Summary of Two Heads Are Better Than One: Averaging Along Fine-tuning to Improve Targeted Transferability, by Hui Zeng et al.

Two Heads Are Better Than One: Averaging along Fine-Tuning to Improve Targeted Transferability

by Hui Zeng, Sanshuai Cui, Biwei Chen, Anjie Peng

First submitted to arxiv on: 30 Dec 2024

GrooveSquid.com Paper Summaries

GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!

Summary difficulty	Written by	Summary
High	Paper authors	High Difficulty Summary Read the original abstract here
Medium	GrooveSquid.com (original content)	Medium Difficulty Summary The paper proposes a new approach to fine-tuning adversarial examples (AE) in feature space, aiming to boost their targeted transferability. The existing fine-tuning schemes only utilize the endpoint and ignore valuable information along the trajectory. This study introduces averaging over the fine-tuning trajectory to pull crafted AEs towards a more centered region of the loss surface. By integrating this method with state-of-the-art targeted attacks in various scenarios, experimental results demonstrate its superiority in enhancing transferability.
Low	GrooveSquid.com (original content)	Low Difficulty Summary The paper improves how to make fake examples work better against different types of targets. It suggests averaging the changes made during fine-tuning to move the example towards a more central spot on the graph. This helps improve how well the example works when used as an attack against different targets. The study compares this new approach with current methods and shows that it performs better in various scenarios.