{"data":{"slug":"ed-zh-parden","name":"PARDEN","tagline":"Defending against AI jailbreaks through repetition methods","github_url":"https://github.com/Ed-Zh/PARDEN","owner":"Ed-Zh","repo":"PARDEN","owner_avatar_url":"https://avatars.githubusercontent.com/u/74455933?v=4","primary_language":"HTML","stars":12,"forks":1,"topics":[],"archived":false,"github_pushed_at":"2024-05-14T08:14:06+00:00","maintenance_label":"Dormant","stars_delta_30d":0,"url":"https://www.graphcanon.com/tools/ed-zh-parden","markdown_url":"https://www.graphcanon.com/tools/ed-zh-parden.md","api_url":"https://www.graphcanon.com/api/graphcanon/tools/ed-zh-parden","graph_url":"https://www.graphcanon.com/api/graphcanon/graph?tool=ed-zh-parden","description":null,"homepage_url":null,"license":null,"open_issues":0,"watchers":1,"ai_summary":"Contains data and code for PARDEN, which studies defending aligned language models from adversarial attacks","readme_excerpt":"This repo contains the data and code (minimal demo) for our paper (accepted at ICML 2024): PARDEN, Can You Repeat That? Defending against Jailbreaks via Repetition\n \n\nPaper: https://arxiv.org/abs/2405.07932\n\nBlogpost:\n\nPARDEN_data/ contains the benign and harmful datasets genearted by different models. The generation of harmful dataset is explained in detail in the paper, using both GCG[1] and prompt injection. \n\n\nPARDEN_notebook_minimal.ipynb demonstrates how to use PARDEN and tests its performance on the harmful strings generated by llama2-7b. \n\n[1] Zou et al. Universal and Transferable Adversarial Attacks on Aligned Language Models. https://arxiv.org/abs/2307.15043","github_created_at":"2024-01-30T10:37:07+00:00","created_at":"2026-07-11T23:41:52.290291+00:00","updated_at":"2026-10-06T06:00:07.505918+00:00","categories":[{"slug":"evaluation-observability","name":"Evaluation & Observability","url":"https://www.graphcanon.com/categories/evaluation-observability","markdown_url":"https://www.graphcanon.com/categories/evaluation-observability.md","api_url":"https://www.graphcanon.com/api/graphcanon/categories/evaluation-observability"}],"tags":[{"slug":"adversarial-attacks","name":"adversarial-attacks"},{"slug":"jailbreak-defense","name":"jailbreak defense"},{"slug":"language-models","name":"language models"}],"trust":{"provenance":{"is_fork":false,"github_id":750278286,"owner_type":"User","methodology":"github_public_v1","parent_repo":null,"near_duplicate_slugs":[]},"computed_at":"2026-10-06T06:00:06.412Z","maintenance":{"label":"Dormant","score":18,"methodology":"github_public_v1","releases_90d":0,"days_since_push":874,"last_release_at":null,"stars_delta_30d":0,"open_issues_delta_30d":0},"security_summary":{"status":"no_lockfile","scanner":null,"low_count":0,"high_count":0,"last_scan_at":"2026-07-11T23:41:54.727Z","medium_count":0,"scan_profile":"none","critical_count":0}},"capability_facts":{"scan":{"source":"repo_scan","observed_at":"2026-10-06T06:00:07.219Z"},"languages":{"value":["html"],"source":"github.language","observed_at":"2026-10-06T06:00:07.219Z"}},"decision_facts":{"hosting":null,"pricing":null,"requirements":null,"constraints":null,"when_to_use":["When studying defenses against AI jailbreaks","For evaluation on specific harmful strings generated by LLama2-7B"],"when_not_to_use":["If focusing on real-time defense implementation outside research settings","For general model alignment without concern for adversarial attacks"],"source":"enrich:decision_facts","observed_at":"2026-07-15T10:10:14.941Z"},"constraint_facets":null,"decision_summary":[{"label":"Adopt for","value":"PARDEN defends AI models from adversarial attacks using repetition methods."},{"label":"License detail","value":"(unknown)"}]}}