Ai2

Our fully open releases give researchers the data, code, checkpoints, and methods they need to inspect claims, reproduce findings, and advance new science. Read more about why that’s so important to us. ⬇️ allenai.org/blog/who-get...

Who gets to understand AI? | Ai2 allenai.org

To gather feedback, we're releasing de-identified annotated tutoring transcripts plus code & scored replays. 💻 Code: https://t.co/BMctU7uEJS 🤗 Data: https://t.co/U6bmDmDuC0 🌐 Learn more: https://t.co/xaRYDUlWIk

allenai/tutormoments-preview · Datasets at Hugging Face huggingface.co

We build tools like infini-gram so anyone can check AI writing against the text a model may have learned from. 📝 Read more in our new blog: buff.ly/CAMPRzB

Tracing distinctive language in AI-written text | Ai2 allenai.org

What's next for OlmoEarth Platform: change-detection alerts, agentic tools for nonexperts, & precomputed global embeddings that could one day skip the forward pass entirely. Read more about the engineering in our latest blog: 🌐 buff.ly/wkAWugp

The OlmoEarth Platform: Geospatial inference at planetary scale | Ai2 allenai.org

This is the work that’s shaping what comes next for OlmoEarth. Learn more: allenai.org/olmoearth

OlmoEarth | Ai2 allenai.org

For a deeper look at what SciArena revealed about how scientists evaluate AI-generated answers to Qs about the scientific literature, read our NeurIPS 2025 Spotlight paper—which includes findings beyond the leaderboard. 📄 buff.ly/A9OrLph

SciArena: An Open Evaluation Platform for Non-Verifiable Scientific Literature-Grounded Tasks arxiv.org

o3 finished on top of the SciArena leaderboard – ahead of Claude Opus 4.1, Gemini 3 Pro Preview, & open-weights models like DeepSeek-R1 – with answers researchers called more detailed + to the point. Learn more in our updated blog: buff.ly/o1EyhhG

SciArena: A new platform for evaluating foundation models in scientific literature tasks | Ai2 allenai.org

Shippy is currently in private preview for select partners. For more on how we built it, what’s next, and how we’re ensuring users can trust and fully trace its outputs, read our new engineering blog: buff.ly/xur4hXq

What building Shippy taught us about building agents | Ai2 allenai.org

Try olmOCR 2 in the Ai2 Playground, check out our blog for more info, & download the weights and data from Hugging Face: ▶️ Playground: buff.ly/pJhoWXN 📝 Blog: buff.ly/8uUabID 🤗 Model & data: buff.ly/EUYHIuN

playground.allenai.org

Modular training is gaining momentum as frontier models become costlier to train & deploy. This project shows how open, distributed approaches can make development more practical for national projects, public institutions, & smaller teams. → Learn more: buff.ly/N0b3zI7

Modular LLMs at scale: How the Danish Foundation Models project is using FlexOlmo to pool national expertise without pooling sensitive data | Ai2 allenai.org

The pre-training dataset for OlmoEarth includes Sentinel-2, Sentinel-1, and Landsat satellite imagery, paired with various "maps" such as ESA WorldCover and the USDA Cropland Data Layer. For details, see: - Hugging Face: huggingface.co/datasets/all... - OlmoEarth v1 paper: arx…

allenai/olmoearth_pretrain_dataset · Datasets at Hugging Face huggingface.co

The pre-training dataset for OlmoEarth includes Sentinel-2, Sentinel-1, and Landsat satellite imagery, paired with various "maps" such as ESA WorldCover and the USDA Cropland Data Layer. For details, see: - Hugging Face: huggingface.co/datasets/all... - OlmoEarth v1 paper: arx…

arxiv.org

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. 🧵

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. 🧵

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. 🧵

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. 🧵

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. 🧵

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. 🧵

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. 🧵

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.

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. 🧵

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 ↓ 🎥