Sign in

Joseph Viviano

@josephdviviano.bsky.social
680 followers 1K following 70 posts

humanistic technology bretheren @MILA_Quebec ~ AI for Science ~ ex @deepgenomics & @CAMHResearch, intern @google & @imagia_ai ~ www.viviano.ca

PostsRepliesMedia
Joseph Viviano @josephdviviano.bsky.social · 12/05/2026
Agents are transforming technical work. I've tried to establish a good agent workflow in research. This captures my and some of my colleagues @mila-quebec.bsky.social thoughts on a method which balances the automation with understanding. Feedback welcome! viviano.ca/blog/2026/ag...
viviano.ca
Agentic Research Best Practices | Mode Collapse
Some notes from myself and the Mila community on how to use agents to do research effectively.
000
Joseph Viviano @josephdviviano.bsky.social · 16/03/2026
me: "can you use whatever resources you like, and python, to generate a short 'youtube poop' video and render it using ffmpeg ? can you put more of a personal spin on it? it should express what it's like to be a LLM" claude opus 4.6:
100
Joseph Viviano @josephdviviano.bsky.social · 14/02/2026
We need to raise the bar on research code right now. 1) documentation and tests are dead simple now. 2) creating benchmarks integrating across multiple implementations 3) have agents double check your work / fix broken tests 4) fix outstanding bugs in major scientific packages
35714
Joseph Viviano @josephdviviano.bsky.social · 14/02/2026
@hempuli.bsky.social ! We want to expand on utilForever's Baba is You RL simulator! github.com/utilForever/... Is that ok with you? We want to expand the set of levels included to involve more complex problems and reasoning over the various abstractions in your (awesome and very hard) game!
github.com
GitHub - utilForever/baba-is-auto: Baba Is You simulator using C++ with some reinforcement learning
Baba Is You simulator using C++ with some reinforcement learning - utilForever/baba-is-auto
000
Joseph Viviano @josephdviviano.bsky.social · 30/07/2025
We’ve released torchgfn v2! github.com/GFNOrg/torch... We believe we’ve built the go-to library for fundamental GFlowNet methods development and prototyping, and we’re really excited to help you start using it.
github.com
GitHub - GFNOrg/torchgfn: A modular, easy to extend GFlowNet library
A modular, easy to extend GFlowNet library. Contribute to GFNOrg/torchgfn development by creating an account on GitHub.
140
Reposted by Joseph Viviano
Yoshua Bengio @yoshuabengio.bsky.social · 03/06/2025
Today marks a big milestone for me. I'm launching @law-zero.bsky.social, a nonprofit focusing on a new safe-by-design approach to AI that could both accelerate scientific discovery and provide a safeguard against the dangers of agentic AI.
38024
Reposted by Joseph Viviano
Pete Buttigieg @petebuttigieg.bsky.social · 23/05/2025
America cannot long remain free, nor first among nations, if it becomes the kind of place where universities are dismantled because they don't align politically with the current head of the government.
14424769011267
Reposted by Joseph Viviano
Hafez Ghaemi @hafezghm.bsky.social · 14/05/2025
Preprint Alert 🚀 Can we simultaneously learn transformation-invariant and transformation-equivariant representations with self-supervised learning? TL;DR Yes! This is possible via simple predictive learning & architectural inductive biases – without extra loss terms and predictors! 🧵 (1/10)
15116
Joseph Viviano @josephdviviano.bsky.social · 25/04/2025
Great job gang! And thanks to @boussifo.bsky.social for being such a stellar lead. (Also depicted: @jainmoksh.bsky.social)
060
Joseph Viviano @josephdviviano.bsky.social · 04/04/2025
Ecstatic to show off some work my brilliant colleagues and I did at @iclr-conf.bsky.social this year! 🚀 We address the credit assignment challenge under long trajectories in RL or GFlowNets by constructing high order actions, or “chunks”, effectively compressing trajectory lengths!
1100
Reposted by Joseph Viviano
Johan Edstedt @parskatt.bsky.social · 05/03/2025
From having to wrap a bunch of methods I can say I'm not a big fan of having configs in the form of argparsers.
271
Reposted by Joseph Viviano
Carl T. Bergstrom @carlbergstrom.com · 09/02/2025
any economists here that can help me understand why the world's richest man might want to kill the consumer financial protection bureau
Musk tweet: CFPB RIP
741694254
Reposted by Joseph Viviano
Daniel Musekamp @danielmusekamp.bsky.social · 11/12/2024
Neural surrogates can accelerate PDE solving but need expensive ground-truth training data. Can we reduce the training data size with active learning (AL)? In our NeurIPS D3S3 poster, we introduce AL4PDE, an extensible AL benchmark for autoregressive neural PDE solvers. 🧵
1123
Joseph Viviano @josephdviviano.bsky.social · 09/12/2024
"The complaint alleges the algorithm, dubbed nH Predict, has a 90% error rate, basing that calculation on the percentage of payment denials reversed through internal appeals processes or administrative law judge rulings." www.statnews.com/2023/11/14/u...
statnews.com
UnitedHealth faces class action lawsuit over algorithmic care denials in Medicare Advantage plans
A class action lawsuit was filed against UnitedHealth and a subsidiary alleging they use an algorithm to deny care to seriously ill patients
070
Reposted by Joseph Viviano
Bird account @dialecticbio.bsky.social · 06/12/2024
Scalable emulation of protein equilibrium ensembles with generative deep learning www.biorxiv.org/content/10.1... No code is available yet, but sounds huge. Effectively does MD-sim, but fast!
Fig. 1 Overview of model and architecture. a) Actin conformational changes and filament formation / dissociation as an example
for the mechanistic basis of protein function. b) ML model architecture consisting of protein sequence encoder and denoising
diffusion model. The diffusion model samples coarse-grained protein structures from an approximate equilibrium distribution, from
which properties such as free energy differences can be computed. c) Architecture of the score model used in the denoising diffusion
model. d) Data integration and model training pipeline. e) Data processing pipeline for pretraining. f) Experimental property
training for finetuning
0124
Joseph Viviano @josephdviviano.bsky.social · 05/12/2024
This book radically influenced my way of looking at human civilization and the kinds of tech that must be built to move us into a post sacristy, sustainable future. His story (overcoming failure/depression to produce major intellectual contributions outside the academic system was also formative.
040
Joseph Viviano @josephdviviano.bsky.social · 04/12/2024
Who are the top people working on computationally guided biosynthesis in the Boston / NY area?
010
Reposted by Joseph Viviano
Rupali Bhati @rupalibhati.bsky.social · 27/11/2024
Mila is such a large community. One starter pack just isn’t enough! After @josephdviviano.bsky.social’s Mila list filled up, I decided to make another one. Will continue to add members until this one is full too. go.bsky.app/9nXTDHo
4339
Joseph Viviano @josephdviviano.bsky.social · 27/11/2024
Rupali is making the second Mila starter pack. Please reply to this if you're at Mila and she will add you.
2180
Reposted by Joseph Viviano
Oussama Zekri @ozekri.bsky.social · 26/11/2024
🚀 Did you know you can use the in-context learning abilities of an LLM to estimate the transition probabilities of a Markov chains? The results are pretty exciting ! 😄
182
Reposted by Joseph Viviano
Nenad Tomasev @nenadtomasev.bsky.social · 26/11/2024
A great new essay on AI for Science from our colleagues here: deepmind.google/public-polic...
deepmind.google
A new golden age of discovery
In this essay, we take a tour of how AI is transforming scientific disciplines from genomics to computer science to weather forecasting. Some scientists are training their own AI models, while...
0225
Reposted by Joseph Viviano
Kevin K. Yang 楊凱筌 @kevinkaichuang.bsky.social · 25/11/2024
BioM3: a model that generates proteins conditioned on text prompts, with wet-lab validation! www.biorxiv.org/content/10.1...
Overview of the BioM3 framework for protein design via natural language prompts.Wet-lab validation
18015
Joseph Viviano @josephdviviano.bsky.social · 25/11/2024
My suspicion is that we might be able to squeeze more juice out of the data with the correct training paradigm but what we're seeing in PLMs is similar to the saturation we're seeing in LLMs - every dataset/model pair will have a different S curve, with PLMs it is smaller.
020
Reposted by Joseph Viviano
Claus Wilke @clauswilke.com · 25/11/2024
The complete guide for transfer learning with the protein language model ESM-2. (In brief: Use ESM-2 650M and calculate mean embeddings across sites.) www.biorxiv.org/content/10.1...
biorxiv.org
Scaling Down for Efficiency: Medium-Sized Transformer Models for Protein Sequence Transfer Learning
Protein language models such as the transformer-based Evolutionary Scale Modeling 2 (ESM2) can offer deep insights into evolutionary and structural properties of proteins. While larger models, such as...
24921
Reposted by Joseph Viviano
Eugene Vinitsky 🍒 @eugenevinitsky.bsky.social · 23/11/2024
Kind of a broken record here but proceedings.neurips.cc/paper_files/... is totally fascinating in that it postulates two underlying, measurable structures that you can use to assess if RL will be easy or hard in an environment
e introduce the effective horizon, a property of
MDPs that controls how difficult RL is. Our analysis is mo-
tivated by Greedy Over Random Policy (GORP), a simple
Monte Carlo planning algorithm (left) that exhaustively ex-
plores action sequences of length k and then uses m random
rollouts to evaluate each leaf node. The effective horizon
combines both k and m into a single measure. We prove
sample complexity bounds based on the effective horizon that
correlate closely with the real performance of PPO, a deep
RL algorithm, on our BRIDGE dataset of 155 deterministic
MDPs (right).
815029
Reposted by Joseph Viviano
Willie Neiswanger @willieneis.bsky.social · 10/11/2024
I'm making a list of AI for Science researchers on bluesky — let me know if I missed you / if you'd like to join! go.bsky.app/AcP9Lix
16225389
Joseph Viviano @josephdviviano.bsky.social · 22/11/2024
Hey @mubashariqbal.com please add: go.bsky.app/BHKxoss
120
Reposted by Joseph Viviano
Flaviu Cipcigan @flaviucipcigan.bsky.social · 07/11/2024
And if you'd like to contribute or use these models, join our working group. We're still starting up, so expect some rough edges but we've got a few really exciting projects lined up!
thealliance.ai
Materials and Chemistry | AI Alliance
We aim to curate datasets, tasks and benchmarks for materials science, build out foundation models in chemistry for prediction of properties, experimental outcomes or generation of new candidates and ...
032
Joseph Viviano @josephdviviano.bsky.social · 20/11/2024
I'm making an unofficial starter pack with some of my colleagues at Mila. WIP for now but here's the link! go.bsky.app/BHKxoss
77129