Adversarial Benchmark

Adversarial benchmarks are designed to rigorously evaluate the robustness of machine learning models by exposing them to carefully crafted inputs intended to induce errors, mimicking real-world challenges like noisy data or malicious attacks. Current research focuses on developing more effective benchmark datasets across various modalities (text, images, sensor data) and metrics to quantify adversarial strength, often employing techniques like reinforcement learning and information bottleneck principles to generate and analyze these challenging examples. These benchmarks are crucial for improving model reliability and safety, particularly in high-stakes applications like autonomous driving and fake news detection, by identifying and mitigating vulnerabilities before deployment.

Papers

October 12, 2024

Are You Human? An Adversarial Benchmark to Expose LLMs
Gilad Gressel, Rahul Pankajakshan, Yisroel Mirsky
Technical Challenge Human Generated LLM Detector LLM Truthfulness Adversarial Benchmark

June 24, 2024

Is your benchmark truly adversarial? AdvScore: Evaluating Human-Grounded Adversarialness
Yoo Yeon Sung, Maharshi Gor, Eve Fleisig, Ishani Mondal, Jordan Lee Boyd-Graber
Adversarial Example Global Evaluation Visual Creation North Star Metric Adversarial Datasets Adversarial Question Adversarial Benchmark

November 28, 2023

RETSim: Resilient and Efficient Text Similarity
Marina Zhang, Owen Vallis, Aysegul Bumin, Tanay Vakharia, Elie Bursztein
Cross Lingual Tiny Refinement Elicit Resilience Adversarial Benchmark

October 28, 2023

Benchmark Generation Framework with Customizable Distortions for Image Classifier Robustness
Soumyendu Sarkar, Ashwin Ramesh Babu, Sajad Mousavi, Zachariah Carmichael, Vineet Gundecha, Sahand Ghorbanpour, Ricardo Luna, Gutierrez Antonio Guillen, Avisek Naug
Supervised ImageNet Adversarial Sample Distortion Aware Image Classifier Image Classification Model Adversarial Benchmark

October 9, 2023

SC-Safety: A Multi-round Open-ended Question Adversarial Safety Benchmark for Large Language Models in Chinese
Liang Xu, Kangkang Zhao, Lei Zhu, Hang Xue
Large Language Model Language Understanding Chinese Character Human SAFETY Adversarial Behavior Multi Round Adversarial Benchmark

August 17, 2023

Sensor Fusion by Spatial Encoding for Autonomous Driving
Quoc-Vinh Lai-Dang, Jihui Lee, Bumgeun Park, Dongsoo Har
Autonomous Driving Transformer Based Mid Range LiDAR Sensor Fusion Spatial Encoding Adversarial Benchmark

February 9, 2023

IB-RAR: Information Bottleneck as Regularizer for Adversarial Robustness
Xiaoyun Xu, Guilherme Perin, Stjepan Picek
Adversarial Attack Adversarial Example Adversarial Training Adversarial Robustness Information Bottleneck Gradient Regularization Adversarial Benchmark

February 5, 2022

Layer-wise Regularized Adversarial Training using Layers Sustainability Analysis (LSA) framework
Mohammad Khalooei, Mohammad Mehdi Homayounpour, Maryam Amirmazlaghani
Adversarial Attack New Framework Adversarial Training Layer Wise Adversarial Benchmark Sustainability Assessment

January 3, 2022

An Adversarial Benchmark for Fake News Detection Models
Lorenzo Jaime Yu Flores, Yiding Hao
Adversarial Attack Fake News Fake News Detection Online Misinformation Fake News Detection Model Adversarial Benchmark