ReCOGS: How Incidental Details of a Logical Form Overshadow an Evaluation of Semantic Interpretation
This work addresses a methodological issue in evaluating compositional generalization for semantic parsing, which is incremental as it refines an existing benchmark.
The authors tackled the problem that the COGS benchmark for semantic parsing may misrepresent model capabilities due to incidental features of its logical forms, and they found that converting these forms to semantically equivalent ones allowed baseline models to perform better, leading to a modified benchmark called ReCOGS.
Compositional generalization benchmarks for semantic parsing seek to assess whether models can accurately compute meanings for novel sentences, but operationalize this in terms of logical form (LF) prediction. This raises the concern that semantically irrelevant details of the chosen LFs could shape model performance. We argue that this concern is realized for the COGS benchmark. COGS poses generalization splits that appear impossible for present-day models, which could be taken as an indictment of those models. However, we show that the negative results trace to incidental features of COGS LFs. Converting these LFs to semantically equivalent ones and factoring out capabilities unrelated to semantic interpretation, we find that even baseline models get traction. A recent variable-free translation of COGS LFs suggests similar conclusions, but we observe this format is not semantically equivalent; it is incapable of accurately representing some COGS meanings. These findings inform our proposal for ReCOGS, a modified version of COGS that comes closer to assessing the target semantic capabilities while remaining very challenging. Overall, our results reaffirm the importance of compositional generalization and careful benchmark task design.