English

How much do language models copy from their training data? Evaluating linguistic novelty in text generation using RAVEN

Computation and Language 2021-11-19 v1

Abstract

Current language models can generate high-quality text. Are they simply copying text they have seen before, or have they learned generalizable linguistic abstractions? To tease apart these possibilities, we introduce RAVEN, a suite of analyses for assessing the novelty of generated text, focusing on sequential structure (n-grams) and syntactic structure. We apply these analyses to four neural language models (an LSTM, a Transformer, Transformer-XL, and GPT-2). For local structure - e.g., individual dependencies - model-generated text is substantially less novel than our baseline of human-generated text from each model's test set. For larger-scale structure - e.g., overall sentence structure - model-generated text is as novel or even more novel than the human-generated baseline, but models still sometimes copy substantially, in some cases duplicating passages over 1,000 words long from the training set. We also perform extensive manual analysis showing that GPT-2's novel text is usually well-formed morphologically and syntactically but has reasonably frequent semantic issues (e.g., being self-contradictory).

Keywords

Cite

@article{arxiv.2111.09509,
  title  = {How much do language models copy from their training data? Evaluating linguistic novelty in text generation using RAVEN},
  author = {R. Thomas McCoy and Paul Smolensky and Tal Linzen and Jianfeng Gao and Asli Celikyilmaz},
  journal= {arXiv preprint arXiv:2111.09509},
  year   = {2021}
}

Comments

10 pages, plus 39 pages of appendices

R2 v1 2026-06-24T07:43:02.639Z