Issues

• adversarial_training_mnist.ipynb fails with SymbolicTensor error when eager execution is disabled

  notebooks
  Trusted-AI/adversarial-robustness-toolbox#2845 · Filter by author mathieuemery opened on Oct 7, 2026
  1

  

• empirical_robustness returns 0.0 when the attack flips no predictions, scoring a fully resisted attack below any successful one

  Trusted-AI/adversarial-robustness-toolbox#2828 · Filter by author shaurya416 opened on Sep 26, 2026

• Optional EvalPort adapter for per-sample robustness results (adversarial_accuracy/empirical_robustness)

  Trusted-AI/adversarial-robustness-toolbox#2827 · Filter by author adhabnr-ux opened on Sep 22, 2026

• ScikitlearnLogisticRegression.loss_gradient and ScikitlearnSVC.loss_gradient fail on index labels despite docstring

  Trusted-AI/adversarial-robustness-toolbox#2825 · Filter by author Francis206 opened on Sep 14, 2026

• sklearn .fit method incorrectly applies arg max

  Trusted-AI/adversarial-robustness-toolbox#2808 · Filter by author simplymathematics opened on Apr 10, 2026 · ART 1.21.0
  2

  

• LLM robustness reference: WFGY 1.0 self-healing framework + ProblemMap

  Trusted-AI/adversarial-robustness-toolbox#2806 · Filter by author onestardao opened on Feb 10, 2026
  1

• BUG web ART

  documentation
  Trusted-AI/adversarial-robustness-toolbox#2330 · Filter by author OrsonTyphanel93 opened on Nov 19, 2023
  1

  

• Check Boundary Attack initialised with target class

  enhancement
  Trusted-AI/adversarial-robustness-toolbox#81 · Filter by author beat-buesser opened on May 29, 2019