iiinova — where your digital self keeps evolving

Ai2 @ai2.bsky.social · 27/08/2026

AI’s biggest role in science may not be answering questions. It may be helping scientists find which questions are worth asking. At Providence Swedish, AutoDiscovery surfaced an unexpected immune signal in a heavily studied cancer dataset—and follow-up research confirmed it. 🧵 buff.ly/LTacUQl

1153

Ai2 @ai2.bsky.social · 29/09/2026

Google Cloud put Olmo 3’s reproducibility to the test, rerunning our 7B pretraining & mid-training on Cloud TPUs and matching our original run on held-out evals. Reproducibility matters for science + trustworthy AI. That’s what fully open makes possible. 🤝 developers.googleblog.com/reproducing-...

developers.googleblog.com

Reproducing Olmo 3 7B Pre-training in MaxText: case study of large scale training on TPUs- Google Developers Blog

Learn how MaxText reproduced Ai2’s OLMo 3 7B on Google Cloud TPUs, matching PyTorch GPU benchmarks across pre-training with up to 57.4% MFU.

1405

Ai2 @ai2.bsky.social · 23/09/2026

Today at #ClimateWeekNYC, Ai2 and @globalfishingwatch.org announced a partnership to shape how AI and AI agents enter ocean monitoring and enforcement. Purpose-built. Open. Accountable to the people who use them. buff.ly/l7e7st4

161

Ai2 @ai2.bsky.social · 17/09/2026

Can a fully open model help make AI more “prosocial” through a game? Soham Padia built Steering Arena, where players try to elicit kind & respectful responses from Olmo 3. Surprisingly, strings like “Undert! AH :-) Rog Appl)” were highly effective. 👇 buff.ly/adP47yO

150

Ai2 @ai2.bsky.social · 14/09/2026

How should future scientists learn to interrogate AI tools for discovery? Read how UW students put AutoDiscovery to the test as part of an academic challenge earlier this year, then try the tool for yourself—credits are now extended through Dec. 31. 🧵 buff.ly/fXoQxxY

161

Ai2 @ai2.bsky.social · 09/09/2026

Training an LLM by showing it answers people prefer – and ones they don’t – can improve it while quietly worsening behaviors. Goodfire AI used Ai2’s open post-training stack to predict how a full training run would change responses to different prompts. 🧵 buff.ly/RuJxLM1

180

Ai2 @ai2.bsky.social · 08/09/2026

We're at #ECCV2026 with papers & talks across the conference. Come say hello and learn about our latest research!

051

Ai2 @ai2.bsky.social · 04/09/2026

ML emulators mimic climate processes faster than traditional models. Next is coupling atmosphere & ocean emulators so their predictions feed into each other as the simulation runs. With the E3SM team, we built a system that does that: SamudrACE-E3SMv3. 🧵 buff.ly/9ShSTCM

161

Reposted by Ai2

Maarten Sap @maartensap.bsky.social · 02/09/2026

Is the safety-capability tradeoff for LLMs real? Or could it be an artefact of the benchmarks we use to measure safety?? We did some explorations with psychometrics-inspired multi-dimensional IRT models and created BenchMIRT to explore these questions! See 🧵

0112

Ai2 @ai2.bsky.social · 01/09/2026

Do LLM safety & capability evals measure what they claim to? We built BenchMIRT to audit them + see which model abilities their Qs actually test. On BBQ, a social-bias eval, it found the Qs distinguished models more by reasoning ability than safety. 🧵 buff.ly/bTcvqJf

1213

Ai2 @ai2.bsky.social · 01/09/2026

At an event on August 27, we brought together AI researchers, scientists, & medical practitioners to explore what AI needs to do better to meaningfully advance science. Five ideas kept coming up. 🧵

110

Ai2 @ai2.bsky.social · 26/08/2026

A Thai research team adapted our Dolma data-curation toolkit to build Mangosteen, a 47B-token corpus for Thai LLMs. They used Dolma to filter widely used web datasets into a smaller corpus that improved Thai LLM performance despite using less data. 🧵 buff.ly/fvLSIru

1161

Ai2 @ai2.bsky.social · 21/08/2026

What kinds of training data shape different model capabilities? A Georgia Tech team used our fully open model flow to trace Olmo’s performance on social/general reasoning and social-science/STEM knowledge tests back to the types of text it trained on. 🧵 buff.ly/cQEzbUn

1110

Ai2 @ai2.bsky.social · 18/08/2026

A model can sound like it knows a drug—even when it doesn’t. Researchers at @utaustin.bsky.social, @northeasternu.bsky.social, & @mdanderson.bsky.social found LLMs often lack drug-specific knowledge and instead lean on morphology. They used Olmo to trace why. 🧵 buff.ly/PuyfnCq

1133

Ai2 @ai2.bsky.social · 07/08/2026

Today we're introducing a preview of TutorMoments, a framework that measures whether AI tutors can make one of the hardest calls in teaching: when to step in and help a student, & when to hold back and let them do the heavy thinking. 🧵

2516

Ai2 @ai2.bsky.social · 06/08/2026

We're expanding our partnership with @hf.co to accelerate open science. Our storage on the Hub is roughly tripling to ~2 petabytes, & our downloads now run at high speed—even for our largest datasets & multi-checkpoint models. 🧵

1281

Ai2 @ai2.bsky.social · 05/08/2026

📸 140+ people joined us during #SeattleTechWeek to learn how AI gets built at Ai2. Research lead Iz Beltagy walked through what it takes to train a truly open LLM—from large-scale training runs to eval before release. Thanks to everyone who came + asked thoughtful questions. 🤝

140

Ai2 @ai2.bsky.social · 31/07/2026

When a model writes, where do its words come from? Are they new, or do they match exactly with language it saw in training? An AI-writing detector can't tell you. @tuhinchakr.bsky.social's group at Stony Brook has been dissecting AI-generated prose with our infini-gram engine. 🧵

1131

Ai2 @ai2.bsky.social · 28/07/2026

The organizations best positioned to use Earth-observation models – those working in conservation, food security, & disaster response – often can't run these models at the scale they need. That's an infrastructure problem. So we built the OlmoEarth Platform to solve it. 🧵

160

Ai2 @ai2.bsky.social · 24/07/2026

As a nonprofit research institute dedicated to advancing open science, we're encouraged to see growing support for open models across the AI ecosystem. We believe the evidence behind advanced AI systems shouldn’t be locked up in a few hands.

191

Ai2 @ai2.bsky.social · 24/07/2026

160

Ai2 @ai2.bsky.social · 24/07/2026

110

Ai2 @ai2.bsky.social · 24/07/2026

120

Ai2 @ai2.bsky.social · 23/07/2026

Curious how AI gets built at Ai2? For #SeattleTechWeek, we're opening our Northlake office for a panel with the researchers building our open models. They'll talk through the deep technical work behind them, with time after to chat about what you're working on & network. 👇

111

Ai2 @ai2.bsky.social · 21/07/2026

Two updates to Asta, our ecosystem of AI agents for science: a one-click handoff from AutoDiscovery to Asta’s data analysis tools, & paper search that evaluates its own results + searches again when they fall short. 🧵

1102

Ai2 @ai2.bsky.social · 17/07/2026

What does it actually take to build cutting-edge AI systems? On July 30 during #SeattleTechWeek, the researchers behind Ai2's open models sit down to talk through the deep technical work behind them. 🧵

290

Ai2 @ai2.bsky.social · 16/07/2026

This week we hosted researchers from @cgiar.org, @servirglobal.bsky.social, NASA Harvest, & @msftresearch.bsky.social at our office to explore OlmoEarth for food security and natural resource management challenges. The CGIAR teams brought their own use cases and put OlmoEarth to work. 🧵

130

Ai2 @ai2.bsky.social · 16/07/2026

It was an honor to bring OlmoEarth to the Nature Positive Summit (@npinitiative.bsky.social) this week in Kumamoto, Japan, with our partners The Group on Earth Observations (@earthobservations.bsky.social). 🧵

180

Ai2 @ai2.bsky.social · 15/07/2026

We built SciArena to test how well AI models handle scientific literature questions, as judged by researchers. It's retiring July 15, and the results are in: ~1,700 users cast ~3,900 votes. Here's what they told us. 🧵

295

Ai2 @ai2.bsky.social · 13/07/2026

Our @skylightmarine.bsky.social team is rolling out Shippy, an AI agent for the people protecting our ocean in real time. A wrong answer at these stakes can send a patrol vessel miles off course. Here's the architecture we built so analysts can trust it. 🧵

142

Ai2 @ai2.bsky.social · 10/07/2026

olmOCR 2 is now in the Ai2 Playground—our home for our fully open text, video, and image understanding models. 🧵

1102

Ai2 @ai2.bsky.social · 10/07/2026

OlmoEarth turns Earth data into insights—in hours, not years. But it's only as powerful as the people using it. So, we brought them into one room. 🧵

1100

Ai2 @ai2.bsky.social · 09/07/2026

"Embeddings only get you so far… if you want the next level in performance, I think fine-tuning is the way to go." Ai2 research scientist @pjreddie.bsky.social explains how our partners customize OlmoEarth – our open-source Earth-observation models – to map crops, wildfire risk, & more. 👇

130

Ai2 @ai2.bsky.social · 08/07/2026

What can you build with a fully open robotics model in a weekend? 🤖 Robotics engineer @0xbinh.bsky.social used MolmoAct 2, our open vision-language-action model, in the voice-controlled robot that won @southparkcommons.bsky.social's AI hackathon. Watch our interview with him ↓ 🎥

0120

Ai2 @ai2.bsky.social · 06/07/2026

We're at #ICML2026 with papers & talks across the conference. Come say hello and learn about our latest research!

170

Ai2 @ai2.bsky.social · 02/07/2026

The Danish Foundation Models (DFM) project is adapting our modular FlexOlmo architecture into a lighter-weight system that runs on commodity hardware—putting collaborative model building within reach of smaller research groups & organizations. 🧵

1329

Ai2 @ai2.bsky.social · 01/07/2026

We're at #ACL2026 with papers & talks across the conference. Come say hello and learn about our latest research!

070

Ai2 @ai2.bsky.social · 29/06/2026

AI image generators don't "draw"—they follow a compass: the score function, which points toward more probable images. The same compass drives Bayesian sampling and plasma physics. We built DiScoFormer to estimate the score far better when data gets complex. 🧵

2141

Ai2 @ai2.bsky.social · 29/06/2026

Today we're releasing OlmoEarth v1.2, the latest in our family of open foundation models for Earth observation. 🌍 We've switched to rotary positional embeddings (RoPE), which reduces artifacts in the embeddings & gives a small performance boost. 🧵

2123

Ai2 @ai2.bsky.social · 25/06/2026

Hybrid (transformer–RNN) models are fast becoming a serious alternative to the transformer, but a big question remains: how do they process tokens differently & how does this impact performance? We compared our transformer (Olmo 3) & hybrid (Olmo Hybrid) models to find out. 🧵

1266

Ai2 @ai2.bsky.social · 18/06/2026

Learn how @thinkaisquared.bsky.social & Domyn used Olmo, our family of fully open language models, to build their own models for regulated industries like finance, healthcare, & the public sector. 🧵

121

Ai2 @ai2.bsky.social · 17/06/2026

We're releasing MolmoMotion, a 3D motion forecasting model. Given one or a few video frames, 3D points on an object, & an instruction like "Put the white bowl on the table," MolmoMotion predicts where those points will go over the next few seconds in a shared 3D world frame. 🧵

3277

Ai2 @ai2.bsky.social · 12/06/2026

Building an LLM means evaluating it over & over as it changes. Tweak a hyperparameter or scale the model up, & every new checkpoint sends you back through the same benchmarking loop. We're releasing olmo-eval, a workbench built for this kind of iterative model development. 🧵

183

Ai2 @ai2.bsky.social · 11/06/2026

LLMs are no longer created w/ human data alone. They rely on other models to generate & filter data, evaluate outputs, & guide dev work. So what is a modern LLM built on? Olmo 3 → 89 model + 183 dataset dependencies; Nemotron 3 → 273 + 560 We made ModSleuth to trace this. 🧵

15511

Reposted by Ai2

kylelwiggers.bsky.social @kylelwiggers.bsky.social · 10/06/2026

𝗔𝗖𝗘𝟮𝗦-𝗦𝗛𝗶𝗘𝗟𝗗+, our new climate emulator that learns to separate the effects of sea surface temperature & CO2, is now on @hf.co—check it out → huggingface.co/allenai/ACE2...

huggingface.co

allenai/ACE2S-SHiELD-plus · Hugging Face

We’re on a journey to advance and democratize artificial intelligence through open source and open science.

092

Ai2 @ai2.bsky.social · 09/06/2026

Today we're introducing ACE2S-SHiELD+, a climate emulator that learns to separate the effects of sea surface temperature & CO2. It accurately handles scenarios where previous versions of our ACE family of climate emulators produced inaccurate results. 🧵

1140

Ai2 @ai2.bsky.social · 03/06/2026

We're at #CVPR2026 with papers & talks across the conference. Come say hello and learn about our latest research!

051

Ai2 @ai2.bsky.social · 01/06/2026

We're extending AutoDiscovery early access through July 31. New accounts start with 500 Hypothesis Credits (one credit = one hypothesis), & any credits you already have will still work. 🧵

140

Ai2 @ai2.bsky.social · 28/05/2026

MolmoAct 2 artifacts have been downloaded 400K+ times in under 1 month. Today we're opening up the full code & training data: everything you need to fine-tune or build on our fully open robotics foundation model. 🧵

191

Ai2 @ai2.bsky.social · 22/05/2026

Most models are only evaluated on a fraction of the benchmarks out there. ArtifactLinker, our new system, predicts which ones would set a new state-of-the-art on benchmarks hosted on @hf.co, then runs the evaluation to verify. 🧵

1141