etherfuse commited on
Commit
5436be6
·
verified ·
1 Parent(s): 1d9cb2a

Card brought up to date for the finished study: what was found, where to read it, how to use this

Browse files
Files changed (1) hide show
  1. README.md +29 -1
README.md CHANGED
@@ -4,7 +4,7 @@ language: en
4
  tags: [authorship, fine-tuned, book-of-mormon-authorship, 1780-1830]
5
  ---
6
 
7
- # Candidate and control models (BOM-28)
8
 
9
  One fine-tuned copy of the time-locked model per candidate author's sample (`cand__...`) and per
10
  control group (`ctrl__...`), at 125M (`pilot/`), 297M (`base/`) and 590M (`mid/`), as scored on the
@@ -21,6 +21,34 @@ post hoc; the paper's corrected rankings). `a44/` holds the models of Joseph Smi
21
  non-revelatory writing (Amendment 44) and `a44/a45/` the equal-dose slice models (Amendment 45),
22
  each named `<unit>[_slice<k>]_<size>.pt`.
23
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
24
  ## About
25
 
26
  Part of an open, pre-registered study of the authorship, borrowing and dating of the
 
4
  tags: [authorship, fine-tuned, book-of-mormon-authorship, 1780-1830]
5
  ---
6
 
7
+ # Candidate models: who writes like the Book of Mormon?
8
 
9
  One fine-tuned copy of the time-locked model per candidate author's sample (`cand__...`) and per
10
  control group (`ctrl__...`), at 125M (`pilot/`), 297M (`base/`) and 590M (`mid/`), as scored on the
 
21
  non-revelatory writing (Amendment 44) and `a44/a45/` the equal-dose slice models (Amendment 45),
22
  each named `<unit>[_slice<k>]_<size>.pt`.
23
 
24
+ ## The study and what it found
25
+
26
+ This is one of the models from a finished, open study of the English text of the Book of Mormon:
27
+ who writes like it, what it borrows, and when its language dates from.
28
+
29
+ - **The findings in plain words, with live charts:** https://lifehug.com/book-of-mormon
30
+ - **The paper, the code and every results file:** https://github.com/lifehug/book-of-mormon-authorship
31
+
32
+ In short: the book copies almost nothing from earlier print. Among the eight people usually
33
+ proposed as its author, the closest writing is Joseph Smith's revelations, and a search of 23,872
34
+ authors of the period puts those revelations first or second. No tool we found, ours included, can
35
+ tell a person from the voice they are writing in, or reliably say "it is none of these people".
36
+ The study takes no position on inspiration.
37
+
38
+ ## How to check the result with these
39
+
40
+ No training and no restricted text are needed.
41
+
42
+ 1. Clone the GitHub repository and download the base model of the size you want
43
+ (`etherfuse/timelocked-1830-125m`, `-300m` or `-590m`) and the candidate files of the same
44
+ size from this page (`a43/pilot/` is 125M, `a43/base/` 297M, `a43/mid/` 590M).
45
+ 2. Load a candidate's `.pt` into the base model's architecture. It is a plain state dict.
46
+ 3. Score a passage: the candidate model's average loss on it, minus the base model's. The lower
47
+ the number, the more that candidate's writing prepared the model for the passage. The scoring
48
+ code is `alm/blind_score.py`.
49
+ 4. Compare with ours: the scores we got for every Book of Mormon segment are in `data/blind/` in
50
+ the GitHub repository.
51
+
52
  ## About
53
 
54
  Part of an open, pre-registered study of the authorship, borrowing and dating of the