Sign in

Ivan Rubachev

@puhsu.bsky.social
326 followers 1.2K following 21 posts

ML Researcher at research.yandex.com | Working on DL for Tabular Data

PostsRepliesMedia
Reposted by Ivan Rubachev
Pablo Samuel Castro @pcastr.bsky.social · 17/03/2026
New paper 🚨 "Stable Deep Reinforcement Learning via Isotropic Gaussian Representations" Deep RL suffers from unstable training, representation collapse, and neuron dormancy. We show that a simple geometric insight, isotropic Gaussian representations, can fix this. Here's how 👇
2295
Ivan Rubachev @puhsu.bsky.social · 13/02/2026
TabICL also released a new version this week, higly recommend checking it out too github.com/soda-inria/t...
github.com
GitHub - soda-inria/tabicl: TabICLv2: A state-of-the-art tabular foundation model
TabICLv2: A state-of-the-art tabular foundation model - soda-inria/tabicl
030
Ivan Rubachev @puhsu.bsky.social · 13/02/2026
P.S. Its interesting how in our little corner of the ML/DL space SOTA foudnation models are actually open (GraphPFN was initialised from the github.com/limix-ldm/Li... model, and used @dholzmueller.bsky.social @gaelvaroquaux.bsky.social prior sampling from TabICL
github.com
GitHub - limix-ldm-ai/LimiX: LimiX: Unleashing Structured-Data Modeling Capability for Generalist Intelligence https://arxiv.org/abs/2509.03505
LimiX: Unleashing Structured-Data Modeling Capability for Generalist Intelligence https://arxiv.org/abs/2509.03505 - limix-ldm-ai/LimiX
140
Ivan Rubachev @puhsu.bsky.social · 13/02/2026
They also wrote up a blogpost about the model research.yandex.com/blog/graphpf...
research.yandex.com
GraphPFN: a graph foundation model pretrained on diverse synthetic graphs
We introduce GraphPFN, a graph foundation model that extends the prior-data fitted network (PFN) framework to graphs, achieving state-of-the-art in-context and finetuned results on real-world graphs.
110
Ivan Rubachev @puhsu.bsky.social · 13/02/2026
To piggy-back a bit on foundation models for structured data discussion here My colleagues at Yandex Research just updated the GraphPFN paper. It's a Graph Foundation Model that works on graph datasets with tabular features, and shows SOTA results both in ICL regimes and when fine-tuned.
arxiv.org
GraphPFN: A Prior-Data Fitted Graph Foundation Model
Graph foundation models face several fundamental challenges including transferability across datasets and data scarcity, which calls into question the very feasibility of graph foundation models. Howe...
122
Ivan Rubachev @puhsu.bsky.social · 03/02/2026
this?
020
Reposted by Ivan Rubachev
embedding-shapes @emsh.cat · 27/01/2026
How hard can it be to build a browser from scratch for three platforms anyways? Apparently 20K lines of code and ~70 hours from first commit to last. emsh.cat/one-human-on... #llm #llms #ai #codex #openai
4389
Ivan Rubachev @puhsu.bsky.social · 27/01/2026
I liked the response better geohot.github.io//blog/jekyll...
geohot.github.io
The Importance of Diversity
I read Dario’s The Adolescence of Technology and it’s scary. It assumes the perspective of a top-down ruler, that someone can and will get to control AI. This is taken as a given. Machines of Loving G...
140
Reposted by Ivan Rubachev
dame @dame.is · 21/01/2026
if you’re going to use AI in your workflow, you have to get extremely good at self-discipline/focus because AI will literally tempt you to pursue every tiny whim/idea that enters your brain and thus will absolutely destroy you and your work if left unchecked slow down before it’s too late
59712
Reposted by Ivan Rubachev
vectorware.com @vectorware.com · 20/01/2026
We are excited to announce that we can successfully use Rust's standard library from the GPU. This has never been done before. www.vectorware.com/blog/rust-st... Supporting Rust's standard library enables existing Rust code to work on the GPU and makes GPU programming feel normal.
vectorware.com
Rust's standard library on the GPU
GPU code can now use Rust's standard library. We share the implementation approach and what this unlocks for GPU programming.
424957
Ivan Rubachev @puhsu.bsky.social · 19/01/2026
www.d12frosted.io/posts/2025-1...
d12frosted.io
Knowing Less, Producing More
On AI, Craftsmanship, and the Slow Erosion of Productive Friction.
010
Reposted by Ivan Rubachev
dan @danabra.mov · 18/01/2026
formats over apps
overreacted.io
A Social Filesystem — overreacted
Formats over apps.
72977205
Ivan Rubachev @puhsu.bsky.social · 04/12/2024
Explicitly adding induction heads helps. Some gains in NLP, seemingly bigger in RL algorithm distillation arxiv.org/abs/2411.01958
arxiv.org
N-Gram Induction Heads for In-Context RL: Improving Stability and Reducing Data Needs
In-context learning allows models like transformers to adapt to new tasks from a few examples without updating their weights, a desirable trait for reinforcement learning (RL). However, existing in-co...
110
Ivan Rubachev @puhsu.bsky.social · 03/12/2024
⚡️
000
Ivan Rubachev @puhsu.bsky.social · 01/12/2024
I just completed "Historian Hysteria" - Day 1 - Advent of Code 2024 #AdventOfCode adventofcode.com/2024/day/1 (in zig btw)
adventofcode.com
Day 1 - Advent of Code 2024
111
Ivan Rubachev @puhsu.bsky.social · 29/11/2024
Yep, just need to find the code. I can share
110
Ivan Rubachev @puhsu.bsky.social · 29/11/2024
Yeah. I've experimented a bit with the existing code. It generalized to some of our specific problems in tabular DL (even though the meta-train was mostly from language and vision tasks). Curious what do you mean by actually worked here? No edge cases and failures, or just easy to use technically?
100
Ivan Rubachev @puhsu.bsky.social · 29/11/2024
#MLsky
0696
Reposted by Ivan Rubachev
Omead Pooladzandi ✈️ NeurIPS'24 @hessianfree.bsky.social · 28/11/2024
The rejects were horribly misinformed self contradictory but extremely confident. PSGD, SOAP and friends are taking over regardless of academia.
002
Ivan Rubachev @puhsu.bsky.social · 29/11/2024
VeLO was something else, I’m a fan arxiv.org/abs/2211.09760
arxiv.org
VeLO: Training Versatile Learned Optimizers by Scaling Up
While deep learning models have replaced hand-designed features across many domains, these models are still trained with hand-designed optimizers. In this work, we leverage the same scaling approach b...
350
Reposted by Ivan Rubachev
alpindale.bsky.social @alpindale.bsky.social · 28/11/2024
Thank you @bsky.app team for correcting the mistake. Glad to be back!
3930224
Reposted by Ivan Rubachev
Jeremy Howard @howard.fm · 28/11/2024
Did you know that 99% of email today is spam? Your inbox isn’t 99% spam because AI is used to filter it. The same 99% will happen here too, but if AI researchers continue to get perma-banned for making available the datasets needed to filter it, it’s going to make this platform unusable.
4251064
Ivan Rubachev @puhsu.bsky.social · 26/11/2024
@trl-research.bsky.social
021
Ivan Rubachev @puhsu.bsky.social · 26/11/2024
Tabular DL and AutoML podcast just dropped. For sure watching this youtu.be/3qpQ-sMRafE
youtu.be
How AutoML Creates New Opportunities for Europe - Frank Hutter // CyberValley Podcast #5
YouTube video by Cyber Valley
1112
Ivan Rubachev @puhsu.bsky.social · 25/11/2024
Hello to all #ICLR reviewers on #MLsky
0274
Ivan Rubachev @puhsu.bsky.social · 25/11/2024
bsky.app/profile/hame...
041
Ivan Rubachev @puhsu.bsky.social · 24/11/2024
But keep the numbers in appendix or code pls So annoying when the only info is in visual form with unclear axes etc. I agree that it’s much better for presentation, but when digging in, I often need raw metrics.
140
Ivan Rubachev @puhsu.bsky.social · 18/11/2024
…extend of customisability? If I understand correctly, we can do a lot with custom feeds. Some examples here github.com/Bossett/bsky...
github.com
GitHub - Bossett/bsky-feeds
Contribute to Bossett/bsky-feeds development by creating an account on GitHub.
000
Ivan Rubachev @puhsu.bsky.social · 18/11/2024
Wow. Didn’t know we can create custom algorithmic feeds here. This is cool! What are your favourites, what’s the extend of (context: docs.bsky.app/docs/starter...)
docs.bsky.app
Custom Feeds | Bluesky
Custom feeds, or feed generators, are services that provide custom algorithms to users through the AT Protocol. This allows users to choose their own timelines, whether it's an algorithmic For You pag...
110
Reposted by Ivan Rubachev
David Holzmüller @dholzmueller.bsky.social · 18/11/2024
Can deep learning finally compete with boosted trees on tabular data? 🌲 In our NeurIPS 2024 paper, we introduce RealMLP, a NN with improvements in all areas and meta-learned default parameters. Some insights about RealMLP and other models on large benchmarks (>200 datasets): 🧵
Paper screenshot and Figure 1 (c) with cumulative ablations for components of RealMLP-TD.
1608
Reposted by Ivan Rubachev
Madelon Hulsebos @madelonhulsebos.bsky.social · 18/11/2024
WIP starterpack w researchers on Table Representation Learning (TRL): all things related to representation learning and generative models for e.g. tables, DBs, spreadsheets! I'll curate but DM/reply w handle+some info welcome! Also follow @trl-research.bsky.social for updates 🤗 go.bsky.app/4SNSMRj
go.bsky.app
Table Representation Learning researchers
Join the conversation
8248