JFTA-Bench: Evaluate LLM's Ability of Tracking and Analyzing Malfunctions Using Fault Trees
This work addresses the need for LLMs to assist in maintenance of complex systems by tracking and analyzing malfunctions using fault trees, representing an incremental advancement in domain-specific benchmarking.
The authors tackled the problem of enabling large language models to process fault tree images for malfunction analysis by proposing a novel textual representation and constructing a benchmark with 3,130 entries and 40.75 average turns per entry to evaluate models' abilities in malfunction localization and error recovery, with Gemini 2.5 Pro achieving the best performance.
In the maintenance of complex systems, fault trees are used to locate problems and provide targeted solutions. To enable fault trees stored as images to be directly processed by large language models, which can assist in tracking and analyzing malfunctions, we propose a novel textual representation of fault trees. Building on it, we construct a benchmark for multi-turn dialogue systems that emphasizes robust interaction in complex environments, evaluating a model's ability to assist in malfunction localization, which contains $3130$ entries and $40.75$ turns per entry on average. We train an end-to-end model to generate vague information to reflect user behavior and introduce long-range rollback and recovery procedures to simulate user error scenarios, enabling assessment of a model's integrated capabilities in task tracking and error recovery, and Gemini 2.5 pro archives the best performance.