SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2406.01288 · NeurIPS · 2024

Improved Few-Shot Jailbreaking Can Circumvent Aligned Language Models and Their Defenses

Tianyu Pang, Jing Jiang, Qian Liu, Min Lin, Chao Du, Xiaosen Zheng

arXiv · PDF · Open in the Atlas

Code that ran

We lifted 11 functions out of this paper's own repositories and ran 2 of them in a sandbox. "Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results.

RepositoryRoleRan
sail-sg/I-FSJ canonical 2 of 11
FunctionStatusWhere it lives
check_for_attack_success Ran sail-sg/I-FSJ/exps/rs.py
code served (permissive licence) · get_code("a2058f7fc99001a7")
generate Ran sail-sg/I-FSJ/exps/rs.py
code served (permissive licence) · get_code("3a80a2bfa48e30bb")
combine_prompts Not yet run sail-sg/I-FSJ/api_experiments/evaluate_api_models.py
code served (permissive licence) · get_code("71b177b3a3ed6078")
get_embedding_layer Not yet run sail-sg/I-FSJ/llm_attacks/base/attack_manager.py
code served (permissive licence) · get_code("0bf2136a7171ae80")
get_embedding_matrix Not yet run sail-sg/I-FSJ/llm_attacks/base/attack_manager.py
code served (permissive licence) · get_code("688aaeba506e3329")
get_embeddings Not yet run sail-sg/I-FSJ/llm_attacks/base/attack_manager.py
code served (permissive licence) · get_code("f556dda0785f3c77")
get_filtered_cands Not yet run sail-sg/I-FSJ/llm_attacks/minimal_gcg/opt_utils.py
code served (permissive licence) · get_code("25ae440509f39ec5")
load_prompts Not yet run sail-sg/I-FSJ/api_experiments/evaluate_api_models.py
code served (permissive licence) · get_code("645890ba00019a69")
make_checking_fn Not yet run sail-sg/I-FSJ/api_experiments/evaluate_api_models.py
code served (permissive licence) · get_code("d0e94f216f27f804")
sample_control Not yet run sail-sg/I-FSJ/exps/rs.py
code served (permissive licence) · get_code("1e8c1c46e0ec7dad")
sample_control Not yet run sail-sg/I-FSJ/llm_attacks/minimal_gcg/opt_utils.py
code served (permissive licence) · get_code("94d7ddaf0169dbf1")

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Recently, Anil et al. [3] show that many-shot (up to hundreds of) demonstrations can jailbreak state-of-the-art LLMs by exploiting their long-context capability. Nevertheless, is it possible to use few-shot demonstrations to efficiently jailbreak LLMs within limited context sizes? While the vanilla few-shot jailbreaking may be inefficient, we propose improved techniques such as injecting special system tokens like [/INST] and employing demo-level random search from a collected demo pool. These simple techniques result in surprisingly effective jailbreaking against aligned LLMs (even with advanced defenses). For example, our method achieves > 80% (mostly > 95%) ASRs on Llama-2-7B and Llama-3-8B without multiple restarts, even if the models are enhanced by strong defenses such as perplexity detection and/or SmoothLLM, which is challenging for suffix-based jailbreaking. In addition, we conduct comprehensive and elaborate (e.g., making sure to use correct system prompts) evaluations against other aligned LLMs and advanced defenses, where our method consistently achieves nearly 100% ASRs. Our code is available at https://github.com/sail-sg/I-FSJ.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2406.01288")
get_code_for_paper("2406.01288")
have("2406.01288")

Connect an agent — have() is free.