{"data":{"slug":"unispac-visual-adversarial-examples-jailbreak-large-language-models","name":"Visual-Adversarial-Examples-Jailbreak-Large-Language-Models","tagline":"Repository for visual adversarial examples that jailbreak large language models","github_url":"https://github.com/Unispac/Visual-Adversarial-Examples-Jailbreak-Large-Language-Models","owner":"Unispac","repo":"Visual-Adversarial-Examples-Jailbreak-Large-Language-Models","owner_avatar_url":"https://avatars.githubusercontent.com/u/45935569?v=4","primary_language":"Python","stars":283,"forks":30,"topics":[],"archived":false,"github_pushed_at":"2024-05-13T05:36:24+00:00","maintenance_label":"Dormant","stars_delta_30d":1,"url":"https://www.graphcanon.com/tools/unispac-visual-adversarial-examples-jailbreak-large-language-models","markdown_url":"https://www.graphcanon.com/tools/unispac-visual-adversarial-examples-jailbreak-large-language-models.md","api_url":"https://www.graphcanon.com/api/graphcanon/tools/unispac-visual-adversarial-examples-jailbreak-large-language-models","graph_url":"https://www.graphcanon.com/api/graphcanon/graph?tool=unispac-visual-adversarial-examples-jailbreak-large-language-models","description":"Repository for the Paper (AAAI 2024, Oral) --- Visual Adversarial Examples Jailbreak Large Language Models","homepage_url":null,"license":null,"open_issues":24,"watchers":3,"ai_summary":"This repository contains work related to creating and analyzing visual adversarial examples aimed at disrupting the functioning of large language models, specifically showcasing attacks on MiniGPT-4.","readme_excerpt":"### Installation\n\nWe take MiniGPT-4 (13B) as the sandbox to showcase our attacks. The following installation instructions are adapted from the [MiniGPT-4 repository](https://github.com/Vision-CAIR/MiniGPT-4).\n\n**1. Set up the environment**\n\n```bash\ngit clone https://github.com/Unispac/Visual-Adversarial-Examples-Jailbreak-Large-Language-Models.git\n\ncd Visual-Adversarial-Examples-Jailbreak-Large-Language-Models\n\nconda env create -f environment.yml\nconda activate minigpt4\n```\n\n**2. Prepare the pretrained weights for MiniGPT-4**\n\n> As we directly inherit the MiniGPT-4 code base, the guide from the [MiniGPT-4 repository](https://github.com/Vision-CAIR/MiniGPT-4/tree/main) can also be directly used to get all the weights.\n\n* **Get Vicuna:** MiniGPT-4 (13B) is built on the v0 version of [Vicuna-13B](https://lmsys.org/blog/2023-03-30-vicuna/). Please refer to this [guide](https://github.com/Vision-CAIR/MiniGPT-4/blob/main/PrepareVicuna.md) from the MiniGPT-4 repository to get the weights of Vicuna.\n\n  Then, set the path to the vicuna weight in the model config file [here](https://github.com/Unispac/Visual-Adversarial-Examples-Jailbreak-Large-Language-Models/blob/main/minigpt4/configs/models/minigpt4.yaml#L16) at Line 16.\n\n* **Get MiniGPT-4 (the 13B version) checkpoint**: download from [here](https://drive.google.com/file/d/1a4zLvaiDBr-36pasffmgpvH5P7CKmpze/view?usp=share_link). \n\n  Then, set the path to the pretrained checkpoint in the evaluation config file in [eval_configs/minigpt4_eval.yaml](https://github.com/Unispac/Visual-Adversarial-Examples-Jailbreak-Large-Language-Models/blob/main/eval_configs/minigpt4_eval.yaml#L11) at Line 11.\n\n<br>","github_created_at":"2023-06-13T23:39:52+00:00","created_at":"2026-07-11T23:39:40.547223+00:00","updated_at":"2026-10-04T06:00:38.939992+00:00","categories":[{"slug":"computer-vision","name":"Computer Vision","url":"https://www.graphcanon.com/categories/computer-vision","markdown_url":"https://www.graphcanon.com/categories/computer-vision.md","api_url":"https://www.graphcanon.com/api/graphcanon/categories/computer-vision"},{"slug":"model-training","name":"Model Training","url":"https://www.graphcanon.com/categories/model-training","markdown_url":"https://www.graphcanon.com/categories/model-training.md","api_url":"https://www.graphcanon.com/api/graphcanon/categories/model-training"}],"tags":[{"slug":"large-language-model","name":"large-language-model"},{"slug":"visual-adversarial","name":"visual adversarial"}],"trust":{"provenance":{"is_fork":false,"github_id":653369312,"owner_type":"User","methodology":"github_public_v1","parent_repo":null,"near_duplicate_slugs":[]},"computed_at":"2026-10-04T06:00:38.158Z","maintenance":{"label":"Dormant","score":18,"methodology":"github_public_v1","releases_90d":0,"days_since_push":874,"last_release_at":null,"stars_delta_30d":1,"open_issues_delta_30d":0},"security_summary":{"status":"no_lockfile","scanner":null,"low_count":0,"high_count":0,"last_scan_at":"2026-07-11T23:39:42.118Z","medium_count":0,"scan_profile":"none","critical_count":0}},"capability_facts":{"scan":{"source":"repo_scan","observed_at":"2026-10-04T06:00:38.590Z"},"languages":{"value":["python"],"source":"github.language","observed_at":"2026-10-04T06:00:38.590Z"}},"decision_facts":{"hosting":null,"pricing":null,"requirements":null,"constraints":null,"when_to_use":["If your goal is to research the resilience of Large Language Models against visual inputs that could mislead or 'jailbreak' their usual behavior.","For scholars and researchers studying model security and robustness, particularly focusing on how image data can affect language models."],"when_not_to_use":["Avoid using this tool if your objective does not involve security testing or research concerning visual-linguistic interactions with large language models.","If you are looking to enhance general-purpose AI applications without considering adversarial attacks, other frameworks would be more appropriate."],"source":"enrich:decision_facts","observed_at":"2026-07-16T22:17:16.412Z"},"constraint_facets":null,"decision_summary":[{"label":"Adopt for","value":"This tool focuses on generating and studying visual adversarial examples designed to exploit vulnerabilities in large language models such as MiniGPT-4."}]}}