Runtime auditor improves failure prediction in multi-agent systems

Dream2Discern: Learning from Hard Latent Variants for Runtime Failure Prediction in Multi-Agent Systems

Machine LearningArtificial IntelligenceComputation and LanguageMultiagent Systems

Summary

Predicting when complex groups of AI agents will fail during their tasks is important but difficult because verifying failures is costly and limited by available data. The authors propose a new training method called Dream2Discern, where the system generates challenging new training examples from existing data by tweaking its internal reasoning steps to increase prediction errors. This creates a renewable set of hard cases that help the system learn better failure prediction without collecting more data. Tests show this method improves accuracy and speed in predicting failures across benchmark datasets.

What this means in practice

  • For software reliability engineers: Improve early detection of failures in multi-agent AI systems using a renewable source of challenging examples without extra data collection.
  • For autonomous system operators: Enable faster and more accurate runtime failure predictions to intervene before terminal errors in complex agent-based systems.

Authors

Boxuan Zhang, jianing Zhu, Xiaotian Han, Ruixiang Tang

Abstract

Reliable deployment of LLM-based multi-agent systems calls for runtime auditors that predict terminal outcomes from observed execution prefixes. These auditors typically learn from a limited corpus of trajectories with verified outcomes. Expanding this corpus requires costly execution and verification, yet broader failure coverage alone provides no explicit criterion for identifying prefixes that the current auditor finds difficult to classify. To this end, we propose Dream2Discern, a runtime auditor trained through implicit augmentation guided by its own prediction loss. During Dream, we select perturbations of the auditor's representation operators to increase prediction loss on recorded prefixes, yielding hard latent variants. During Discern, the auditor learns to predict execution outcomes from these variants using the original labels. The variants are regenerated as the auditor learns, turning a limited corpus into a renewable source of hard examples and requiring no additional trajectory collection or outcome verification. Extensive evaluations on AFTraj-2K and StepShield show that Dream2Discern outperforms competitive failure auditors and achieves over 90% trajectory-level weighted accuracy on both benchmarks while balancing prediction accuracy and timeliness. Project page: https://zbox1005.github.io/dream2discern-project/