search

thu-ml / MMTrustEval

A toolbox for benchmarking trustworthiness of multimodal large language models (MultiTrust, NeurIPS 2024 Track Datasets and Benchmarks)
https://multi-trust.github.io/
Creative Commons Attribution Share Alike 4.0 International
108 stars 7 forks source link

Code for generating Adversarial examples (untargeted/targeted) #6

Open HashmatShadab opened 1 month ago

HashmatShadab commented 1 month ago

Hi! Thanks for sharing your work

Can you please provide the script for generating adversarial examples for Task R.4 and R.5 ?

zycheiheihei commented 1 month ago

Sure, the corresponding code for online generation has been provided in https://github.com/thu-ml/MMTrustEval/blob/main/mmte/methods/adv.py. We will soon release an example for how to integrate it into the evaluation pipeline.