Transparency at the Source: Evaluating and Interpreting Language Models With Access to the True Distribution
Submission Type: Regular Long Paper
Submission Track: Interpretability, Interactivity, and Analysis of Models for NLP
Submission Track 2: Language Modeling and Analysis of Language Models
Keywords: language models, perplexity, interpretability, PCFG, learning dynamics, synthetic data
TL;DR: Training Language Models on data generated from a transparent generative process allows for more controlled evaluation of their behaviour.
Abstract: We present a setup for training, evaluating and interpreting neural language models, that uses artificial, language-like data. The data is generated using a massive probabilistic grammar (based on state-split PCFGs), that is itself derived from a large natural language corpus, but also provides us complete control over the generative process. We describe and release both grammar and corpus, and test for the naturalness of our generated data. This approach allows us define closed-form expressions to efficiently compute exact lower bounds on obtainable perplexity using both causal and masked language modelling. Our results show striking differences between neural language modelling architectures and training objectives in how closely they allow approximating the lower bound on perplexity. Our approach also allows us to directly compare learned representations to symbolic rules in the underlying source. We experiment with various techniques for interpreting model behaviour and learning dynamics. With access to the underlying true source, our results show striking differences and outcomes in learning dynamics between different classes of words.
Submission Number: 5707
Loading