Mohit Bansal
Mohit Bansal studies multimodal learning and language generation, including research on reasoning, interpretability and faithful responses. Bansal earned an engineering degree at IIT Kanpur from 2004 to 2008 and completed doctoral study at Berkeley from 2008 to 2013. His dissertation with Dan Klein concerned web semantics for structured language processing. He also wrote a 2012 master’s thesis on parsing. Bansal coauthored ICML papers on symbolic mixtures of experts and multimodal fact-level attribution for verifiable reasoning. They are specific publications within his work on reasoning agents and language generation.
- Works
- 459
- Citations
- 20,023
- h-index
- 66
Publication numbers from OpenAlex, as of 14 October 2026.
Timeline
-
2008–2013
Training in structured language processing
Bansal earned an engineering degree at IIT Kanpur from 2004 to 2008 and completed doctoral study at Berkeley from 2008 to 2013. His dissertation with Dan Klein concerned web semantics for structured language processing. He also wrote a 2012 master’s thesis on parsing.
-
2013–2016
Research appointments
Bansal held a three-year research assistant professorship at Toyota Technological Institute at Chicago from 2013 to 2016. He joined UNC Chapel Hill in 2016 and led the MURGe Lab.
-
2017–2019
Support for an AI research program
He received a DARPA Young Faculty Award in 2017, followed by a DARPA Director’s Fellowship and an NSF CAREER Award in 2019. His research covers natural language processing, vision and multimodal learning, with attention to faithful generation and interpretable models.
-
2026
Reasoning and verifiable multimodal work
Bansal coauthored ICML papers on symbolic mixtures of experts and multimodal fact-level attribution for verifiable reasoning. They are specific publications within his work on reasoning agents and language generation.
Sources and credits
Photo: cs.unc.edu; Mohit Bansal; creator not stated. License: Not stated by source. Source: cs.unc.edu. Cropped to a square and resized. Ask us to remove or replace this photo.