News

DSML Reading Club Meeting #3

This week, Mikhail Shkorin will present graph embeddings through the paper “Optimal Time Complexity Algorithms for Computing General Random Walk Graph Kernels on Sparse Graphs”.

Current graph-embedding methods either lack theoretical grounding, as with GNNs, or have high computational complexity, as with kernel approaches.

The paper proposes a simple and scalable algorithm that compares graphs and their nodes efficiently through a linear approximation of random-walk kernels.

At the meeting, we will discuss:

• Why graph embeddings are needed
• How to create node embeddings in linear time
• How to move from node embeddings to graph embeddings on the fly without computing a huge adjacency matrix
• Graph transformers on point clouds in robotics
• Add to calendar: calendar.app.google/wci6yDfF8M68tHCv7
• Thursday, May 15, 11:00 Kazakhstan time
• Google Meet
• Host: Yelaman Abdullin

An offline LLM arena for the Kazakh language

Community residents Sanzhar Murzakhmetov, Sanzhar Umbet, Beksultan Sagyndyk, and Kirill Yakunin launched the first offline LLM arena for Kazakh. Its main goal is to test not merely next-token generation, but understanding of cultural context as a whole.

What the team built:

• A custom culture-focused QA dataset
• Topics and keywords assembled with linguists using Serper, Perplexity, and LLM generation
• Pairwise model comparisons, first judged by GPT-4o and finalized with a Bradley–Terry model

Current results:

• Google DeepMind's Gemma models are consistently strong
• Sherkala-8B from MBZUAI ranks second, outperforming even larger models
• ISSAI at Nazarbayev University has good multiple-choice scores but some of the weakest generation results

Kazakhstan's first national AI olympiad for school students has concluded!

It was a demanding AI marathon with seven Kaggle competitions and an entirely new competition format for participants and organizers alike. Forty finalists reached the end after two remote rounds and two days of in-person competition in Astana.

We hope the olympiad draws talented young people in Kazakhstan toward this promising field and contributes to the development of artificial intelligence in our region.

Congratulations to all the prize winners, and we are proud of every finalist who put in tremendous effort to become a pioneer of this new olympiad. We look forward to welcoming all of you to our small DSML.kz community!

DSML Reading Club Meeting #2

Next week, Anuar Taskynov will present VGGT—Visual Geometry Grounded Transformer.

VGGT is a next-generation foundation model for 3D computer-vision tasks. From one, several, or even hundreds of scene images, it can immediately predict key 3D properties: camera parameters, depth maps, dense point clouds, and 3D tracking.

Unlike traditional approaches, VGGT works as a single universal model without complex post-processing. It remains fast—under one second per reconstruction—and accurate, achieving state-of-the-art results across several 3D tasks.

At the meeting, we will discuss: a brief introduction to 3D computer-vision tasks; an overview of VGGT's architecture and capabilities; why the model matters and how it can be applied in practice.

📄 Paper: VGGT: Visual Geometry Grounded Transformer (https://vgg-t.github.io/)

🗓 Calendar: https://calendar.app.google/3MLTmm99NSdBjZ3q9

📞 Google Meet: https://meet.google.com/wqi-crwg-ztp

🕔 Date & Time: Thursday, May 8, 11:00 Kazakhstan time

🗣 Speaker: Anuar Taskynov (https://linkedin.com/in/anuartask)

React with a fire emoji and add the meeting to your calendar if you are joining us!