Abstract
We examine three evaluation paradigms: standard benchmarks (e.g., MMLU and BBH), interactive games (e.g., Signalling Games or Taboo), and cognitive tests (e.g., for working memory or theory of mind). First, we investigate which of the former two—benchmarks or games—is most effective at discriminating LLMs of varying quality. Then, inspired by human cognitive assessments, we compile a suite of targeted tests that measure cognitive abilities deemed essential for effective language use, and we investigate their correlation with model performance in benchmarks and games. Our analyses reveal that interactive games are superior to standard benchmarks in discriminating models. Causal and logical reasoning correlate with both static and interactive tests, while differences emerge regarding core executive functions and social/emotional skills, which correlate more with games. We propose a new evaluation framework triangulating LLMs progress. Our findings highlight the importance of defining evaluation regimes that consider multiple paradigms. We advocate for the development of new interactive benchmarks and targeted cognitive tasks inspired by human ability assessments but designed specifically for LLMs. The code for running the experiments is released at: https://github.com/momentino/playpen_eval/tree/triangulating.