Adaptive evaluation finds more impactful failures with fewer tests
Risk-Aware Adaptive Evaluation: Finding High-Impact Failures Under Limited Budgets
Artificial IntelligenceMachine LearningMultiagent Systems
Summary
Testing interactive agents to find failures can be very expensive, especially because failures are rare and differ in importance. The authors propose a smarter way to decide which test scenarios to repeat, using a technique that considers the risk and expected impact of failures. Their method finds many more significant failures using far fewer tests compared to testing all scenarios equally. This approach works best when testing resources are very limited.
What this means in practice
- •For software testing teams: Allocate limited testing resources to find high-impact bugs more efficiently in complex systems with stochastic behaviors.
- •For airline operations analysts: Prioritize evaluation of flight scenarios to identify costly failure modes with fewer trials, improving safety assessments under budget constraints.
Tested on one dataset.
Authors
Priyanath Maji, Spandan Ghose Chowdhury
Abstract
Evaluating interactive agents is expensive. Agent behavior is stochastic, so reliability must be measured over repeated trials, but failures are rare and differ widely in how much they matter. Standard benchmarks spend this budget uniformly: a read-only lookup is sampled as often as an irreversible payment action. We instead formulate evaluation as a sequential allocation problem. Given a fixed trial budget and a set of scenarios whose failure behavior is unknown, which scenarios should be run, and run again? We propose a risk-aware contextual Thompson Sampling policy that combines a pre-execution scenario context vector and a fixed impact score with the failure outcomes observed during evaluation, and we test it by offline replay over 70 $τ$-bench airline scenarios and 824 recorded trials. Our main result is at the smallest budget: with only 50 trials ($6\%$ of the corpus), the policy recovers $86\%$ of the impact-weighted failures an oracle could find, compared to $25\%$ for uniform allocation. It discovers $3.5\times$ more impact-weighted failures (215.4 vs. 62.2) with the same number of trials, delivers $5\times$ the discovery per dollar, and cuts the budget wasted on scenarios that never fail from $34\%$ to $2.8\%$. The rest of our analysis demonstrates and qualifies this result: a budget sweep shows the advantage shrinks as the budget approaches the corpus size, and paired significance tests show that scenario context helps mainly at small budgets while posterior-based exploration helps at moderate ones. Risk-aware adaptive allocation therefore helps most exactly where evaluation budget is scarcest.