Typed decision models struggle to reliably use fallback options
Lost in the Fallback: How Reliably Do Typed Decision Models Use Fallback Options?
Cryptography and SecurityArtificial Intelligence
Summary
Typed decision models help systems choose between options, and fallback options let them ask for human help or better models when unsure. This paper shows these models often fail to use fallback options correctly, even when the rules are clearly defined. Slight changes in how fallback rules are written can cause the models to behave inconsistently. The authors reveal that fallback mechanisms meant to make decisions safer can sometimes make more mistakes instead.
What this means in practice
- •For system architects: Improve reliability in routing or model selection by verifying fallback rules for consistent execution across different formulations.
- •For customer support teams: Ensure automated systems correctly defer complex cases to humans when fallback options are triggered to maintain service quality.
Authors
Yuanbo Xie, Mingli Zhu, Tianyun Liu, Yingjie Zhang, Haixiao Xiong, Tingwen Liu
Abstract
Typed decision models (TDMs) are increasingly used in applications such as request routing and model selection, making reliable decisions critical to system performance. Fallback options accommodate categories outside regular choices, enabling deferral to human experts or more capable models. While recent works have examined adversarial vulnerabilities in TDMs, their ability to follow fallback rules in non-adversarial settings remains underexplored. We introduce the Paired Fallback Robustness (PFR) Test to evaluate two aspects of fallback reliability: whether models correctly switch between regular and fallback options, and whether they make consistent decisions under logically equivalent definitions of the same rule. We evaluate open-source decision models on three benchmarks, with controlled comparisons across architectures and parameter counts. We find that correct recognition does not reliably transfer to correct fallback options, and errors persist even when the correct category is explicitly supplied.Moreover, logically equivalent formulations of the same fallback rule can also produce substantial disagreement in model decisions. These findings reveal a crucial reliability gap in current TDMs: fallback mechanisms designed to improve reliability can instead introduce errors. It highlights faithful execution of fallback rules and consistency under equivalent definitions as essential requirements for reliable typed decision-making.