Statistical Multicriteria Evaluation of LLM-Generated Text
This work addresses the problem of nuanced text quality assessment for NLP researchers and practitioners, but it is incremental as it adapts an existing statistical framework to a new application.
The paper tackled the challenge of evaluating LLM-generated text by adapting a Generalized Stochastic Dominance framework to address limitations in existing methods, such as single-metric evaluation and lack of statistical guarantees, and demonstrated its ability to identify statistically significant performance differences across multiple quality dimensions.
Assessing the quality of LLM-generated text remains a fundamental challenge in natural language processing. Current evaluation approaches often rely on isolated metrics or simplistic aggregations that fail to capture the nuanced trade-offs between coherence, diversity, fluency, and other relevant indicators of text quality. In this work, we adapt a recently proposed framework for statistical inference based on Generalized Stochastic Dominance (GSD) that addresses three critical limitations in existing benchmarking methodologies: the inadequacy of single-metric evaluation, the incompatibility between cardinal automatic metrics and ordinal human judgments, and the lack of inferential statistical guarantees. The GSD-front approach enables simultaneous evaluation across multiple quality dimensions while respecting their different measurement scales, building upon partial orders of decoding strategies, thus avoiding arbitrary weighting of the involved metrics. By applying this framework to evaluate common decoding strategies against human-generated text, we demonstrate its ability to identify statistically significant performance differences while accounting for potential deviations from the i.i.d. assumption of the sampling design.