Known-author test models (benchmark fine-tunes)

The language models fine-tuned for the validation benchmark (110 held-out tests), kept so every benchmark number can be checked without retraining. runs/<size>_<variant>/weights/ holds one fine-tuned copy per benchmark author (V1, the primary method's procedure) or per task (V2, with its classifier head), and the folder's .json files their scores; results/alm_*.json are the benchmark results. Sizes: pilot 125M, base 297M, mid 590M. Runs: _v1_lrsel and _v2_lrsel at the learning rate chosen on held-out training text (Amendment 41); _v2_seed1..5 V2 with five further seeds (Amendment 42; V2's false-match rate ranged 9-98% across seeds). The registered runs' copies (primary, V1, V2 as first run) were scored and discarded at the time and were not rebuilt: no result depends on them, and they can be rebuilt from the base models, the code and the config. Their scores are in runs/ and results/.

The study and what it found

This is one of the models from a finished, open study of the English text of the Book of Mormon: who writes like it, what it borrows, and when its language dates from.

In short: the book copies almost nothing from earlier print. Among the eight people usually proposed as its author, the closest writing is Joseph Smith's revelations, and a search of 23,872 authors of the period puts those revelations first or second. No tool we found, ours included, can tell a person from the voice they are writing in, or reliably say "it is none of these people". The study takes no position on inspiration.

About

Part of an open, pre-registered study of the authorship, borrowing and dating of the earliest Book of Mormon text (https://github.com/lifehug/book-of-mormon-authorship; pre-registration https://osf.io/6ugxc/). The project does not argue for or against divine inspiration. Everything is published so it can be checked and redone; the corpus texts we agreed to keep private are not, and the repository's docs/replication.md gives the steps to obtain the same access.

Licence: CC BY-NC 4.0. Training data: English printed 1780-1830 (Internet Archive, Harvard Institutional Books 1.0, HathiTrust public-domain volumes, Text Creation Partnership), with the Book of Mormon and everything after 1830 removed; see the base model cards (etherfuse/timelocked-1830-125m). Contains derivatives of Harvard Institutional Books 1.0, Institutional Data Initiative, Harvard Law School Library; used for noncommercial research. Checksums: every file's SHA-256 is in the GitHub repository, data/manifests/hf_<repo>.jsonl.

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support