iiinova — where your digital self keeps evolving

Reposted by Nicolas Beltran-Velez

Irving Institute for Cancer Dynamics @cancerdynamics.bsky.social · 21/05/2025

🎓 Hats off to the 2025 IICD graduates: Yining Ma Junze Huang Yichi Yang Ruilin Dai Boan Zhu Cameron Park @jlfan.bsky.social & Achille Nazaret! Wishing you all the best in your next chapter — we’re proud of you! 💙 #Columbia2025 @bleilab.bsky.social @khanhndinh.bsky.social @elhamazizi.bsky.social

071 0

Reposted by Nicolas Beltran-Velez

Freda Shi @fredashi.bsky.social · 28/03/2025

I received a review like this five years ago. It’s probably the right time now to share it with everyone who wrote or got random discouraging reviews from ICML/ACL.

1635 0

Reposted by Nicolas Beltran-Velez

Nathan Lambert @natolambert.bsky.social · 26/02/2025

First 11 chapters of RLHF Book have v0 draft done. Should be useful now. Next: * Crafting more blog content into future topics, * DPO+ chapter, * Meeting with publishers to get wheels turning on physical copies, * Cleaning & cohesiveness rlhfbook.com

0489 0

Reposted by Nicolas Beltran-Velez

briantrippe.bsky.social @briantrippe.bsky.social · 19/02/2025

🔥 Benchmark Alert! MotifBench sets a new standard for evaluating protein design methods in motif scaffolding. Why does this matter? Reproducibility & fair comparison have been lacking—until now. Paper: arxiv.org/abs/2502.12479 | Repo: github.com/blt2114/Moti... A thread ⬇️

14117 0

Reposted by Nicolas Beltran-Velez

Alexander Doria @dorialexander.bsky.social · 19/02/2025

The HuggingFace/Nanotron team just shipped an entire pretraining textbook in interactive format. huggingface.co/spaces/nanot... It’s not just a great pedagogic support, but many unprecedented data and experiments presented for the first time in a systematic way.

0409 0

Nicolas Beltran-Velez @velezbeltran.bsky.social · 19/02/2025

I just wanted to see what it looked like 😭

030 0

Nicolas Beltran-Velez @velezbeltran.bsky.social · 17/02/2025

Good God, please. I just want some gradients that don't vanish 😭

140 0

Reposted by Nicolas Beltran-Velez

Juan Diego Rodriguez @juand-r.bsky.social · 05/02/2025

I was hoping that recent events would lead to a mass exodus from X. Many have left, but most of the ML and LLM people have not. I have lost a lot of respect for the ML community.

9714 0

Reposted by Nicolas Beltran-Velez

lebellig @lebellig.bsky.social · 05/02/2025

Now that bluesky has gifs (it didn't work?), I can share (again) my educational notebook on discrete flow matching (by Itai Gat et al.). Also please check the original article and official implementation by Meta! 🐍 github.com/gle-bellier/... 🐍 github.com/facebookrese... 📄 arxiv.org/abs/2407.15595

1152 0

Reposted by Nicolas Beltran-Velez

Christian A. Naesseth @canaesseth.bsky.social · 05/02/2025

Really excited about this! We note a connection between diffusion/flow models and neural/latent SDEs. We show how to use this for simulation-free learning of fully flexible SDEs. We refer to this as SDE Matching and show speed improvements of several orders of magnitude. arxiv.org/abs/2502.02472

arxiv.org

SDE Matching: Scalable and Simulation-Free Training of Latent Stochastic Differential Equations

The Latent Stochastic Differential Equation (SDE) is a powerful tool for time series and sequence modeling. However, training Latent SDEs typically relies on adjoint sensitivity methods, which depend ...

05010 0

Reposted by Nicolas Beltran-Velez

Ted Underwood @tedunderwood.com · 03/02/2025

I have a sinking feeling that by 2029 I'm going to be faking a British accent so no one will think I was one of the *Americans* working on AI during the regime.

This is a scatterplot with the following key features:

Axes:
The x-axis represents "Interest in AI," with values ranging approximately from -2 to 2.
The y-axis represents "Willingness to Tolerate Closed, Autocratic Systems," also ranging from about -2 to 2.
Data Points:
Black dots dominate the plot, distributed across all four quadrants, indicating diverse positions on both variables.
A few red dots labeled "my peeps" are clustered in the bottom-right quadrant, signifying high interest in AI but low tolerance for closed, autocratic systems.
Blue Lines:
The plot includes horizontal and vertical blue lines at zero, dividing it into four quadrants for visual reference.
This visualization highlights a subset of individuals ("my peeps") who stand out from the majority based on their distinct combination of interest and values.

910910 0

Nicolas Beltran-Velez @velezbeltran.bsky.social · 03/02/2025

NGL, it's kind of surprising that more people haven't migrated here, especially given what Musk has been doing these days. I don't get it.

010 0

Reposted by Nicolas Beltran-Velez

Nathan Lambert @natolambert.bsky.social · 01/02/2025

Since everyone wants to learn RL for language models now post DeepSeek, reminder that I've been working on this book quietly in the background for months. Policy gradient chapter is coming together. Plugging away at the book every day now. rlhfbook.com/c/11-policy-...

215519 0

Reposted by Nicolas Beltran-Velez

Eugene Vinitsky 🍒 @eugenevinitsky.bsky.social · 29/01/2025

Please stop anthropomorphizing language models, it makes them feel really bad

3702 0

Reposted by Nicolas Beltran-Velez

Ben Collins @bencollins.bsky.social · 29/01/2025

This comments section is the first time I've felt even a shred of hope in eight days.

reddit.com

From the fednews community on Reddit

Explore this post and more from the fednews community

566202853824 0

Reposted by Nicolas Beltran-Velez

Jeff Dean @jeffdean.bsky.social · 26/01/2025

Nazi salutes and speaking at neo-Nazi rallies seems bad. There's history that we should learn from.

31155 0

Nicolas Beltran-Velez @velezbeltran.bsky.social · 25/01/2025

Something I really like about NLP research is that it makes everything super intuitive. This week I have been thinking about variational inference in NLP and a lot of the things that seemed to require mathematical intuition just become trivial when thinking about language. So cool:)

020 0

Nicolas Beltran-Velez @velezbeltran.bsky.social · 25/01/2025

But the memory needed for the value function kills the ones that don't have good GPUs 😭

000 0

Reposted by Nicolas Beltran-Velez

Ethan Mollick @emollick.bsky.social · 15/01/2025

New randomized, controlled trial by the World Bank of students using GPT-4 as a tutor in Nigeria. Six weeks of after-school AI tutoring = 2 years of typical learning gains, outperforming 80% of other educational interventions. And it helped all students, especially girls who were initially behind.

1535388 0

Nicolas Beltran-Velez @velezbeltran.bsky.social · 08/01/2025

I mostly use copilot for writing code (as auto complete), gpt4-o for boiler plate, and o1 for serious debugging or boilerplate with some complexity or a lot of requirements. I also use o1 for quick but slightly involved experiments but not as often.

050 0

Nicolas Beltran-Velez @velezbeltran.bsky.social · 07/01/2025

I use chatgpt over Google for a lot of things because it is really good at fuzzy queries + data aggregation from many sources. I feel that as long as you double check results it is much faster and convenient.

040 0

Nicolas Beltran-Velez @velezbeltran.bsky.social · 28/12/2024

Does anyone have any good resources to learn about quantization? Any essential papers to read and resources about how to use/quantize models in practice are greatly appreciated!

030 0

Nicolas Beltran-Velez @velezbeltran.bsky.social · 22/12/2024

That seems very reasonable no? I assumed the public set was used by most (almost all?) algorithms that have been benchmarked against the task. Isn't that the case? (legitimate question)

040 0

Reposted by Nicolas Beltran-Velez

Mark Riedl @markriedl.bsky.social · 20/12/2024

1-> 2 -> 3 -> 3.5 -> 4 -> 4o -> o1 -> o3 I guess we need AGI just to figure out how to name things

7716 0

Reposted by Nicolas Beltran-Velez

Csaba Szepesvari @skiandsolve.bsky.social · 19/12/2024

If you are into ML theory (RL or not) with a proven track record, and you are interested in an industry research position, PM me. Feel free to spread the word.

27531 0

Reposted by Nicolas Beltran-Velez

Joy Fan @jlfan.bsky.social · 18/12/2024

🧵 Excited to share #Echidna, a Bayesian framework for quantifying the impact of gene dosage on phenotypic plasticity: tinyurl.com/296kf7hf! With @elhamazizi.bsky.social and @mingxz.bsky.social, we integrate scRNA-seq & WGS to uncover how CNAs drive tumor evolution and transcriptional variability.

biorxiv.org

2156 0

Reposted by Nicolas Beltran-Velez

Elham Azizi @elhamazizi.bsky.social · 18/12/2024

Proud of this work spearheaded by the phenomenal @jlfan.bsky.social and @mingxz.bsky.social in collaboration w/ Ben Izar! The past 3 years we've worked hard to unravel how #CNVs shape #tumor phenotypic plasticity seen in #singlecell #RNAseq data ➡️ #Echidna 🦔

1186 0

Nicolas Beltran-Velez @velezbeltran.bsky.social · 12/12/2024

Hello! We will be presenting Estimating the Hallucination Rate of Generative AI at NeurIPS. Come if you'd like to chat about epistemic uncertainty for In-Context Learning, or uncertainty more generally. :) Location: East Exhibit Hall A-C #2703 Time: Friday @ 4:30 Paper: arxiv.org/abs/2406.07457

0234 0

Reposted by Nicolas Beltran-Velez

Yuli Slavutsky @yulislavutsky.bsky.social · 10/12/2024

I'm on my way to #NeurIPS2024. On Friday I'm going to present my latest paper with Yuval Benjamini. The gist is in the comments, and come chat with me to hear more!

174 0

Reposted by Nicolas Beltran-Velez

claudia shi @claudiashi.bsky.social · 10/12/2024

The circuit hypothesis proposes that LLM capabilities emerge from small subnetworks within the model. But how can we actually test this? 🤔 joint work with @velezbeltran.bsky.social @maggiemakar.bsky.social @anndvision.bsky.social @bleilab.bsky.social Adria @far.ai Achille and Caro

2156 0

Reposted by Nicolas Beltran-Velez

David Cox @neurobongo.bsky.social · 08/12/2024

I have been a little bit scarce on social media in the last few months. Some of that has just been from being busy at work, but some of it has had to do with my father's passing. He hated the very idea of social media, but he religiously followed my Twitter, and then Bluesky, feeds. /n

5451 0

Reposted by Nicolas Beltran-Velez

Elham Azizi @elhamazizi.bsky.social · 16/11/2023

We are thrilled to share #Decipher 🔍! Extremely proud of Achille and Joy for leading the development of this creative #ML tool combining VAEs with deep exponential models for comparing disease & healthy trajectories ➕ applying it to study leukemic onset! biorxiv.org/conte…

163 0

Reposted by Nicolas Beltran-Velez

Andrew Gordon Wilson @andrewgwils.bsky.social · 05/12/2024

I wanted to make my first post about a project close to my heart. Linear algebra is an underappreciated foundation for machine learning. Our new framework CoLA (Compositional Linear Algebra) exploits algebraic structure arising from modelling assumptions for significant computational savings! 1/4

313821 0

Reposted by Nicolas Beltran-Velez

Keyon Vafa @keyonv.bsky.social · 04/12/2024

Thank you Nature and @anilananth.bsky.social for this great feature on LLMs and AGI (and for highlighting our work arxiv.org/abs/2406.03689)

0104 0

Nicolas Beltran-Velez @velezbeltran.bsky.social · 04/12/2024

From our lab finally I convinced Yuli Slavutsky (generalization/uncertainty/robustness) @yulislavutsky.bsky.social and Claudia Shi (science of LLMs and AI safety) @claudiashi.bsky.social to use bluesky. Go give them a follow! 😁

080 0

Nicolas Beltran-Velez @velezbeltran.bsky.social · 04/12/2024

But yes this makes sense. Thank you for answering my question and taking the time to parse what I wrote! I took a second read and realized my original post was a bit difficult to understand. Thank you as well for the reference. I 'll check that out as well!

010 0

Nicolas Beltran-Velez @velezbeltran.bsky.social · 04/12/2024

Rather, the main situation where this happens is when having high epistemic uncertainty means placing a lot of mass in models π with high predictive entropy, which is often the case when we model for example the variance. In those cases I would expect to see this even if the model has converged.

110 0

Nicolas Beltran-Velez @velezbeltran.bsky.social · 04/12/2024

Yeah that's roughly my argument! However, I don't think it happens only when we have small n.

110 0

Nicolas Beltran-Velez @velezbeltran.bsky.social · 03/12/2024

Yeah, I just looked at the algos in the paper again and I can see why it would be confusing.😅 Glad it helped!

100 0

Nicolas Beltran-Velez @velezbeltran.bsky.social · 03/12/2024

Algo 2 (sampling) tinyurl.com/treeffuser-s... tinyurl.com/treeffuser-s...

Pseudocode sampling algorithm for Treeffuser.

130 0

Nicolas Beltran-Velez @velezbeltran.bsky.social · 03/12/2024

Algo 1 (training) tinyurl.com/treeffuser-s...

Pseudocode training algorithm for Treeffuser.

130 0

Nicolas Beltran-Velez @velezbeltran.bsky.social · 03/12/2024

Algo 1 and Algo 2 are very close to the standard training and use of diffusions under the denoising score matching framework. I have some pseudocode version below. Let me know if they help. Happy to explain more! (I also included the relevant files in the actual repo if you wanna take a look)

120 0