
Predoctoral Researcher
Allen Institute for AI
Jacob Morrison
Hi! I'm a predoctoral researcher on the AllenNLP team at Ai2. I'm also an incoming NLP PhD student at UW, where I'll be working with Banghua Zhu, and I'll be supported by an NSF Computer Science Graduate Fellowship.I've previously worked on code generation at Google [x] and platform health at Twitter, and in a previous life I was a software engineer at Tableau and Google. See my CV for more details.
Research
My research is generally focused on making language models broadly useful and reliable, and understanding their downstream societal impacts. In addition to my research I also spend a portion of my time helping policymakers understand and address the societal impacts of AI.
A few selected interests of mine and relevant publications:
data-centric methods for improving model capabilities
Selected Work
- Developing new post-training recipes and methods: Tülu 3, Merge to Learn
- New and interesting model evaluations: RewardBench, RewardBench 2, Bidimensional Leaderboards
open science and democratizing access to large models
Selected Awards & Fellowships
- May 2025: Mercor Graduate Fellowship Finalist
- August 2024: ACL Theme Paper Award
- August 2024: ACL Best Resource Paper Award
- April 2024: GeekWire Innovation of the Year
- August 2023: NSF Computer Science Graduate Fellowship
Publications
2025
-
FlexOLMo: Open Language Models for Flexible Data Use
Under review
-
RewardBench 2: Advancing Reward Model Evaluation
Under review
- 2 OLMo 2 Furious
2024
- Tülu 3: Pushing Frontiers in Open Language Model Post-Training
-
Holistically Evaluating the Environmental Impact of Creating Language Models
ICLR 2025 🔦 Spotlight Paper 🔦
-
OLMoE: Open Mixture-of-Experts Language Models
NeurIPS 2024 Workshop on Efficient Natural Language and Speech Processing (spotlight) 🔦 Spotlight Paper 🔦ICLR 2025 paper 🗣️ Oral Presentation 🗣️
-
Merge to Learn: Efficiently Adding Skills to Language Models with Model Merging
Findings of EMNLP 2024 paper
-
SciRIFF: A Resource to Enhance Language Model Instruction-Following over Scientific Literature
arXiv paper
- RewardBench: A Benchmark for Evaluating Reward Models
-
Intentionally Unintentional Speech: Why Generative AI Models Are Not Protected by the First Amendment
First Amendment Law Review (University of North Carolina), Spring 2025 paper
-
Unsettled Law in the Age of Generative AI: Time to Generate New Approaches?
Journal of Law and Technology at Texaspaper
-
A Legal Risk Taxonomy for Generative Artificial Intelligence
arXiv preprint paper
-
OLMo: Accelerating the Science of Language Models
ACL 2024 paper 🥇 Theme Paper Award 🥇
-
Dolma: an Open Corpus of Three Trillion Tokens for Language Model Pretraining Research
ACL 2024 paper 🥇 Best Resource Paper Award 🥇