reading_time_model.json.gz is a gradient-boosted tree model fitted to word
reading times from the OneStop Eye Movements dataset, ordinary reading
regime (Berzak et al., OneStop: A 360-Participant English Eye Tracking
Dataset with Different Reading Regimes, Scientific Data 12, 1995 (2025);
https://github.com/lacclab/OneStop-Eye-Movements; data at
https://osf.io/2prdq/). The eye-tracking data is licensed under Creative
Commons Attribution 4.0 International. The model contains no text.

The model was also tested on the Provo Corpus (Luke and Christianson 2018,
Behavior Research Methods 50, 826-833; https://osf.io/sjefs/), licensed
under Creative Commons Attribution 4.0 International. Provo data was not
used for fitting.

word_freq.txt.gz holds word counts from two shards of FineWeb-Edu
(CC-MAIN-2019-04); see FINEWEB_NOTICE for its licence (ODC-By 1.0).

Built with scripts/reading_time_prep.py, scripts/reading_time_train.py and
scripts/build_word_freq.py. The feature set follows Cekini et al. 2026,
"The Cognitive Cost of LLM Writing" (arXiv 2610.23715).
