PairAudit: Guiding Human Review with Graph Tokens under Distribution Shift

Key points

• We introduce PairAudit to find overlooked errors and improve review under a fixed budget.1
• Its graph tokens capture prediction patterns across connected nodes.1
• Rather than building another predictor through feature aggregation, PairAudit uses unusual relational patterns to uncover potential errors in existing predictions.1
• Experiments across security tasks show that PairAudit corrects more errors on average than uncertainty-based review, including more errors on unseen attacks.1

Sources (1)

• [1]PairAudit: Guiding Human Review with Graph Tokens under Distribution Shift
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 7, 03:35 PM

  Intrusion detectors can confidently misclassify attacks that were not seen during training.

  We introduce PairAudit to find overlooked errors and improve review under a fixed budget.

Extractive summary: sentences quoted from the sources.