Carrie Wu

Jailbreaking large-language models (LLMs) involves testing their robustness against adversarial prompts and evaluating their ability to withstand prompt attacks that could elicit unauthorized or malicious responses. In this paper, we present TurboFuzzLLM, a mutation-based fuzzing technique for efficiently finding a collection of effective jailbreaking templates that, when combined with harmful questions, can lead a target LLM to produce harmful responses through black-box access via user prompts.

Carrie Wu · Follow 2 min read · -- Where to find mentors: external opportunities- networking online/offline more: join a professional learning group Who is the right mentor for me? Mentor criteria checklist: type skillset- specific industry knowledge- different priority availability communication style challenge resolution skills For example, my checklist: leadership influence without authority, write tech doc, present in conference, product/case/proposal building/vision writing.

Carrie Wu · Follow Just now -- Goal: Know the key trends that highlight the economic impact of women as consumers, investors, and leaders. Understand how the low numbers of women working in technology industries stifles innovation. Learn how to create an inclusive team or organisation where women want to work and thrive. SMART Goals: Elevate the visibility of our technical women talent by increasing the underrepresented in tech senior leaders.