Inventor · disambiguated record
Mostafa Dehghani
Also filed as: DEHGHANI MOSTAFA
10 granted patents·12 pending applications·17 citations·filing 2019–2025
83Inventor score
Top patents by PatentIndex Score
22 records- 0190US11983903B2Processing images using self-attention based neural networksGOOGLE LLC·Filed 2023·Granted May 14, 2024·2 cites·21 claims
- 0289US12112538B2Systems and methods for improved video understandingGOOGLE LLC·Filed 2021·Granted Oct 8, 2024·2 cites·16 claims
- 0389US10740433B2Universal transformersGOOGLE LLC·Filed 2019·Granted Aug 11, 2020·9 cites·27 claims
- 0488US12125247B2Processing images using self-attention based neural networksGOOGLE LLC·Filed 2021·Granted Oct 22, 2024·2 cites·20 claims
- 0586US11860969B2Universal transformersGOOGLE LLC·Filed 2020·Granted Jan 2, 2024·2 cites·20 claims
- 0678US2025005797A1Processing images using self-attention based neural networksGOOGLE LLC·Filed 2024·Application pending·0 cites
- 0778US2025005798A1Processing images using self-attention based neural networksGOOGLE LLC·Filed 2024·Application pending·0 cites
- 0877US2024428587A1Systems and Methods for Improved Video UnderstandingGOOGLE LLC·Filed 2024·Application pending·0 cites
- 0977US2024428586A1Systems and Methods for Improved Video UnderstandingGOOGLE LLC·Filed 2024·Application pending·0 cites
- 1064US2024169184A1Efficient decoding of output sequences using adaptive early exitingGOOGLE LLC·Filed 2024·Application pending·0 cites
- 1162US11886976B1Efficient decoding of output sequences using adaptive early exitingGOOGLE LLC·Filed 2023·Granted Jan 30, 2024·0 cites·20 claims
- 1257US2024346824A1Action localization in videos using learned queriesGOOGLE LLC·Filed 2024·Application pending·0 cites
- 1356US2024256964A1Pretraining Already-Pretrained Models for Diverse Downstream TasksGOOGLE LLC·Filed 2024·Application pending·0 cites
- 1454US12511539B2Computer vision neural networks with learned tokenizationGOOGLE LLC·Filed 2022·Granted Dec 30, 2025·0 cites·20 claims
- 1553US2024256835A1Training ultra-large-scale vision transformer neural networksGOOGLE LLC·Filed 2024·Application pending·0 cites
- 1652US12394191B2Neural networks based multimodal transformer for multi-task user interface modelingGOOGLE LLC·Filed 2022·Granted Aug 19, 2025·0 cites·24 claims
- 1751US2024403636A1Self-attention based neural networks for processing network inputs from multiple modalitiesGOOGLE LLC·Filed 2022·Application pending·0 cites
- 1851US2025156756A1Systems and Methods for Pretraining Models for Diverse Downstream TasksGOOGLE LLC·Filed 2022·Application pending·0 cites
- 1951US2023244938A1Using Chains of Thought to Prompt Machine-Learned Models Pre-Trained on Diversified ObjectivesGOOGLE LLC·Filed 2023·Application pending·0 cites
- 2050US2024257511A1Neural Network Architectures with Multiple Normalization Layers for Machine VisionGOOGLE LLC·Filed 2024·Application pending·0 cites
- 2146US12511521B2Machine-learned attention models featuring echo-attention layersGOOGLE LLC·Filed 2022·Granted Dec 30, 2025·0 cites·20 claims
- 2246US12469186B2Systems and methods for generating multimodal data using a single-tower architecture with a data generation subsystemGDM HOLDING LLC·Filed 2025·Granted Nov 11, 2025·0 cites·25 claims
Join the waitlist — get patent alerts
Get an alert when Mostafa Dehghani files or is granted a new patent.
We store only your email — no account needed. See our privacy policy.
Identity basis: PatentsView inventor disambiguation (2025Q4-odp release). How scoring works →