Leonardo Cotta
@cottascience
Followers
1K
Following
7K
Media
32
Statuses
1K
floptimistic @EITOxford from BH 🔺 🇧🇷
London, UK
Joined July 2018
Simple, but well-engineered, pretrain+SFT takes you really far 🫡 Excited to read the report and learn from this amazing team!
0
0
2
Curing diseases is not a jobs program for humans to feel special about their little existence.
I was at an event on AI for science yesterday, a panel discussion here at NeurIPS. The panelists discussed how they plan to replace humans at all levels in the scientific process. So I stood up and protested that what they are doing is evil. Look around you, I said. The room is
0
2
10
We are beyond thrilled to share our first flagship models, Rnj-1 base and instruct 8B parameter models. Rnj-1 is the culmination of 10 months of hard work by a phenomenal team, dedicated to advancing American SOTA OSS AI. Lots of wins with Rnj-1. 1. SWE bench performance close
Today, we’re excited to introduce Rnj-1, @essential_ai's first open model; a world-class 8B base + instruct pair, built with scientific rigor, intentional design, and a belief that the advancement and equitable distribution of AI depend on building in the open. We bring
101
172
2K
+1. @TmlrOrg could remove the track to ICLR. Their system is much better, and the papers are A LOT more trustworthy than current ML confs
This ICLR is something else. We all know the review system is broken and the big ML conferences are basically lotteries, etc. But watching this year’s mess unfold in public hits differently. It feels like watching our academic field slowly die on livestream.
0
0
7
This was a great group effort ❤️. Check the thread below! My 2c: we train a 32B coding agent by distilling a strong teacher model on a mix of real and synthetic bugs generated by our new approach BugPilot 🛩️! BugPilot creates bugs unintentionally, by asking the teacher to
Excited to introduce our SoTA coding models, FrogBoss (32B) and FrogMini (14B), on SWE-Bench-Verified! (FrogBoss eats bugs… like a boss) 🐸🪲 These models were trained with bugs from a mix of existing and our new synthetic bug generation approach, called BugPilot. (1/n)
0
8
40
OTOH, feeling desperate after seeing a weird loss curve or some nan gradients... is our all-time low 😅
waking up and feeling excited to check your experiment jobs from last night is what peak ai researcher lifestyle looks like.
0
0
3
We'll host 2 interns next summer on @sergioksas's team at Waymo next summer, one of them hosted by myself. Apply if you'd like to work on problems at the very frontier of ML-based planning for L4 self-driving! https://t.co/qc7ve2zJEA
https://t.co/gMDzWBobNJ
9
16
161
(1/7) New paper!🚀 https://t.co/dq6yEzWyHg ✅Boltzmann distribution sampling for peptides up to 8 residues ✅4.3ms of training MD trajectories ✅Open-source codebase With @charliebtan, @leonklein26, Saifuddin Syed, @dom_beaini
@mmbronstein @AlexanderTong7 @k_neklyudov Read
8
53
212
this is 100% true and it holds for language, vision and other domains. none of the breakthroughs happened in a vacuum, e.g. data filtering relies on decades of NLP classifier advances.
This is truly an incredible breakthrough IMO. Really exemplifies what you get when deep domain expertise (population/disease genetics in this case) fuses with cleverly crafted ML. What u get r sleek, well thought out architectures that absolutely destroy the behemoths. Wow!! 1/
0
0
3
I don’t work in tech, so I can’t speak to whether that field has the right number of middle managers But I know that tech celebrates people who write actual code. It is a high status thing to do, not a job for losers who couldn’t get tenure
2
3
38
We @EITOxford just released these fully funded graduate scholarships in AI: multimodal deep learning discrete diffusion agents for scientific reasoning generalisation in ML machine learning interatomic potentials ML beyond attention
Applications for our Graduate Scholarships are now open! We’re looking for exceptional individuals from around the world to support our work and shape the future of science and technology.
0
4
15
Applications for our Graduate Scholarships are now open! We’re looking for exceptional individuals from around the world to support our work and shape the future of science and technology.
1
3
8
Even if you don't have papers, the PhD track is not the shortest path to wealth or "career success". It’s an investment in the human experience itself —in how you see the world, question knowledge, and define your own purpose.
i wonder- realistically what benefit does a phd give to a person who already has multiple A* papers and a 1000 citations? sounds like they know how to do research well, which is what the major learning of a phd is. if you're already done with that, why not go straight into
0
0
10
this is empirical science at its best. identify failure modes and update your belief/strategy
Hinton ~1983 thought Boltzmann Machines > backprop, but debugged himself out of the infatuation Boltzmann Machines failed to learn, so he printed out weights, 8 cm thick, and inspected them for weeks it's the dreaded local minimum so 1 yr later, in desperation, he tried backprop
0
0
3
The news is out! We're starting Blank Bio to build a computational toolkit assisted with RNA foundation models. If you want to see my flip between being eerily still and overly animated check out the video below! The core hypothesis is that RNA is the most customizable molecule
Blank Bio (@blankbio_) is building foundation models to power a computational toolkit for RNA therapeutics, starting with mRNA design and expanding to target ID, biomarker discovery, and more. https://t.co/7VRxSRgSKK Congrats on the launch, @hsu_jonny, @phil_fradkin & @ianshi3!
13
25
176
ICLR 2026 will take place in 📍Rio de Janeiro, Brazil 📅 April 23–27, 2026 Save the date - see you in Rio! #ICLR2026
9
102
1K
Lots of hype around “AI scientists” but they have all been focused on the fun parts, not the hard parts. The focus has been on hypothesis generation, rather than *validation*: the expensive and grueling part. It’s like having an AI write your poetry rather than do your
0
3
12
📄 Paper: https://t.co/aeVeADJDtq 💻 Code: https://t.co/Y1FjgyGz2U 🌍 Website: https://t.co/w6bIjCdNzg 🗂️ Dataset: https://t.co/LNIMItvGsU 📊 Eval:
huggingface.co
1
1
4
When designing AI scientist benchmarks, the challenge lies in how to simulate realistic experimental data. We find that the systems biology provide a great simulator for this! I am currently at ICML. Very excited to chat with anyone interested in this work!
What makes a great scientist? Most AI scientist benchmarks miss the key skill: designing and analyzing experiments. 🧪 We're introducing SciGym: the first simulated lab environment to benchmark #LLM on experimental design and analysis capabilities. #AI4SCIENCE #ICML25
0
3
6
What makes a great scientist? Most AI scientist benchmarks miss the key skill: designing and analyzing experiments. 🧪 We're introducing SciGym: the first simulated lab environment to benchmark #LLM on experimental design and analysis capabilities. #AI4SCIENCE #ICML25
6
22
145