Neural Models and Language Model Prompting for the Multidimensional Evaluation of Open-Ended Conversations
This work addresses the evaluation problem for dialogue systems, but it is incremental as it builds on existing methods and benchmarks within a specific challenge.
The paper tackled the challenge of evaluating generative AI-based dialogue systems by developing models to predict dimension-specific scores for open-ended conversations, achieving modest correlations with human judgments using LM prompting and high correlations on some dimensions with smaller encoder-based models, though performance decreased on a test set with different score ranges.
The growing number of generative AI-based dialogue systems has made their evaluation a crucial challenge. This paper presents our contribution to this important problem through the Dialogue System Technology Challenge (DSTC-12, Track 1), where we developed models to predict dialogue-level, dimension-specific scores. Given the constraint of using relatively small models (i.e. fewer than 13 billion parameters) our work follows two main strategies: employing Language Models (LMs) as evaluators through prompting, and training encoder-based classification and regression models. Our results show that while LM prompting achieves only modest correlations with human judgments, it still ranks second on the test set, outperformed only by the baseline. The regression and classification models, with significantly fewer parameters, demonstrate high correlation for some dimensions on the validation set. Although their performance decreases on the test set, it is important to note that the test set contains annotations with significantly different score ranges for some of the dimensions with respect to the train and validation sets.