I’m a research scientist at Google DeepMind in Seattle. My current interests are centered around agents, tool use, and multimodal learning. I often work on solutions involving self-supervised methods and reinforcement learning.

Before joining Google, I completed my PhD at the University of Washington where I was advised by Luke Zettlemoyer and Dan Weld. When I am not doing Computer Science-y stuff, I like to hike, travel, and read.

Selected Publications

For the full list, please refer to my Google Scholar page.

• Gemini Team, Google. Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities. ArXiv 2507.06261, 2025.

• Shikhar Murty, Christopher Manning, Peter Shaw, Mandar Joshi, Kenton Lee. BAGEL: Bootstrapping Agents by Guiding Exploration with Language. ICML 2024.

• Peter Shaw, Mandar Joshi, James Cohan, Jonathan Berant, Panupong Pasupat, Hexiang Hu, Urvashi Khandelwal, Kenton Lee, Kristina Toutanova. From Pixels to UI Actions: Learning to Follow Instructions via Graphical User Interfaces. NeurIPS 2023. * equal contribution

• Google AI Team. PaLI-X: On Scaling up a Multilingual Vision and Language Model. ArXiv 2305.18565, 2023.

• Kenton Lee*, Mandar Joshi*, Iulia Turc, Hexiang Hu, Fangyu Liu, Julian Eisenschlos, Urvashi Khandelwal, Peter Shaw, Ming-Wei Chang, Kristina Toutanova. Pix2Struct: Screenshot Parsing as Pretraining for Visual Language Understanding. ArXiv 2210.03347, 2022.
  * equal contribution

• Mandar Joshi, Terra Blevins, Mike Lewis, Daniel S. Weld, Luke Zettlemoyer. Few-shot Mining of Naturally Occurring Inputs and Outputs . ArXiv 2205.04050, 2022.

• Armen Aghajanyan, Bernie Huang, Candace Ross, Vladimir Karpukhin, Hu Xu, Naman Goyal, Dmytro Okhonko, Mandar Joshi, Gargi Ghosh, Mike Lewis, Luke Zettlemoyer. CM3: A Causal Masked Multimodal Model of the Internet . ArXiv 2201.07520, 2022.

• Armen Aghajanyan, Dmytro Okhonko, Mike Lewis, Mandar Joshi, Hu Xu, Gargi Ghosh, Luke Zettlemoyer. HTLM: Hyper-Text Pre-Training and Prompting of Language Models. ArXiv 2107.06955, 2021.

• Weijia Shi, Mandar Joshi, Luke Zettlemoyer. DESCGEN: A Distantly Supervised Dataset for Generating Abstractive Entity Descriptions. ACL 2021.

• Arie Cattan, Alon Eirew, Gabriel Stanovsky, Mandar Joshi, Ido Dagan. Cross-document Coreference Resolution over Predicted Mentions. ACL 2021 Findings (Short).

• Arie Cattan, Alon Eirew, Gabriel Stanovsky, Mandar Joshi, Ido Dagan. Streamlining Cross-Document Coreference Resolution: Evaluation and Modeling. ArXiv 2009.11032, 2020.

• Terra Blevins, Mandar Joshi, Luke Zettlemoyer. FEWS: Large-Scale, Low-Shot Word Sense Disambiguation with the Dictionary. EACL 2021.

• Bhargavi Paranjape, Mandar Joshi, John Thickstun, Hannaneh Hajishirzi, Luke Zettlemoyer. An Information Bottleneck Approach to Controlling Conciseness in Rationale Extraction. EMNLP, 2020. [ code ]

• Mandar Joshi, Kenton Lee, Yi Luan, Kristina Toutanova. Contextualized Representations Using Textual Encyclopedic Knowledge. ArXiv:2004.12006, 2020.

• Yinhan Liu, Myle Ott, Naman Goyal, Jingfei Du, Mandar Joshi, Danqi Chen, Omer Levy, Mike Lewis, Luke Zettlemoyer, Veselin Stoyanov. RoBERTa: A Robustly Optimized BERT Pretraining Approach . ArXiv:1907.11692, 2019. [ code ]

• Mandar Joshi*, Danqi Chen*, Yinhan Liu, Daniel S. Weld, Luke Zettlemoyer, Omer Levy. SpanBERT: Improving Pre-training by Representing and Predicting Spans. TACL, 2019. Equal Contribution [ code ]
  * equal contribution

• Mandar Joshi, Omer Levy, Daniel S. Weld, Luke Zettlemoyer. BERT for Coreference Resolution: Baselines and Analysis. (Short) Proceedings of Emperical Methods in Natural Language Processing (EMNLP), 2019. [ code ]

• Mandar Joshi, Eunsol Choi, Omer Levy, Daniel S. Weld, Luke Zettlemoyer. pair2vec: Compositional Word-Pair Embeddings for Cross-Sentence Inference. Conference of the North American Chapter of the Association for Computational Linguistics (NAACL), 2019. [<a href=”https://github.com/mandarjoshi90/pair2vec”, target=’_blank’> code </a>]

• Mandar Joshi, Eunsol Choi, Daniel Weld, Luke Zettlemoyer. TriviaQA: A Large Scale Distantly Supervised Challenge Dataset for Reading Comprehension. Association for Computational Linguistics (ACL) 2017. [ website ] [ bib ]

• Mandar Joshi, Uma Sawant, Soumen Chakrabarti. Knowledge Graph and Corpus Driven Segmentation and Answer Inference for Telegraphic Entity-seeking Queries. Empirical Methods in Natural Language Processing (EMNLP) 2014.