Tag
This paper introduces Trajel, a dataset and evaluation framework for auditing trajectory-level hallucinations in multi-agent industrial workflows, proposing a five-type hallucination taxonomy and showing that trajectory-aware detection outperforms standard post-hoc verification.
The paper introduces LEAPBench, a 55-task framework for trajectory-level evaluation of LLMs in iterative scientific design, revealing that outcome-based scoring misses efficiency gains and that domain-agnostic prompting can outperform domain-aware prompting in matching published best designs.