Risk-Aware Adaptive Evaluation: Finding High-Impact Failures Under Limited Budgets

arXiv:2609.38914v1 Announce Type: new 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 au-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 3.5imes more impact-weighted failures (215.4 vs. 62.2) with the same number of trials, delivers 5imes 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.

Related

• AURA-Eval: Evaluation Framework for Acting Under Risk Awareness in LLM Agent Trajectories
• Exploit More, Explore Smarter for Budget-Constrained Agentic Search
• Risk Under Pressure: Compute-Aware Evaluation of Adversarial Robustness in Language Models
• One Policy, Any Budget: Internalizing Budget-Aware Search via Reinforcement Learning

Source: arXiv cs.AI | 2026-10-01