Three papers by the Yandex Research team (shown in bold below) and our collaborators have been accepted for publication at the International Conference on Learning Representations (ICLR 2025), and one paper has been accepted for publication at the Annual Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics (NAACL 2025).

Papers accepted to ICLR and NAACL 2025

In 2024, Yandex Research has evolved into several dedicated research groups, each focused on a specific scientific area: Generative Computer Vision, Efficient Large Language Models (LLMs), Graph Neural Networks, and Tabular Deep Learning.

We presented 17 papers at major conferences, including NeurIPS, ICLR, ICML, CVPR, and ACL. Several papers were co-authored with researchers from prominent global institutions, including IST Austria, KAUST Saudi Arabia, Carnegie Mellon University, Meta AI, and Together AI.

Yandex Research in 2024

Our paper has been accepted for an oral presentation at the Learning on Graphs Conference (LoG 2024).

Paper accepted to LoG 2024

At Yandex Research, we advance progress in science and develop technologies for a better tomorrow.

large_join.jpg

medium_join.jpg

small_join.jpg

thumbnail_join.jpg

join.jpg

Join our research team and get an opportunity to engage in world-class applied and theoretical research, gain experience working on high-impact projects and publish at top-tier academic conferences. You will contribute to high-tech services like computer vision technologies, dialogue systems, neural machine translation and self-driving cars. 

Work with us

Tabular data involves two-dimensional tables with objects (rows) and features (columns), which are used in numerous applied tasks such as classification, regression, ranking and many others.

Tabular data

Today, training most powerful models often takes significant resources. Our research aims to make large-scale training more efficient and accessible to the entire machine learning community.

Large-scale machine learning

Generative models in computer vision are powerful tool for various applications.

Generative models

Graphs are a natural way to represent data from various domains such as social networks, molecules, text, code, etc. We develop and analyze algorithms for graph-structured data.

Graph machine learning

Algorithmic reasoning focuses on building models that can execute classic algorithms. It allows one to combine the advantages of neural networks with theoretical guarantees of algorithms.

Neural algorithmic reasoning

Yandex Research team regularly contributes to the computer vision research community, mostly in the field of image retrieval and generative modelling.

Computer vision

icon_post_8.svg

In this work, we achieve perfect neural execution of several algorithms by forcing the node and edge representations to be from a fixed finite set. Also, the proposed architectural choice allows us to prove the correctness of the learned algorithms for any test data.

Discrete neural algorithmic reasoning

icon_post_28.svg


We introduce SpecExec, a speculative decoding method that delivers up to 20 tokens per iteration and up to 15x speedups in offloading settings. This enables usable LLM inference speeds for users with limited GPU VRAM sizes, who have had to use offloading or settle for lower quality models until now. 

SpecExec: Massively Parallel Speculative Decoding for Interactive LLM Inference on Consumer Devices

icon_post_14.svg

Diffusion models are currently the most powerful approach to generating images from a text description. One popular application is the adaptation of such models to new concepts: usually, it is quite difficult or even impossible to write a prompt that accurately describes the target object.

At the same time, existing methods are too slow for practical purposes. In “Is This Loss Informative? Faster Text-to-Image Customization by Tracking Objective Dynamics”, our paper accepted at NeurIPS 2023, we propose DVAR  — a lightweight early stopping criterion for gradient-based adaptation methods. Our experiments with Stable Diffusion show up to 8 times acceleration with minor losses in the quality of identity preservation and less overfitting. We release the source code of DVAR so that anyone can speed up their adaptation pipeline.

Follow the cutting-edge research in computer science

Blog

Publications

SpecExec: Massively Parallel Speculative Decoding for Interactive LLM Inference on Consumer Devices

PV-Tuning: Beyond Straight-Through Estimation for Extreme LLM Compression

Sequoia: Scalable, Robust, and Hardware-aware Speculative Decoding

Challenges of Generating Structurally Diverse Graphs