Pradeep Dasigi
Lead Research Scientist on the Olmo team at Ai2; credited with technical leadership on Olmo 3
- Lead Research Scientist, Olmo team, Allen Institute for AI (Ai2)13
Pradeep Dasigi is a Lead Research Scientist on the Olmo team at the Allen Institute for AI (Ai2), working on open language models such as OLMo and Tülu with a focus on post-training. The Olmo 3 report lists Dasigi among the authors credited with technical leadership and cross-workstream contributions, and Dasigi is a co-author of the Tülu 3 post-training report. Dasigi's earlier work covers reading-comprehension benchmarks, training data curation and robustness of learning algorithms.1245
Career
Dasigi is a Lead Research Scientist on the Olmo team at Ai2 and is listed in Ai2's team directory. Dasigi's page notes past work with the AllenNLP team and describes current interest in training language and vision-language models for long-horizon agentic tasks such as coding, search and web browsing.13
Research
Dasigi's early benchmark work includes DROP (2019), a reading-comprehension benchmark requiring discrete reasoning, and Quoref (2019) and Qasper (2021), both as first author. Dasigi's research page groups later work under efficient adaptation of language models and robustness to distribution shifts.7892
On open language-model recipes, Dasigi co-authored "How Far Can Camels Go?" (2023) on instruction tuning with open resources, the Tülu 3 report (2024) and the Olmo 3 report (2025), where Dasigi is credited for technical leadership alongside work on post-training and evaluation.10542
Work
Olmo 3 7B InstructOpen weights
Credited in the Olmo 3 report for technical leadership and cross-workstream contributions, and for post-training infrastructure, evaluation and mid- and post-training data work.4
Olmo 3.1 32B ThinkOpen weights
Among the authors credited with technical leadership on the Olmo 3 report, which introduces Olmo 3.1 Think 32B as the project's flagship reasoning model.4
Llama 3.1 Tülu 3.1 8BOpen weights
Co-author of the Tülu 3 report on open language model post-training, which the Tülu 3.1 8B model card cites as its paper.56
Qasper (external site: arxiv.org)Open data
First author of the paper introducing a dataset of information-seeking questions and answers anchored in research papers.9
Career timeline
Mar 2019
Co-authored DROP, a reading-comprehension benchmark requiring discrete reasoning over paragraphs.7
Aug 2019
Posted Quoref, a reading-comprehension dataset on coreferential reasoning, as first author.8
May 2021
Posted Qasper, a dataset of questions and answers anchored in research papers, as first author.9
Jun 2023
Co-authored "How Far Can Camels Go?" on instruction tuning with open resources.10
Nov 2024
Co-authored the Tülu 3 report on open language model post-training.5
Dec 2025
Credited with technical leadership on the Olmo 3 report.4
Related people
More sourced profiles at the same organization, in the same field, or from the same school.
- Akshita BhagiaAlso at Ai2
- Adina WilliamsEvaluation and benchmarks
- Christopher ClarkAlso at Allen Institute for AI
- Eugene YanLanguage models
- Dieter FoxAlso at Ai2
- Jesse DodgeAlso at Ai2
- Hannaneh HajishirziAlso at Ai2
- Kyle LoAlso at Ai2
- Maarten SapAlso at Allen Institute for AI
- Leo GaoLanguage models
- Noah A. SmithAlso at Ai2
- Marina DanilevskyEvaluation and benchmarks
Is this you?
Profiles cover public professional work only, and every line cites a source. If something is out of date, missing or wrong, send a correction with a link to an official page and we will update it. Suggest a correction
Add this badge to your website or GitHub profile README. It links back to this page. It is not an endorsement or a rating.
Markdown (for a README)
[](https://unitedstatesofamericasuperintelligence.com/people/pradeep-dasigi/)
HTML
<a href="https://unitedstatesofamericasuperintelligence.com/people/pradeep-dasigi/"><img src="https://unitedstatesofamericasuperintelligence.com/badges/profile.svg" alt="Pradeep Dasigi — Profile on USASI" height="20"></a>
This profile covers public professional work only, from the sources below. It is not an endorsement or a ranking. To correct or remove it, use the correction form.