Zobrazeno 1 - 10
of 9 148
pro vyhledávání: '"A Feizi"'
Image-text contrastive models such as CLIP learn transferable and robust representations for zero-shot transfer to a variety of downstream tasks. However, to obtain strong downstream performances, prompts need to be carefully curated, which can be a
Externí odkaz:
http://arxiv.org/abs/2406.13683
The increasing size of large language models (LLMs) challenges their usage on resource-constrained platforms. For example, memory on modern GPUs is insufficient to hold LLMs that are hundreds of Gigabytes in size. Offloading is a popular method to es
Externí odkaz:
http://arxiv.org/abs/2406.11674
Autor:
Zarei, Arman, Rezaei, Keivan, Basu, Samyadeep, Saberi, Mehrdad, Moayeri, Mazda, Kattakinda, Priyatham, Feizi, Soheil
Recent text-to-image diffusion-based generative models have the stunning ability to generate highly detailed and photo-realistic images and achieve state-of-the-art low FID scores on challenging image generation benchmarks. However, one of the primar
Externí odkaz:
http://arxiv.org/abs/2406.07844
Autor:
Basu, Samyadeep, Grayson, Martin, Morrison, Cecily, Nushi, Besmira, Feizi, Soheil, Massiceti, Daniela
Understanding the mechanisms of information storage and transfer in Transformer-based models is important for driving model understanding progress. Recent work has studied these mechanisms for Large Language Models (LLMs), revealing insights on how i
Externí odkaz:
http://arxiv.org/abs/2406.04236
Identifying the origin of data is crucial for data provenance, with applications including data ownership protection, media forensics, and detecting AI-generated content. A standard approach involves embedding-based retrieval techniques that match qu
Externí odkaz:
http://arxiv.org/abs/2406.02836
Inference on large language models can be expensive in terms of the compute and memory costs involved, especially when long sequence lengths are used. In particular, the self-attention mechanism used in such models contributes significantly to these
Externí odkaz:
http://arxiv.org/abs/2406.02542
Recent works have explored how individual components of the CLIP-ViT model contribute to the final representation by leveraging the shared image-text representation space of CLIP. These components, such as attention heads and MLPs, have been shown to
Externí odkaz:
http://arxiv.org/abs/2406.01583
Autor:
Kalibhat, Neha, Kattakinda, Priyatham, Zarei, Arman, Seleznev, Nikita, Sharpe, Samuel, Kumar, Senthil, Feizi, Soheil
Vision transformers have established a precedent of patchifying images into uniformly-sized chunks before processing. We hypothesize that this design choice may limit models in learning comprehensive and compositional representations from visual data
Externí odkaz:
http://arxiv.org/abs/2405.16401
Autor:
Christodorescu, Mihai, Craven, Ryan, Feizi, Soheil, Gong, Neil, Hoffmann, Mia, Jha, Somesh, Jiang, Zhengyuan, Kamarposhti, Mehrdad Saberi, Mitchell, John, Newman, Jessica, Probasco, Emelia, Qi, Yanjun, Shams, Khawaja, Turek, Matthew
The rise of Generative AI (GenAI) brings about transformative potential across sectors, but its dual-use nature also amplifies risks. Governments globally are grappling with the challenge of regulating GenAI, balancing innovation against safety. Chin
Externí odkaz:
http://arxiv.org/abs/2407.12999
Autor:
Meymandi, Arash Rasti, Hosseini, Zahra, Davari, Sina, Moshiri, Abolfazl, Rahimi-Golkhandan, Shabnam, Namdar, Khashayar, Feizi, Nikta, Tavakoli-Targhi, Mohamad, Khalvati, Farzad
This study explores the integration of advanced Natural Language Processing (NLP) and Artificial Intelligence (AI) techniques to analyze and interpret Persian literature, focusing on the poetry of Forough Farrokhzad. Utilizing computational methods,
Externí odkaz:
http://arxiv.org/abs/2405.06760