Ai2 ORG

Modeling, training, eval, and inference code for OLMo

Data and tools for generating and inspecting OLMo pre-training data.

An open-source platform for Visual AI.

Toolkit for linearizing PDFs for LLM datasets/training

olmOCR converts scanned PDFs and image documents into clean plain text or Markdown using a 7-billion-parameter vision model, handling tables, equations, multi-column layouts, and handwriting that classical OCR misses.

OLMoE: Open Mixture-of-Experts Language Models

An open-source NLP research library, built on PyTorch.

A now-archived Python NLP research library from the Allen Institute for AI that provided reusable components for training and evaluating language models, active development ended December 2022 with Hugging Face Transformers recommended as the successor.

AllenAI's post-training codebase

A modular RL library to fine-tune language models to human preferences

Longformer: The Long-Document Transformer

A full spaCy pipeline and models for scientific/biomedical documents.

A BERT model for scientific text.

Tensorflow implementation of contextualized word representations from bi-directional language models

Bi-directional Attention Flow (BiDAF) network is a multi-stage hierarchical process that represents context at different levels of granularity and uses a bi-directional attention flow mechanism to achieve a query-aware context representation without early summarization.

PyTorch building blocks for the OLMo ecosystem

🪐 Objaverse-XL is a Universe of 10M+ 3D Objects. Contains API Scripts for Downloading and Processing!

S2ORC: The Semantic Scholar Open Research Corpus: https://www.aclweb.org/anthology/2020.acl-main.447/

Expanding natural instructions

MultimodalC4 is a multimodal extension of c4 that interleaves millions of images with text.

Code for the Molmo Vision-Language Model

ImageNet classification using binary Convolutional Neural Networks

library supporting NLP and CV research on scientific papers

Official code for VisProg (CVPR 2023 Best Paper!)

No description.

Given a scholarly PDF, extract figures, tables, captions, and section titles.

RewardBench: the first evaluation tool for reward models.

Science Parse parses scientific papers (in PDF form) and returns them in structured form.

Code for the Molmo2 Vision-Language Model

No description.

Official Repository for MolmoAct2

SPECTER: Document-level Representation Learning using Citation-informed Transformers

Allen Institute for AI: WildDet3D: Scaling Promptable 3D Detection in the Wild

No description.

Organize your experiments into discrete steps that can be cached and reused throughout the lifetime of your research project.

Officially supported AllenNLP models

CVPR 2024: Language Guided Generation of 3D Embodied AI Environments.

Code associated with the Don't Stop Pretraining ACL 2020 paper

A template repo for Python packages

An open-source implementation of Whisper

Code and data for "Lumos: Learning Agents with Unified Data, Modular Design, and Open-Source LLMs"

Parsers for scientific papers (PDF2JSON, TEX2JSON, JATS2JSON)

🏘️ Scaling Embodied AI by Procedurally Generating Interactive 3D Houses

Software that makes labeling PDFs easy.

An interactive PDF reader.

A deep NLP library, based on Keras / tf, focused on question answering (but useful for other NLP too)

Provides a common interface to many IR ranking datasets.

An open source framework for research in Embodied-AI from AI2.

Reproducible, flexible LLM evaluations

One framework to evaluate any VLA model on any robot simulation benchmark.

Official Repository for MolmoAct

An end-to-end open ecosystem for robot learning

ScienceWorld is a text-based virtual environment centered around accomplishing tasks from the standardized elementary science curriculum.

Friends of OLMo and their links.

CodeScientist: An automated scientific discovery system for code-based experiments

No description.

Quality information extraction at web scale. Edit

No description.

Repo housing the open sourced code for the ai2 scholar qa app and also the corresponding library

No description.

Public space for the user community of Semantic Scholar APIs to share scripts, report issues, and make suggestions.

Data and models for the SciFact verification task.

📷 Scripts for rendering Objaverse

No description.

Benchmarking LLMs with Challenging Tasks from Real Users

Earth system foundation model data, training, and eval

frozen-in-time version of our Paper Finder agent for reproducing evaluation results

A tool to use the Ai2 Open Coding Agents Soft-Verified Efficient Repository Agents (SERA) model with Claude Code

No description.

What's In My Big Data (WIMBD) - a toolkit for analyzing large text datasets

Dataset Cartography: Mapping and Diagnosing Datasets with Training Dynamics

A virtual environment for developing and evaluating automated scientific discovery agents.

No description.

Get started with CORD-19

Official code for NeurIPS 2025 paper "AutoDiscovery: Open-ended Scientific Discovery via Bayesian Surprise"

Pretraining Efficiently on S2ORC!

Dataset of medical images, captions, subfigure-subcaption annotations, and inline textual references

Staging area for a public release of Theorizer

multimodal document analysis

ACL 2025: Synthetic data generation pipelines for text-rich images.

No description.

SPOC: Imitating Shortest Paths in Simulation Enables Effective Navigation and Manipulation in the Real World

No description.

Code and training scripts for FlexOlmo

Discovering Data-driven Hypotheses in the Wild

Dataset accompanying the SPECTER model

Data generation and training repository for SERA: Soft-Verified Efficient Repository Agents.

No description.

No description.

No description.

Code for Bolmo: Byteifying the Next Generation of Language Models

Repository for NAACL 2019 paper on Citation Intent prediction

Open One-Stop Moderation Tools for Safety Risks, Jailbreaks, and Refusals of LLMs

The ProcTHOR-10K Houses Dataset

Official repository for the A-OKVQA dataset

The Semantic Scholar Search Reranker

No description.

Semantic Scholar's Author Disambiguation Algorithm & Evaluation Suite

PoliFormer: Scaling On-Policy RL with Transformers Results in Masterful Navigators

No description.

A simple evaluation of generative language models and safety classifiers.

Repository for Decomposed Prompting

RoboTHOR Challenge

No description.

No description.

SciRepEval benchmark training and evaluation scripts

A tool for developing remote sensing datasets and models.

Code and website for "MolmoB0T: Large-Scale Simulation Enables Zero-Shot Manipulation".

No description.

No description.

Tooling for exact and MinHash deduplication of large-scale text datasets

No description.

No description.

OlmoEarth projects

Probabilistic Neural Programming

OLMost every training recipe you need to perform data interventions with the OLMo family of models.

[arXiv 2025] SAGE: Training Smart Any-Horizon Agents for Long Video Reasoning with Reinforcement Learning

CodeNav is an LLM agent that navigates and leverages previously unseen code repositories to solve user queries.

No description.

Efficient and low latency real-time global-scale GPS trajectory modeling

📱👉🏠 Perform conditional procedural generation to generate houses like your own!

Code/data for MARG (multi-agent review generation)

Collection of public APIs for embedding scientific papers

No description.

A corpus and code for understanding norms and subjectivity. 🤖

Data mapping framework for rust stuff

No description.

No description.

A file utility for accessing both local and remote files through a unified interface.

No description.

NLP toolkit (tokenizer, POS-tagger, parser, etc.)

No description.

No description.

No description.

No description.

No description.

ArtifactLinker: Linking Scientific Artifacts for Automatic State-of-the-Art Discovery

MultiCite code and data. Models are available on Huggingface.

No description.

decontamination

PreScience: A Benchmark for Forecasting Scientific Contributions

Measuring the Signal to Noise Ratio in Language Model Evaluation

Reasoning by Communicating with Agents

Fluid Language Model Benchmarking

No description.

Code for NAACL 2022 paper "Reframing Human-AI Collaboration for Generating Free-Text Explanations"

Learning to route instances for Human vs AI Feedback (ACL Main '25)

Benchmarking Generalization to New Tasks from Natural Language Instructions

This repository contains data, code and models for contextual noncompliance.

A machine learning software for extracting information from scholarly documents

SBT Plugins for AI2 projects

No description.

S2APLER: S2 Agglomeration of Papers with Low Error Rate (it's for academic paper clustering)

No description.

No description.

Demo for https://arxiv.org/abs/2411.14199

Can VLMs understand students' hand-drawn math work?

A web application for playing 20 Questions to crowdsource common sense. 🤖

No description.

No description.

Synthetic GUI Pointing Data Generation

No description.

🤗 LeRobot: Making AI for Robotics more accessible with end-to-end learning

A fast AWS S3 browser, with inspiration from s5cmd

No description.

Code associated with the paper: "Few-Shot Self-Rationalization with Natural Language Prompts"

distance to the nearest coastline is all you need.

No description.

Panda ("plan-and-act") agent for Autonomous Scientific Discovery

Official Repository for STTS.

Code for the paper "Cracks in the Foundation: Seemingly Minor Architectural Choices Impact Long Context Extension"

No description.

A collection of useful utility classes and functions.

No description.

GrapAL landing page, user guide & documentation

Format and normalize Chinese names into Western forms

Deprecated: use https://github.com/allenai/playground

Deprecated: use https://github.com/allenai/playground

Repository for task scaling laws using model ladders

Service for sharing data resource between asta users, clients, and agents

A Python Library for Document Layout Understanding

Official implementation of "DRACULA: Hunting for the Actions Users Want Deep Research Agents to Execute"

No description.

Resource manager for MolmoSpaces

A set of helper functions for processing and integrating visual inputs with Molmo

No description.

No description.

Common real-robot infra for Ai2 Robotics

GitHub actions for skiff2 repositories.

Workspace manager for coding agents. Interactively solve and develop Harbor tasks.

Repo for collection of feedback on OlmoEarth

No description.

Evaluation code for the paper "Language Models Don't Know What You Want: Evaluating Personalization in Deep Research Needs Real Users"

Execution environments for OLMES

No description.

No description.

No description.

No description.

Homebrew tap to install allenai/s6ui

CLI for merging tokenized shard .npy files and remapping .csv.gz metadata offsets.

Policy zoo for data generation + evaluation in MolmoSpaces

Facebook AI Research Sequence-to-Sequence Toolkit written in Python.