Pradeep Dasigi

Pradeep Dasigi is an AI researcher whose work spans neural question answering, reading-comprehension benchmarks and open language models. He coauthored a 2017 semantic parser for questions over Wikipedia tables, earned a Carnegie Mellon PhD in 2019 and joined Ai2 as a research scientist. He coauthored DROP and Qasper, and open model projects Tulu 3, Olmo 3 and DR Tulu. As of October 2026, he is a Lead Research Scientist on Ai2's Olmo team, where he focuses on post-training and long-horizon language-model agents.

Now at: Allen Institute for AI (Ai2)
Works: 90
Citations: 3,447
h-index: 20

Publication numbers from OpenAlex, as of 22 October 2026.

Timeline

2017

Parsing questions over tables

In 2017, he coauthors a neural semantic parser for compositional questions over semi-structured Wikipedia tables. Its decoder uses a type-constrained grammar and entity linking to answer those questions.
2019

A Carnegie Mellon PhD

He earns a PhD in Language and Information Technology at Carnegie Mellon's Language Technologies Institute in 2019 and joins Ai2 as a research scientist. He co-hosts Ai2's NLP Highlights podcast with Matt Gardner and Waleed Ammar.
2019

The DROP benchmark

In 2019, he coauthors DROP, an English reading-comprehension benchmark that requires discrete reasoning over paragraph content.
2021

Questions over research papers

In 2021, he coauthors Qasper, a dataset of 5,049 questions over 1,585 natural-language-processing papers. It targets questions that require reading full papers, and its answers include supporting evidence.
2024

Tulu 3 post-training

In 2024, he coauthors Tulu 3, a family of fully open post-trained language models. The release includes its data, code and training recipes.
2025

The Olmo 3 model family

In 2025, he coauthors Olmo 3, a fully open family of 7B and 32B language models. The release includes the complete model flow, with every stage, checkpoint, data point and dependency.
2026

Deep research with DR Tulu

In 2026, he coauthors DR Tulu, a deep-research project that trains agents with Reinforcement Learning with Evolving Rubrics. The team releases DR Tulu-8B, its data and code.
As of October 2026

Lead scientist on Olmo

As of October 2026, he is a Lead Research Scientist on the Olmo team at the Allen Institute for AI. He works on open language models and explores training vision-language models for open-ended, long-horizon agent tasks such as coding, search and web browsing.