_version_ 1866913690808221696
author Granite Vision Team
Karlinsky, Leonid
Arbelle, Assaf
Daniels, Abraham
Nassar, Ahmed
Alfassi, Amit
Wu, Bo
Schwartz, Eli
Joshi, Dhiraj
Kondic, Jovana
Shabtay, Nimrod
Li, Pengyuan
Herzig, Roei
Abedin, Shafiq
Perek, Shaked
Harary, Sivan
Barzelay, Udi
Goldfarb, Adi Raz
Oliva, Aude
Wieles, Ben
Bhattacharjee, Bishwaranjan
Huang, Brandon
Auer, Christoph
Gutfreund, Dan
Beymer, David
Wood, David
Kuehne, Hilde
Hansen, Jacob
Shtok, Joseph
Wong, Ken
Bathen, Luis Angel
Mishra, Mayank
Lysak, Maksym
Dolfi, Michele
Yurochkin, Mikhail
Livathinos, Nikolaos
Harel, Nimrod
Azulai, Ophir
Naparstek, Oshri
de Lima, Rafael Teixeira
Panda, Rameswar
Doveh, Sivan
Gupta, Shubham
Das, Subhro
Zawad, Syed
Kim, Yusik
He, Zexue
Brooks, Alexander
Goodhart, Gabe
Govindjee, Anita
Leist, Derek
Ibrahim, Ibrahim
Soffer, Aya
Cox, David
Soule, Kate
Lastras, Luis
Desai, Nirmit
Ofek-koifman, Shila
Raghavan, Sriram
Syeda-Mahmood, Tanveer
Staar, Peter
Drory, Tal
Feris, Rogerio
author_facet Granite Vision Team
Karlinsky, Leonid
Arbelle, Assaf
Daniels, Abraham
Nassar, Ahmed
Alfassi, Amit
Wu, Bo
Schwartz, Eli
Joshi, Dhiraj
Kondic, Jovana
Shabtay, Nimrod
Li, Pengyuan
Herzig, Roei
Abedin, Shafiq
Perek, Shaked
Harary, Sivan
Barzelay, Udi
Goldfarb, Adi Raz
Oliva, Aude
Wieles, Ben
Bhattacharjee, Bishwaranjan
Huang, Brandon
Auer, Christoph
Gutfreund, Dan
Beymer, David
Wood, David
Kuehne, Hilde
Hansen, Jacob
Shtok, Joseph
Wong, Ken
Bathen, Luis Angel
Mishra, Mayank
Lysak, Maksym
Dolfi, Michele
Yurochkin, Mikhail
Livathinos, Nikolaos
Harel, Nimrod
Azulai, Ophir
Naparstek, Oshri
de Lima, Rafael Teixeira
Panda, Rameswar
Doveh, Sivan
Gupta, Shubham
Das, Subhro
Zawad, Syed
Kim, Yusik
He, Zexue
Brooks, Alexander
Goodhart, Gabe
Govindjee, Anita
Leist, Derek
Ibrahim, Ibrahim
Soffer, Aya
Cox, David
Soule, Kate
Lastras, Luis
Desai, Nirmit
Ofek-koifman, Shila
Raghavan, Sriram
Syeda-Mahmood, Tanveer
Staar, Peter
Drory, Tal
Feris, Rogerio
contents We introduce Granite Vision, a lightweight large language model with vision capabilities, specifically designed to excel in enterprise use cases, particularly in visual document understanding. Our model is trained on a comprehensive instruction-following dataset, including document-related tasks, such as content extraction from tables, charts, diagrams, sketches, and infographics, as well as general image tasks. The architecture of Granite Vision is centered around visual modality alignment with a decoder-only, 2 billion parameter Granite large language model. Additionally, we introduce a dedicated safety classification approach in test-time that leverages a sparse set of attention vectors to identify potential harmful inputs. Despite its lightweight architecture, Granite Vision achieves strong results in standard benchmarks related to visual document understanding, as well as on the LiveXiv benchmark, which is designed to avoid test set contamination by using a constantly updated corpus of recently published Arxiv papers. We are releasing the model under the Apache-2 license, allowing for both research and commercial use, while offering complete visibility into the training data and other relevant details. See https://huggingface.co/ibm-granite/ for model weights.
format Preprint
id arxiv_https___arxiv_org_abs_2502_09927
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Granite Vision: a lightweight, open-source multimodal model for enterprise Intelligence
Granite Vision Team
Karlinsky, Leonid
Arbelle, Assaf
Daniels, Abraham
Nassar, Ahmed
Alfassi, Amit
Wu, Bo
Schwartz, Eli
Joshi, Dhiraj
Kondic, Jovana
Shabtay, Nimrod
Li, Pengyuan
Herzig, Roei
Abedin, Shafiq
Perek, Shaked
Harary, Sivan
Barzelay, Udi
Goldfarb, Adi Raz
Oliva, Aude
Wieles, Ben
Bhattacharjee, Bishwaranjan
Huang, Brandon
Auer, Christoph
Gutfreund, Dan
Beymer, David
Wood, David
Kuehne, Hilde
Hansen, Jacob
Shtok, Joseph
Wong, Ken
Bathen, Luis Angel
Mishra, Mayank
Lysak, Maksym
Dolfi, Michele
Yurochkin, Mikhail
Livathinos, Nikolaos
Harel, Nimrod
Azulai, Ophir
Naparstek, Oshri
de Lima, Rafael Teixeira
Panda, Rameswar
Doveh, Sivan
Gupta, Shubham
Das, Subhro
Zawad, Syed
Kim, Yusik
He, Zexue
Brooks, Alexander
Goodhart, Gabe
Govindjee, Anita
Leist, Derek
Ibrahim, Ibrahim
Soffer, Aya
Cox, David
Soule, Kate
Lastras, Luis
Desai, Nirmit
Ofek-koifman, Shila
Raghavan, Sriram
Syeda-Mahmood, Tanveer
Staar, Peter
Drory, Tal
Feris, Rogerio
Computer Vision and Pattern Recognition
Artificial Intelligence
We introduce Granite Vision, a lightweight large language model with vision capabilities, specifically designed to excel in enterprise use cases, particularly in visual document understanding. Our model is trained on a comprehensive instruction-following dataset, including document-related tasks, such as content extraction from tables, charts, diagrams, sketches, and infographics, as well as general image tasks. The architecture of Granite Vision is centered around visual modality alignment with a decoder-only, 2 billion parameter Granite large language model. Additionally, we introduce a dedicated safety classification approach in test-time that leverages a sparse set of attention vectors to identify potential harmful inputs. Despite its lightweight architecture, Granite Vision achieves strong results in standard benchmarks related to visual document understanding, as well as on the LiveXiv benchmark, which is designed to avoid test set contamination by using a constantly updated corpus of recently published Arxiv papers. We are releasing the model under the Apache-2 license, allowing for both research and commercial use, while offering complete visibility into the training data and other relevant details. See https://huggingface.co/ibm-granite/ for model weights.
title Granite Vision: a lightweight, open-source multimodal model for enterprise Intelligence
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2502.09927