Cem Anil, Esin Durmus, Mrinank Sharma, Joe Benton, Sandipan Kundu, Joshua Batson, Nina Rimsky, Meg Tong, Jesse Mu, Daniel Ford, Francesco Mosconi, Rajashree Agrawal, Rylan Schaeffer, Naomi Bashkansky, Samuel Svenningsen, Mike Lambert, Ansh Radhakrishnan, Carson Denison, Evan J Hubinger, Yuntao Bai, Trenton Bricken, Timothy Maxwell, Nicholas Schiefer, Jamie Sully, Alex Tamkin, Tamera Lanham, Karina Nguyen, Tomasz Korbak, Jared Kaplan, Deep Ganguli, Samuel R. Bowman, Ethan Perez, Roger Grosse, David Duvenaud

Many-shot Jailbreaking

We investigate a family of simple long-context attacks on large language models: prompting with hundreds of demonstrations of undesirable behavior. This is newly feasible with the larger context windows recently deployed by Anthropic, OpenAI and Google DeepMind. We find that in diverse, realistic circumstances, the effectiveness of this attack follows a power law, up to hundreds of shots. We demonstrate the success of this attack on the most widely used state-of-the-art closed weight models, and across various tasks. Our results suggest very long contexts present a rich new attack surface for LLMs.

Read more here

Previous
Previous

Rapid Response: Mitigating LLM Jailbreaks with a Few Examples

Next
Next

Towards Understanding Sycophancy in Language Models