Salvato in:
Dettagli Bibliografici
Autori principali: Mirza, M. Jehanzeb, Zhao, Mengjie, Mao, Zhuoyuan, Doveh, Sivan, Lin, Wei, Gavrikov, Paul, Dorkenwald, Michael, Yang, Shiqi, Jha, Saurav, Wakaki, Hiromi, Mitsufuji, Yuki, Possegger, Horst, Feris, Rogerio, Karlinsky, Leonid, Glass, James
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:https://arxiv.org/abs/2410.06154
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866912545879621632
author Mirza, M. Jehanzeb
Zhao, Mengjie
Mao, Zhuoyuan
Doveh, Sivan
Lin, Wei
Gavrikov, Paul
Dorkenwald, Michael
Yang, Shiqi
Jha, Saurav
Wakaki, Hiromi
Mitsufuji, Yuki
Possegger, Horst
Feris, Rogerio
Karlinsky, Leonid
Glass, James
author_facet Mirza, M. Jehanzeb
Zhao, Mengjie
Mao, Zhuoyuan
Doveh, Sivan
Lin, Wei
Gavrikov, Paul
Dorkenwald, Michael
Yang, Shiqi
Jha, Saurav
Wakaki, Hiromi
Mitsufuji, Yuki
Possegger, Horst
Feris, Rogerio
Karlinsky, Leonid
Glass, James
contents In this work, we propose GLOV, which enables Large Language Models (LLMs) to act as implicit optimizers for Vision-Language Models (VLMs) to enhance downstream vision tasks. GLOV prompts an LLM with the downstream task description, querying it for suitable VLM prompts (e.g., for zero-shot classification with CLIP). These prompts are ranked according to their fitness for the downstream vision task. In each respective optimization step, the ranked prompts are fed as in-context examples (with their accuracies) to equip the LLM with the knowledge of the type of prompts preferred by the downstream VLM. Furthermore, we explicitly guide the LLM's generation at each optimization step by adding an offset vector -- calculated from the embedding differences between previous positive and negative solutions -- to the intermediate layer of the network for the next generation. This offset vector biases the LLM generation toward the type of language the downstream VLM prefers, resulting in enhanced performance on the downstream vision tasks. We comprehensively evaluate our GLOV on two tasks: object recognition and the critical task of enhancing VLM safety. Our GLOV shows performance improvement by up to 15.0% and 57.5% for dual-encoder (e.g., CLIP) and encoder-decoder (e.g., LlaVA) models for object recognition and reduces the attack success rate (ASR) on state-of-the-art VLMs by up to $60.7\%$.
format Preprint
id arxiv_https___arxiv_org_abs_2410_06154
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle GLOV: Guided Large Language Models as Implicit Optimizers for Vision Language Models
Mirza, M. Jehanzeb
Zhao, Mengjie
Mao, Zhuoyuan
Doveh, Sivan
Lin, Wei
Gavrikov, Paul
Dorkenwald, Michael
Yang, Shiqi
Jha, Saurav
Wakaki, Hiromi
Mitsufuji, Yuki
Possegger, Horst
Feris, Rogerio
Karlinsky, Leonid
Glass, James
Computer Vision and Pattern Recognition
In this work, we propose GLOV, which enables Large Language Models (LLMs) to act as implicit optimizers for Vision-Language Models (VLMs) to enhance downstream vision tasks. GLOV prompts an LLM with the downstream task description, querying it for suitable VLM prompts (e.g., for zero-shot classification with CLIP). These prompts are ranked according to their fitness for the downstream vision task. In each respective optimization step, the ranked prompts are fed as in-context examples (with their accuracies) to equip the LLM with the knowledge of the type of prompts preferred by the downstream VLM. Furthermore, we explicitly guide the LLM's generation at each optimization step by adding an offset vector -- calculated from the embedding differences between previous positive and negative solutions -- to the intermediate layer of the network for the next generation. This offset vector biases the LLM generation toward the type of language the downstream VLM prefers, resulting in enhanced performance on the downstream vision tasks. We comprehensively evaluate our GLOV on two tasks: object recognition and the critical task of enhancing VLM safety. Our GLOV shows performance improvement by up to 15.0% and 57.5% for dual-encoder (e.g., CLIP) and encoder-decoder (e.g., LlaVA) models for object recognition and reduces the attack success rate (ASR) on state-of-the-art VLMs by up to $60.7\%$.
title GLOV: Guided Large Language Models as Implicit Optimizers for Vision Language Models
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2410.06154