.app.goo.gl%2Ffjx9t4zG8GnEmXLr9%3Fg_st%3Dac) )
Looking forward to a great discussion this Saturday!

Aparently there is no way of fixing meetup's markdown mess, sorry for this guys, just use your abstract imagination and experience with latech to read anouncement, this markdown render problem seem so be some kind of bug, i cleaned it in the redactor but with no effect, probably copying and pasting it clean into editor will solve it, but generating clean landing page with similar capabilities in claude code would also solve it in a way, so i didn't bother, see you at saturday

Temas relacionados

Artificial Intelligence
Machine Learning
Neural Networks
Mathematics
Computer Science

También te puede gustar