$\rm{C {\small IS}}^2$: A Simplified Commonsense Inference Evaluation for Story Prose
This work addresses a methodological issue in evaluating commonsense reasoning for NLP researchers, highlighting the need to disentangle tasks to avoid overestimating model capabilities.
The paper tackles the problem of conflating language generation ability with commonsense reasoning in transformer evaluations by introducing a simplified task, C IS^2, which eliminates generation and focuses on sentence selection, revealing that current models' performance may be inflated by their text generation skills.
Transformers have been showing near-human performance on a variety of tasks, but they are not without their limitations. We discuss the issue of conflating results of transformers that are instructed to do multiple tasks simultaneously. In particular, we focus on the domain of commonsense reasoning within story prose, which we call contextual commonsense inference (CCI). We look at the GLUCOSE (Mostafazadeh et al. 2020) dataset and task for predicting implicit commonsense inferences between story sentences. Since the GLUCOSE task simultaneously generates sentences and predicts the CCI relation, there is a conflation in the results. Is the model really measuring CCI or is its ability to generate grammatical text carrying the results? In this paper, we introduce the task contextual commonsense inference in sentence selection ($\rm{C {\small IS}}^2$), a simplified task that avoids conflation by eliminating language generation altogether. Our findings emphasize the necessity of future work to disentangle language generation from the desired NLP tasks at hand.