AbstractPhil commited on
Commit
59cd3d8
·
verified ·
1 Parent(s): 66d9aad

caption pack: lexicon difficulty and surprise (bytelex atlas, 11 tokenizers, the diet table)

Browse files
mini-beatrix-3/caption_pack/lexicon_difficulty.json ADDED
@@ -0,0 +1,1339 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "utc": "2026-10-06T04:12:56Z",
3
+ "pack": "/root/pack_v2",
4
+ "rows_per_source": 300,
5
+ "seed": 20261005,
6
+ "fleet": [
7
+ "gpt2",
8
+ "smollm2",
9
+ "cl100k_base",
10
+ "o200k_base",
11
+ "llama3",
12
+ "qwen25",
13
+ "deepseek-v3",
14
+ "mistral-v03",
15
+ "xlmr",
16
+ "t5-unigram",
17
+ "bert-wordpiece"
18
+ ],
19
+ "bands": {
20
+ "easy_top_le": 0.42,
21
+ "trap_top_ge": 0.55
22
+ },
23
+ "diet_table": {
24
+ "witness_floor": 8.0,
25
+ "thresholds": [
26
+ 0.5,
27
+ 1.0,
28
+ 1.5,
29
+ 2.0,
30
+ 2.5,
31
+ 3.0
32
+ ],
33
+ "witnessed_cells": 57667
34
+ },
35
+ "references": {
36
+ "minted_easy": {
37
+ "words_total": 96,
38
+ "words_distinct": 96,
39
+ "hapax_share_of_types": 1.0,
40
+ "mean_word_len": 5.38,
41
+ "difficulty_fleet_consensus": {
42
+ "top_share_q10_50_90_types": [
43
+ 0.208,
44
+ 0.303,
45
+ 0.399
46
+ ],
47
+ "top_share_q10_50_90_tokens": [
48
+ 0.207,
49
+ 0.3,
50
+ 0.4
51
+ ],
52
+ "mean_top_types": 0.3,
53
+ "mean_top_tokens": 0.3,
54
+ "entropy_bits_q10_50_90_types": [
55
+ 2.348,
56
+ 2.932,
57
+ 3.525
58
+ ],
59
+ "closability_q10_50_90_types": [
60
+ 0.06,
61
+ 0.217,
62
+ 0.547
63
+ ],
64
+ "band_share_types": {
65
+ "easy": 1.0,
66
+ "mid": 0.0,
67
+ "trap": 0.0,
68
+ "novel": 0.0
69
+ },
70
+ "band_share_tokens": {
71
+ "easy": 1.0,
72
+ "mid": 0.0,
73
+ "trap": 0.0,
74
+ "novel": 0.0
75
+ },
76
+ "novel_word_ends_types": 0,
77
+ "novel_word_ends_tokens": 0
78
+ },
79
+ "surprise_per_tokenizer_tokens": {
80
+ "gpt2": {
81
+ "dead_cell_share": 0.0247,
82
+ "mean_surprisal_bits_witnessed_steps": 2.195,
83
+ "unwitnessed_step_share": 0.136
84
+ },
85
+ "smollm2": {
86
+ "dead_cell_share": 0.0309,
87
+ "mean_surprisal_bits_witnessed_steps": 2.254,
88
+ "unwitnessed_step_share": 0.1447
89
+ },
90
+ "cl100k_base": {
91
+ "dead_cell_share": 0.0185,
92
+ "mean_surprisal_bits_witnessed_steps": 2.367,
93
+ "unwitnessed_step_share": 0.0877
94
+ },
95
+ "o200k_base": {
96
+ "dead_cell_share": 0.0,
97
+ "mean_surprisal_bits_witnessed_steps": 2.628,
98
+ "unwitnessed_step_share": 0.0175
99
+ },
100
+ "llama3": {
101
+ "dead_cell_share": 0.0123,
102
+ "mean_surprisal_bits_witnessed_steps": 2.41,
103
+ "unwitnessed_step_share": 0.0746
104
+ },
105
+ "qwen25": {
106
+ "dead_cell_share": 0.0185,
107
+ "mean_surprisal_bits_witnessed_steps": 2.373,
108
+ "unwitnessed_step_share": 0.0877
109
+ },
110
+ "deepseek-v3": {
111
+ "dead_cell_share": 0.0154,
112
+ "mean_surprisal_bits_witnessed_steps": 2.386,
113
+ "unwitnessed_step_share": 0.0833
114
+ },
115
+ "mistral-v03": {
116
+ "dead_cell_share": 0.0432,
117
+ "mean_surprisal_bits_witnessed_steps": 2.048,
118
+ "unwitnessed_step_share": 0.2193
119
+ },
120
+ "xlmr": {
121
+ "dead_cell_share": 0.0,
122
+ "mean_surprisal_bits_witnessed_steps": 2.947,
123
+ "unwitnessed_step_share": 0.0395
124
+ },
125
+ "t5-unigram": {
126
+ "dead_cell_share": 0.0123,
127
+ "mean_surprisal_bits_witnessed_steps": 2.303,
128
+ "unwitnessed_step_share": 0.1404
129
+ },
130
+ "bert-wordpiece": {
131
+ "dead_cell_share": 0.0247,
132
+ "mean_surprisal_bits_witnessed_steps": 2.258,
133
+ "unwitnessed_step_share": 0.136
134
+ }
135
+ },
136
+ "diet_tokens": {
137
+ "cells": 324,
138
+ "under_floor_share": 0.0093,
139
+ "never_seen_share": 0.0,
140
+ "mean_entropy_bits_known_cells": 2.088
141
+ }
142
+ },
143
+ "minted_trap": {
144
+ "words_total": 96,
145
+ "words_distinct": 96,
146
+ "hapax_share_of_types": 1.0,
147
+ "mean_word_len": 5.06,
148
+ "difficulty_fleet_consensus": {
149
+ "top_share_q10_50_90_types": [
150
+ 0.587,
151
+ 0.733,
152
+ 0.967
153
+ ],
154
+ "top_share_q10_50_90_tokens": [
155
+ 0.585,
156
+ 0.732,
157
+ 0.968
158
+ ],
159
+ "mean_top_types": 0.751,
160
+ "mean_top_tokens": 0.751,
161
+ "entropy_bits_q10_50_90_types": [
162
+ 0.244,
163
+ 1.39,
164
+ 1.973
165
+ ],
166
+ "closability_q10_50_90_types": [
167
+ 0.008,
168
+ 0.07,
169
+ 0.387
170
+ ],
171
+ "band_share_types": {
172
+ "easy": 0.0,
173
+ "mid": 0.0,
174
+ "trap": 1.0,
175
+ "novel": 0.0
176
+ },
177
+ "band_share_tokens": {
178
+ "easy": 0.0,
179
+ "mid": 0.0,
180
+ "trap": 1.0,
181
+ "novel": 0.0
182
+ },
183
+ "novel_word_ends_types": 0,
184
+ "novel_word_ends_tokens": 0
185
+ },
186
+ "surprise_per_tokenizer_tokens": {
187
+ "gpt2": {
188
+ "dead_cell_share": 0.051,
189
+ "mean_surprisal_bits_witnessed_steps": 2.389,
190
+ "unwitnessed_step_share": 0.2071
191
+ },
192
+ "smollm2": {
193
+ "dead_cell_share": 0.0544,
194
+ "mean_surprisal_bits_witnessed_steps": 2.404,
195
+ "unwitnessed_step_share": 0.2121
196
+ },
197
+ "cl100k_base": {
198
+ "dead_cell_share": 0.0204,
199
+ "mean_surprisal_bits_witnessed_steps": 2.429,
200
+ "unwitnessed_step_share": 0.1465
201
+ },
202
+ "o200k_base": {
203
+ "dead_cell_share": 0.0068,
204
+ "mean_surprisal_bits_witnessed_steps": 2.732,
205
+ "unwitnessed_step_share": 0.0505
206
+ },
207
+ "llama3": {
208
+ "dead_cell_share": 0.017,
209
+ "mean_surprisal_bits_witnessed_steps": 2.492,
210
+ "unwitnessed_step_share": 0.1263
211
+ },
212
+ "qwen25": {
213
+ "dead_cell_share": 0.0102,
214
+ "mean_surprisal_bits_witnessed_steps": 2.513,
215
+ "unwitnessed_step_share": 0.1313
216
+ },
217
+ "deepseek-v3": {
218
+ "dead_cell_share": 0.0102,
219
+ "mean_surprisal_bits_witnessed_steps": 2.522,
220
+ "unwitnessed_step_share": 0.1313
221
+ },
222
+ "mistral-v03": {
223
+ "dead_cell_share": 0.0884,
224
+ "mean_surprisal_bits_witnessed_steps": 2.163,
225
+ "unwitnessed_step_share": 0.298
226
+ },
227
+ "xlmr": {
228
+ "dead_cell_share": 0.0068,
229
+ "mean_surprisal_bits_witnessed_steps": 3.079,
230
+ "unwitnessed_step_share": 0.096
231
+ },
232
+ "t5-unigram": {
233
+ "dead_cell_share": 0.0476,
234
+ "mean_surprisal_bits_witnessed_steps": 2.475,
235
+ "unwitnessed_step_share": 0.2222
236
+ },
237
+ "bert-wordpiece": {
238
+ "dead_cell_share": 0.0476,
239
+ "mean_surprisal_bits_witnessed_steps": 2.458,
240
+ "unwitnessed_step_share": 0.2172
241
+ }
242
+ },
243
+ "diet_tokens": {
244
+ "cells": 294,
245
+ "under_floor_share": 0.0306,
246
+ "never_seen_share": 0.0068,
247
+ "mean_entropy_bits_known_cells": 1.721
248
+ }
249
+ },
250
+ "minted_pairs": {
251
+ "words_total": 24,
252
+ "words_distinct": 24,
253
+ "hapax_share_of_types": 1.0,
254
+ "mean_word_len": 5.42,
255
+ "difficulty_fleet_consensus": {
256
+ "top_share_q10_50_90_types": [
257
+ 0.245,
258
+ 0.462,
259
+ 0.814
260
+ ],
261
+ "top_share_q10_50_90_tokens": [
262
+ 0.244,
263
+ 0.373,
264
+ 0.814
265
+ ],
266
+ "mean_top_types": 0.503,
267
+ "mean_top_tokens": 0.503,
268
+ "entropy_bits_q10_50_90_types": [
269
+ 0.811,
270
+ 2.072,
271
+ 3.377
272
+ ],
273
+ "closability_q10_50_90_types": [
274
+ 0.013,
275
+ 0.214,
276
+ 0.685
277
+ ],
278
+ "band_share_types": {
279
+ "easy": 0.5,
280
+ "mid": 0.0,
281
+ "trap": 0.5,
282
+ "novel": 0.0
283
+ },
284
+ "band_share_tokens": {
285
+ "easy": 0.5,
286
+ "mid": 0.0,
287
+ "trap": 0.5,
288
+ "novel": 0.0
289
+ },
290
+ "novel_word_ends_types": 0,
291
+ "novel_word_ends_tokens": 0
292
+ },
293
+ "surprise_per_tokenizer_tokens": {
294
+ "gpt2": {
295
+ "dead_cell_share": 0.0488,
296
+ "mean_surprisal_bits_witnessed_steps": 2.911,
297
+ "unwitnessed_step_share": 0.1552
298
+ },
299
+ "smollm2": {
300
+ "dead_cell_share": 0.0244,
301
+ "mean_surprisal_bits_witnessed_steps": 2.798,
302
+ "unwitnessed_step_share": 0.1379
303
+ },
304
+ "cl100k_base": {
305
+ "dead_cell_share": 0.0122,
306
+ "mean_surprisal_bits_witnessed_steps": 2.689,
307
+ "unwitnessed_step_share": 0.1207
308
+ },
309
+ "o200k_base": {
310
+ "dead_cell_share": 0.0,
311
+ "mean_surprisal_bits_witnessed_steps": 2.98,
312
+ "unwitnessed_step_share": 0.0517
313
+ },
314
+ "llama3": {
315
+ "dead_cell_share": 0.0122,
316
+ "mean_surprisal_bits_witnessed_steps": 2.7,
317
+ "unwitnessed_step_share": 0.1207
318
+ },
319
+ "qwen25": {
320
+ "dead_cell_share": 0.0122,
321
+ "mean_surprisal_bits_witnessed_steps": 2.696,
322
+ "unwitnessed_step_share": 0.1207
323
+ },
324
+ "deepseek-v3": {
325
+ "dead_cell_share": 0.0244,
326
+ "mean_surprisal_bits_witnessed_steps": 2.716,
327
+ "unwitnessed_step_share": 0.1379
328
+ },
329
+ "mistral-v03": {
330
+ "dead_cell_share": 0.061,
331
+ "mean_surprisal_bits_witnessed_steps": 2.487,
332
+ "unwitnessed_step_share": 0.2414
333
+ },
334
+ "xlmr": {
335
+ "dead_cell_share": 0.0244,
336
+ "mean_surprisal_bits_witnessed_steps": 3.325,
337
+ "unwitnessed_step_share": 0.1034
338
+ },
339
+ "t5-unigram": {
340
+ "dead_cell_share": 0.0488,
341
+ "mean_surprisal_bits_witnessed_steps": 2.737,
342
+ "unwitnessed_step_share": 0.2069
343
+ },
344
+ "bert-wordpiece": {
345
+ "dead_cell_share": 0.0244,
346
+ "mean_surprisal_bits_witnessed_steps": 3.042,
347
+ "unwitnessed_step_share": 0.1552
348
+ }
349
+ },
350
+ "diet_tokens": {
351
+ "cells": 82,
352
+ "under_floor_share": 0.0122,
353
+ "never_seen_share": 0.0,
354
+ "mean_entropy_bits_known_cells": 2.088
355
+ }
356
+ },
357
+ "legacy_aliens": {
358
+ "words_total": 8,
359
+ "words_distinct": 8,
360
+ "hapax_share_of_types": 1.0,
361
+ "mean_word_len": 4.0,
362
+ "difficulty_fleet_consensus": {
363
+ "top_share_q10_50_90_types": [
364
+ 0.346,
365
+ 0.413,
366
+ 0.641
367
+ ],
368
+ "top_share_q10_50_90_tokens": [
369
+ 0.315,
370
+ 0.41,
371
+ 0.667
372
+ ],
373
+ "mean_top_types": 0.457,
374
+ "mean_top_tokens": 0.457,
375
+ "entropy_bits_q10_50_90_types": [
376
+ 1.592,
377
+ 2.512,
378
+ 2.74
379
+ ],
380
+ "closability_q10_50_90_types": [
381
+ 0.051,
382
+ 0.195,
383
+ 0.341
384
+ ],
385
+ "band_share_types": {
386
+ "easy": 0.625,
387
+ "mid": 0.125,
388
+ "trap": 0.25,
389
+ "novel": 0.0
390
+ },
391
+ "band_share_tokens": {
392
+ "easy": 0.625,
393
+ "mid": 0.125,
394
+ "trap": 0.25,
395
+ "novel": 0.0
396
+ },
397
+ "novel_word_ends_types": 0,
398
+ "novel_word_ends_tokens": 0
399
+ },
400
+ "surprise_per_tokenizer_tokens": {
401
+ "gpt2": {
402
+ "dead_cell_share": 0.0,
403
+ "mean_surprisal_bits_witnessed_steps": 3.573,
404
+ "unwitnessed_step_share": 0.25
405
+ },
406
+ "smollm2": {
407
+ "dead_cell_share": 0.0625,
408
+ "mean_surprisal_bits_witnessed_steps": 3.65,
409
+ "unwitnessed_step_share": 0.25
410
+ },
411
+ "cl100k_base": {
412
+ "dead_cell_share": 0.0,
413
+ "mean_surprisal_bits_witnessed_steps": 3.784,
414
+ "unwitnessed_step_share": 0.25
415
+ },
416
+ "o200k_base": {
417
+ "dead_cell_share": 0.0,
418
+ "mean_surprisal_bits_witnessed_steps": 3.853,
419
+ "unwitnessed_step_share": 0.25
420
+ },
421
+ "llama3": {
422
+ "dead_cell_share": 0.0,
423
+ "mean_surprisal_bits_witnessed_steps": 3.786,
424
+ "unwitnessed_step_share": 0.25
425
+ },
426
+ "qwen25": {
427
+ "dead_cell_share": 0.0,
428
+ "mean_surprisal_bits_witnessed_steps": 3.825,
429
+ "unwitnessed_step_share": 0.25
430
+ },
431
+ "deepseek-v3": {
432
+ "dead_cell_share": 0.0625,
433
+ "mean_surprisal_bits_witnessed_steps": 3.613,
434
+ "unwitnessed_step_share": 0.25
435
+ },
436
+ "mistral-v03": {
437
+ "dead_cell_share": 0.0625,
438
+ "mean_surprisal_bits_witnessed_steps": 2.773,
439
+ "unwitnessed_step_share": 0.375
440
+ },
441
+ "xlmr": {
442
+ "dead_cell_share": 0.0,
443
+ "mean_surprisal_bits_witnessed_steps": 4.075,
444
+ "unwitnessed_step_share": 0.125
445
+ },
446
+ "t5-unigram": {
447
+ "dead_cell_share": 0.0625,
448
+ "mean_surprisal_bits_witnessed_steps": 3.658,
449
+ "unwitnessed_step_share": 0.25
450
+ },
451
+ "bert-wordpiece": {
452
+ "dead_cell_share": 0.0,
453
+ "mean_surprisal_bits_witnessed_steps": 3.308,
454
+ "unwitnessed_step_share": 0.25
455
+ }
456
+ },
457
+ "diet_tokens": {
458
+ "cells": 16,
459
+ "under_floor_share": 0.0,
460
+ "never_seen_share": 0.0,
461
+ "mean_entropy_bits_known_cells": 1.949
462
+ }
463
+ }
464
+ },
465
+ "sources": {
466
+ "coco": {
467
+ "words_total": 7941,
468
+ "words_distinct": 785,
469
+ "hapax_share_of_types": 0.031,
470
+ "mean_word_len": 6.13,
471
+ "difficulty_fleet_consensus": {
472
+ "top_share_q10_50_90_types": [
473
+ 0.229,
474
+ 0.355,
475
+ 0.566
476
+ ],
477
+ "top_share_q10_50_90_tokens": [
478
+ 0.23,
479
+ 0.371,
480
+ 0.797
481
+ ],
482
+ "mean_top_types": 0.384,
483
+ "mean_top_tokens": 0.44,
484
+ "entropy_bits_q10_50_90_types": [
485
+ 1.814,
486
+ 2.884,
487
+ 3.437
488
+ ],
489
+ "closability_q10_50_90_types": [
490
+ 0.233,
491
+ 0.506,
492
+ 0.86
493
+ ],
494
+ "band_share_types": {
495
+ "easy": 0.722,
496
+ "mid": 0.161,
497
+ "trap": 0.117,
498
+ "novel": 0.0
499
+ },
500
+ "band_share_tokens": {
501
+ "easy": 0.659,
502
+ "mid": 0.116,
503
+ "trap": 0.225,
504
+ "novel": 0.0
505
+ },
506
+ "novel_word_ends_types": 0,
507
+ "novel_word_ends_tokens": 0
508
+ },
509
+ "surprise_per_tokenizer_tokens": {
510
+ "gpt2": {
511
+ "dead_cell_share": 0.0016,
512
+ "mean_surprisal_bits_witnessed_steps": 1.598,
513
+ "unwitnessed_step_share": 0.0115
514
+ },
515
+ "smollm2": {
516
+ "dead_cell_share": 0.0016,
517
+ "mean_surprisal_bits_witnessed_steps": 1.572,
518
+ "unwitnessed_step_share": 0.0084
519
+ },
520
+ "cl100k_base": {
521
+ "dead_cell_share": 0.0018,
522
+ "mean_surprisal_bits_witnessed_steps": 1.587,
523
+ "unwitnessed_step_share": 0.0102
524
+ },
525
+ "o200k_base": {
526
+ "dead_cell_share": 0.0011,
527
+ "mean_surprisal_bits_witnessed_steps": 1.734,
528
+ "unwitnessed_step_share": 0.0052
529
+ },
530
+ "llama3": {
531
+ "dead_cell_share": 0.0018,
532
+ "mean_surprisal_bits_witnessed_steps": 1.598,
533
+ "unwitnessed_step_share": 0.0102
534
+ },
535
+ "qwen25": {
536
+ "dead_cell_share": 0.0018,
537
+ "mean_surprisal_bits_witnessed_steps": 1.595,
538
+ "unwitnessed_step_share": 0.0102
539
+ },
540
+ "deepseek-v3": {
541
+ "dead_cell_share": 0.0013,
542
+ "mean_surprisal_bits_witnessed_steps": 1.652,
543
+ "unwitnessed_step_share": 0.0093
544
+ },
545
+ "mistral-v03": {
546
+ "dead_cell_share": 0.0052,
547
+ "mean_surprisal_bits_witnessed_steps": 1.375,
548
+ "unwitnessed_step_share": 0.0339
549
+ },
550
+ "xlmr": {
551
+ "dead_cell_share": 0.0043,
552
+ "mean_surprisal_bits_witnessed_steps": 2.566,
553
+ "unwitnessed_step_share": 0.0337
554
+ },
555
+ "t5-unigram": {
556
+ "dead_cell_share": 0.002,
557
+ "mean_surprisal_bits_witnessed_steps": 1.733,
558
+ "unwitnessed_step_share": 0.0148
559
+ },
560
+ "bert-wordpiece": {
561
+ "dead_cell_share": 0.0013,
562
+ "mean_surprisal_bits_witnessed_steps": 1.678,
563
+ "unwitnessed_step_share": 0.0097
564
+ }
565
+ },
566
+ "diet_tokens": {
567
+ "cells": 32822,
568
+ "under_floor_share": 0.0011,
569
+ "never_seen_share": 0.0001,
570
+ "mean_entropy_bits_known_cells": 1.786
571
+ },
572
+ "rows_sampled": 300,
573
+ "rows_total": 19341,
574
+ "bytes_per_row": 250.3,
575
+ "words_per_row": 26.5,
576
+ "registers": {
577
+ "tags": null,
578
+ "english": {
579
+ "words_total": 7341,
580
+ "words_distinct": 783,
581
+ "hapax_share_of_types": 0.031,
582
+ "mean_word_len": 5.98,
583
+ "difficulty_fleet_consensus": {
584
+ "top_share_q10_50_90_types": [
585
+ 0.229,
586
+ 0.355,
587
+ 0.566
588
+ ],
589
+ "top_share_q10_50_90_tokens": [
590
+ 0.23,
591
+ 0.371,
592
+ 0.797
593
+ ],
594
+ "mean_top_types": 0.384,
595
+ "mean_top_tokens": 0.438,
596
+ "entropy_bits_q10_50_90_types": [
597
+ 1.814,
598
+ 2.889,
599
+ 3.437
600
+ ],
601
+ "closability_q10_50_90_types": [
602
+ 0.232,
603
+ 0.506,
604
+ 0.86
605
+ ],
606
+ "band_share_types": {
607
+ "easy": 0.723,
608
+ "mid": 0.161,
609
+ "trap": 0.116,
610
+ "novel": 0.0
611
+ },
612
+ "band_share_tokens": {
613
+ "easy": 0.673,
614
+ "mid": 0.125,
615
+ "trap": 0.202,
616
+ "novel": 0.0
617
+ },
618
+ "novel_word_ends_types": 0,
619
+ "novel_word_ends_tokens": 0
620
+ },
621
+ "surprise_per_tokenizer_tokens": {
622
+ "gpt2": {
623
+ "dead_cell_share": 0.0018,
624
+ "mean_surprisal_bits_witnessed_steps": 1.649,
625
+ "unwitnessed_step_share": 0.013
626
+ },
627
+ "smollm2": {
628
+ "dead_cell_share": 0.0018,
629
+ "mean_surprisal_bits_witnessed_steps": 1.64,
630
+ "unwitnessed_step_share": 0.0095
631
+ },
632
+ "cl100k_base": {
633
+ "dead_cell_share": 0.002,
634
+ "mean_surprisal_bits_witnessed_steps": 1.652,
635
+ "unwitnessed_step_share": 0.0116
636
+ },
637
+ "o200k_base": {
638
+ "dead_cell_share": 0.0012,
639
+ "mean_surprisal_bits_witnessed_steps": 1.808,
640
+ "unwitnessed_step_share": 0.0059
641
+ },
642
+ "llama3": {
643
+ "dead_cell_share": 0.002,
644
+ "mean_surprisal_bits_witnessed_steps": 1.663,
645
+ "unwitnessed_step_share": 0.0116
646
+ },
647
+ "qwen25": {
648
+ "dead_cell_share": 0.002,
649
+ "mean_surprisal_bits_witnessed_steps": 1.659,
650
+ "unwitnessed_step_share": 0.0116
651
+ },
652
+ "deepseek-v3": {
653
+ "dead_cell_share": 0.0015,
654
+ "mean_surprisal_bits_witnessed_steps": 1.725,
655
+ "unwitnessed_step_share": 0.0106
656
+ },
657
+ "mistral-v03": {
658
+ "dead_cell_share": 0.0058,
659
+ "mean_surprisal_bits_witnessed_steps": 1.416,
660
+ "unwitnessed_step_share": 0.0385
661
+ },
662
+ "xlmr": {
663
+ "dead_cell_share": 0.0049,
664
+ "mean_surprisal_bits_witnessed_steps": 2.655,
665
+ "unwitnessed_step_share": 0.0383
666
+ },
667
+ "t5-unigram": {
668
+ "dead_cell_share": 0.0022,
669
+ "mean_surprisal_bits_witnessed_steps": 1.796,
670
+ "unwitnessed_step_share": 0.0169
671
+ },
672
+ "bert-wordpiece": {
673
+ "dead_cell_share": 0.0015,
674
+ "mean_surprisal_bits_witnessed_steps": 1.739,
675
+ "unwitnessed_step_share": 0.0111
676
+ }
677
+ },
678
+ "diet_tokens": {
679
+ "cells": 29222,
680
+ "under_floor_share": 0.0013,
681
+ "never_seen_share": 0.0001,
682
+ "mean_entropy_bits_known_cells": 1.766
683
+ }
684
+ }
685
+ }
686
+ },
687
+ "scf": {
688
+ "words_total": 60909,
689
+ "words_distinct": 1579,
690
+ "hapax_share_of_types": 0.218,
691
+ "mean_word_len": 5.44,
692
+ "difficulty_fleet_consensus": {
693
+ "top_share_q10_50_90_types": [
694
+ 0.219,
695
+ 0.355,
696
+ 0.598
697
+ ],
698
+ "top_share_q10_50_90_tokens": [
699
+ 0.199,
700
+ 0.352,
701
+ 0.559
702
+ ],
703
+ "mean_top_types": 0.388,
704
+ "mean_top_tokens": 0.369,
705
+ "entropy_bits_q10_50_90_types": [
706
+ 1.673,
707
+ 2.812,
708
+ 3.448
709
+ ],
710
+ "closability_q10_50_90_types": [
711
+ 0.222,
712
+ 0.506,
713
+ 0.912
714
+ ],
715
+ "band_share_types": {
716
+ "easy": 0.679,
717
+ "mid": 0.177,
718
+ "trap": 0.144,
719
+ "novel": 0.001
720
+ },
721
+ "band_share_tokens": {
722
+ "easy": 0.71,
723
+ "mid": 0.18,
724
+ "trap": 0.11,
725
+ "novel": 0.0
726
+ },
727
+ "novel_word_ends_types": 1,
728
+ "novel_word_ends_tokens": 25
729
+ },
730
+ "surprise_per_tokenizer_tokens": {
731
+ "gpt2": {
732
+ "dead_cell_share": 0.0016,
733
+ "mean_surprisal_bits_witnessed_steps": 2.023,
734
+ "unwitnessed_step_share": 0.0181
735
+ },
736
+ "smollm2": {
737
+ "dead_cell_share": 0.0014,
738
+ "mean_surprisal_bits_witnessed_steps": 2.052,
739
+ "unwitnessed_step_share": 0.0206
740
+ },
741
+ "cl100k_base": {
742
+ "dead_cell_share": 0.0008,
743
+ "mean_surprisal_bits_witnessed_steps": 2.062,
744
+ "unwitnessed_step_share": 0.0135
745
+ },
746
+ "o200k_base": {
747
+ "dead_cell_share": 0.0002,
748
+ "mean_surprisal_bits_witnessed_steps": 2.211,
749
+ "unwitnessed_step_share": 0.0091
750
+ },
751
+ "llama3": {
752
+ "dead_cell_share": 0.0007,
753
+ "mean_surprisal_bits_witnessed_steps": 2.086,
754
+ "unwitnessed_step_share": 0.0135
755
+ },
756
+ "qwen25": {
757
+ "dead_cell_share": 0.0008,
758
+ "mean_surprisal_bits_witnessed_steps": 2.075,
759
+ "unwitnessed_step_share": 0.0131
760
+ },
761
+ "deepseek-v3": {
762
+ "dead_cell_share": 0.0005,
763
+ "mean_surprisal_bits_witnessed_steps": 2.102,
764
+ "unwitnessed_step_share": 0.0168
765
+ },
766
+ "mistral-v03": {
767
+ "dead_cell_share": 0.0042,
768
+ "mean_surprisal_bits_witnessed_steps": 1.828,
769
+ "unwitnessed_step_share": 0.0674
770
+ },
771
+ "xlmr": {
772
+ "dead_cell_share": 0.0015,
773
+ "mean_surprisal_bits_witnessed_steps": 2.753,
774
+ "unwitnessed_step_share": 0.0466
775
+ },
776
+ "t5-unigram": {
777
+ "dead_cell_share": 0.0016,
778
+ "mean_surprisal_bits_witnessed_steps": 2.059,
779
+ "unwitnessed_step_share": 0.0223
780
+ },
781
+ "bert-wordpiece": {
782
+ "dead_cell_share": 0.0011,
783
+ "mean_surprisal_bits_witnessed_steps": 2.046,
784
+ "unwitnessed_step_share": 0.0119
785
+ }
786
+ },
787
+ "diet_tokens": {
788
+ "cells": 209568,
789
+ "under_floor_share": 0.0002,
790
+ "never_seen_share": 0.0001,
791
+ "mean_entropy_bits_known_cells": 1.986
792
+ },
793
+ "rows_sampled": 301,
794
+ "rows_total": 42257,
795
+ "bytes_per_row": 1442.8,
796
+ "words_per_row": 202.4,
797
+ "registers": {
798
+ "tags": {
799
+ "words_total": 9014,
800
+ "words_distinct": 323,
801
+ "hapax_share_of_types": 0.241,
802
+ "mean_word_len": 5.64,
803
+ "difficulty_fleet_consensus": {
804
+ "top_share_q10_50_90_types": [
805
+ 0.207,
806
+ 0.34,
807
+ 0.564
808
+ ],
809
+ "top_share_q10_50_90_tokens": [
810
+ 0.211,
811
+ 0.355,
812
+ 0.522
813
+ ],
814
+ "mean_top_types": 0.364,
815
+ "mean_top_tokens": 0.371,
816
+ "entropy_bits_q10_50_90_types": [
817
+ 1.887,
818
+ 2.969,
819
+ 3.475
820
+ ],
821
+ "closability_q10_50_90_types": [
822
+ 0.188,
823
+ 0.425,
824
+ 0.86
825
+ ],
826
+ "band_share_types": {
827
+ "easy": 0.759,
828
+ "mid": 0.13,
829
+ "trap": 0.111,
830
+ "novel": 0.0
831
+ },
832
+ "band_share_tokens": {
833
+ "easy": 0.705,
834
+ "mid": 0.203,
835
+ "trap": 0.092,
836
+ "novel": 0.0
837
+ },
838
+ "novel_word_ends_types": 0,
839
+ "novel_word_ends_tokens": 0
840
+ },
841
+ "surprise_per_tokenizer_tokens": {
842
+ "gpt2": {
843
+ "dead_cell_share": 0.0021,
844
+ "mean_surprisal_bits_witnessed_steps": 2.081,
845
+ "unwitnessed_step_share": 0.0226
846
+ },
847
+ "smollm2": {
848
+ "dead_cell_share": 0.0023,
849
+ "mean_surprisal_bits_witnessed_steps": 2.1,
850
+ "unwitnessed_step_share": 0.0275
851
+ },
852
+ "cl100k_base": {
853
+ "dead_cell_share": 0.0005,
854
+ "mean_surprisal_bits_witnessed_steps": 2.163,
855
+ "unwitnessed_step_share": 0.0182
856
+ },
857
+ "o200k_base": {
858
+ "dead_cell_share": 0.0,
859
+ "mean_surprisal_bits_witnessed_steps": 2.27,
860
+ "unwitnessed_step_share": 0.0135
861
+ },
862
+ "llama3": {
863
+ "dead_cell_share": 0.0005,
864
+ "mean_surprisal_bits_witnessed_steps": 2.199,
865
+ "unwitnessed_step_share": 0.0182
866
+ },
867
+ "qwen25": {
868
+ "dead_cell_share": 0.0005,
869
+ "mean_surprisal_bits_witnessed_steps": 2.179,
870
+ "unwitnessed_step_share": 0.0174
871
+ },
872
+ "deepseek-v3": {
873
+ "dead_cell_share": 0.0001,
874
+ "mean_surprisal_bits_witnessed_steps": 2.176,
875
+ "unwitnessed_step_share": 0.0222
876
+ },
877
+ "mistral-v03": {
878
+ "dead_cell_share": 0.0066,
879
+ "mean_surprisal_bits_witnessed_steps": 1.882,
880
+ "unwitnessed_step_share": 0.0863
881
+ },
882
+ "xlmr": {
883
+ "dead_cell_share": 0.0002,
884
+ "mean_surprisal_bits_witnessed_steps": 2.791,
885
+ "unwitnessed_step_share": 0.0501
886
+ },
887
+ "t5-unigram": {
888
+ "dead_cell_share": 0.0031,
889
+ "mean_surprisal_bits_witnessed_steps": 2.106,
890
+ "unwitnessed_step_share": 0.0233
891
+ },
892
+ "bert-wordpiece": {
893
+ "dead_cell_share": 0.0028,
894
+ "mean_surprisal_bits_witnessed_steps": 2.063,
895
+ "unwitnessed_step_share": 0.023
896
+ }
897
+ },
898
+ "diet_tokens": {
899
+ "cells": 32791,
900
+ "under_floor_share": 0.0,
901
+ "never_seen_share": 0.0,
902
+ "mean_entropy_bits_known_cells": 2.041
903
+ }
904
+ },
905
+ "english": {
906
+ "words_total": 34773,
907
+ "words_distinct": 1484,
908
+ "hapax_share_of_types": 0.282,
909
+ "mean_word_len": 5.34,
910
+ "difficulty_fleet_consensus": {
911
+ "top_share_q10_50_90_types": [
912
+ 0.22,
913
+ 0.355,
914
+ 0.6
915
+ ],
916
+ "top_share_q10_50_90_tokens": [
917
+ 0.19,
918
+ 0.343,
919
+ 0.562
920
+ ],
921
+ "mean_top_types": 0.389,
922
+ "mean_top_tokens": 0.365,
923
+ "entropy_bits_q10_50_90_types": [
924
+ 1.665,
925
+ 2.812,
926
+ 3.444
927
+ ],
928
+ "closability_q10_50_90_types": [
929
+ 0.223,
930
+ 0.507,
931
+ 0.916
932
+ ],
933
+ "band_share_types": {
934
+ "easy": 0.675,
935
+ "mid": 0.181,
936
+ "trap": 0.144,
937
+ "novel": 0.001
938
+ },
939
+ "band_share_tokens": {
940
+ "easy": 0.715,
941
+ "mid": 0.168,
942
+ "trap": 0.115,
943
+ "novel": 0.001
944
+ },
945
+ "novel_word_ends_types": 1,
946
+ "novel_word_ends_tokens": 25
947
+ },
948
+ "surprise_per_tokenizer_tokens": {
949
+ "gpt2": {
950
+ "dead_cell_share": 0.0015,
951
+ "mean_surprisal_bits_witnessed_steps": 1.991,
952
+ "unwitnessed_step_share": 0.0169
953
+ },
954
+ "smollm2": {
955
+ "dead_cell_share": 0.0012,
956
+ "mean_surprisal_bits_witnessed_steps": 2.023,
957
+ "unwitnessed_step_share": 0.0187
958
+ },
959
+ "cl100k_base": {
960
+ "dead_cell_share": 0.0009,
961
+ "mean_surprisal_bits_witnessed_steps": 2.027,
962
+ "unwitnessed_step_share": 0.0123
963
+ },
964
+ "o200k_base": {
965
+ "dead_cell_share": 0.0003,
966
+ "mean_surprisal_bits_witnessed_steps": 2.174,
967
+ "unwitnessed_step_share": 0.0085
968
+ },
969
+ "llama3": {
970
+ "dead_cell_share": 0.0008,
971
+ "mean_surprisal_bits_witnessed_steps": 2.047,
972
+ "unwitnessed_step_share": 0.0122
973
+ },
974
+ "qwen25": {
975
+ "dead_cell_share": 0.0009,
976
+ "mean_surprisal_bits_witnessed_steps": 2.039,
977
+ "unwitnessed_step_share": 0.012
978
+ },
979
+ "deepseek-v3": {
980
+ "dead_cell_share": 0.0007,
981
+ "mean_surprisal_bits_witnessed_steps": 2.066,
982
+ "unwitnessed_step_share": 0.0153
983
+ },
984
+ "mistral-v03": {
985
+ "dead_cell_share": 0.0041,
986
+ "mean_surprisal_bits_witnessed_steps": 1.806,
987
+ "unwitnessed_step_share": 0.0594
988
+ },
989
+ "xlmr": {
990
+ "dead_cell_share": 0.0018,
991
+ "mean_surprisal_bits_witnessed_steps": 2.725,
992
+ "unwitnessed_step_share": 0.0456
993
+ },
994
+ "t5-unigram": {
995
+ "dead_cell_share": 0.0015,
996
+ "mean_surprisal_bits_witnessed_steps": 2.021,
997
+ "unwitnessed_step_share": 0.0227
998
+ },
999
+ "bert-wordpiece": {
1000
+ "dead_cell_share": 0.0009,
1001
+ "mean_surprisal_bits_witnessed_steps": 2.024,
1002
+ "unwitnessed_step_share": 0.0102
1003
+ }
1004
+ },
1005
+ "diet_tokens": {
1006
+ "cells": 116040,
1007
+ "under_floor_share": 0.0003,
1008
+ "never_seen_share": 0.0002,
1009
+ "mean_entropy_bits_known_cells": 1.928
1010
+ }
1011
+ }
1012
+ }
1013
+ }
1014
+ },
1015
+ "registers": {
1016
+ "tags": {
1017
+ "words_total": 9014,
1018
+ "words_distinct": 323,
1019
+ "hapax_share_of_types": 0.241,
1020
+ "mean_word_len": 5.64,
1021
+ "difficulty_fleet_consensus": {
1022
+ "top_share_q10_50_90_types": [
1023
+ 0.207,
1024
+ 0.34,
1025
+ 0.564
1026
+ ],
1027
+ "top_share_q10_50_90_tokens": [
1028
+ 0.211,
1029
+ 0.355,
1030
+ 0.522
1031
+ ],
1032
+ "mean_top_types": 0.364,
1033
+ "mean_top_tokens": 0.371,
1034
+ "entropy_bits_q10_50_90_types": [
1035
+ 1.887,
1036
+ 2.969,
1037
+ 3.475
1038
+ ],
1039
+ "closability_q10_50_90_types": [
1040
+ 0.188,
1041
+ 0.425,
1042
+ 0.86
1043
+ ],
1044
+ "band_share_types": {
1045
+ "easy": 0.759,
1046
+ "mid": 0.13,
1047
+ "trap": 0.111,
1048
+ "novel": 0.0
1049
+ },
1050
+ "band_share_tokens": {
1051
+ "easy": 0.705,
1052
+ "mid": 0.203,
1053
+ "trap": 0.092,
1054
+ "novel": 0.0
1055
+ },
1056
+ "novel_word_ends_types": 0,
1057
+ "novel_word_ends_tokens": 0
1058
+ },
1059
+ "surprise_per_tokenizer_tokens": {
1060
+ "gpt2": {
1061
+ "dead_cell_share": 0.0021,
1062
+ "mean_surprisal_bits_witnessed_steps": 2.081,
1063
+ "unwitnessed_step_share": 0.0226
1064
+ },
1065
+ "smollm2": {
1066
+ "dead_cell_share": 0.0023,
1067
+ "mean_surprisal_bits_witnessed_steps": 2.1,
1068
+ "unwitnessed_step_share": 0.0275
1069
+ },
1070
+ "cl100k_base": {
1071
+ "dead_cell_share": 0.0005,
1072
+ "mean_surprisal_bits_witnessed_steps": 2.163,
1073
+ "unwitnessed_step_share": 0.0182
1074
+ },
1075
+ "o200k_base": {
1076
+ "dead_cell_share": 0.0,
1077
+ "mean_surprisal_bits_witnessed_steps": 2.27,
1078
+ "unwitnessed_step_share": 0.0135
1079
+ },
1080
+ "llama3": {
1081
+ "dead_cell_share": 0.0005,
1082
+ "mean_surprisal_bits_witnessed_steps": 2.199,
1083
+ "unwitnessed_step_share": 0.0182
1084
+ },
1085
+ "qwen25": {
1086
+ "dead_cell_share": 0.0005,
1087
+ "mean_surprisal_bits_witnessed_steps": 2.179,
1088
+ "unwitnessed_step_share": 0.0174
1089
+ },
1090
+ "deepseek-v3": {
1091
+ "dead_cell_share": 0.0001,
1092
+ "mean_surprisal_bits_witnessed_steps": 2.176,
1093
+ "unwitnessed_step_share": 0.0222
1094
+ },
1095
+ "mistral-v03": {
1096
+ "dead_cell_share": 0.0066,
1097
+ "mean_surprisal_bits_witnessed_steps": 1.882,
1098
+ "unwitnessed_step_share": 0.0863
1099
+ },
1100
+ "xlmr": {
1101
+ "dead_cell_share": 0.0002,
1102
+ "mean_surprisal_bits_witnessed_steps": 2.791,
1103
+ "unwitnessed_step_share": 0.0501
1104
+ },
1105
+ "t5-unigram": {
1106
+ "dead_cell_share": 0.0031,
1107
+ "mean_surprisal_bits_witnessed_steps": 2.106,
1108
+ "unwitnessed_step_share": 0.0233
1109
+ },
1110
+ "bert-wordpiece": {
1111
+ "dead_cell_share": 0.0028,
1112
+ "mean_surprisal_bits_witnessed_steps": 2.063,
1113
+ "unwitnessed_step_share": 0.023
1114
+ }
1115
+ },
1116
+ "diet_tokens": {
1117
+ "cells": 32791,
1118
+ "under_floor_share": 0.0,
1119
+ "never_seen_share": 0.0,
1120
+ "mean_entropy_bits_known_cells": 2.041
1121
+ }
1122
+ },
1123
+ "english": {
1124
+ "words_total": 42114,
1125
+ "words_distinct": 1962,
1126
+ "hapax_share_of_types": 0.196,
1127
+ "mean_word_len": 5.45,
1128
+ "difficulty_fleet_consensus": {
1129
+ "top_share_q10_50_90_types": [
1130
+ 0.222,
1131
+ 0.355,
1132
+ 0.598
1133
+ ],
1134
+ "top_share_q10_50_90_tokens": [
1135
+ 0.19,
1136
+ 0.352,
1137
+ 0.643
1138
+ ],
1139
+ "mean_top_types": 0.39,
1140
+ "mean_top_tokens": 0.378,
1141
+ "entropy_bits_q10_50_90_types": [
1142
+ 1.677,
1143
+ 2.82,
1144
+ 3.437
1145
+ ],
1146
+ "closability_q10_50_90_types": [
1147
+ 0.227,
1148
+ 0.51,
1149
+ 0.901
1150
+ ],
1151
+ "band_share_types": {
1152
+ "easy": 0.682,
1153
+ "mid": 0.177,
1154
+ "trap": 0.14,
1155
+ "novel": 0.001
1156
+ },
1157
+ "band_share_tokens": {
1158
+ "easy": 0.708,
1159
+ "mid": 0.161,
1160
+ "trap": 0.131,
1161
+ "novel": 0.001
1162
+ },
1163
+ "novel_word_ends_types": 1,
1164
+ "novel_word_ends_tokens": 25
1165
+ },
1166
+ "surprise_per_tokenizer_tokens": {
1167
+ "gpt2": {
1168
+ "dead_cell_share": 0.0016,
1169
+ "mean_surprisal_bits_witnessed_steps": 1.918,
1170
+ "unwitnessed_step_share": 0.0161
1171
+ },
1172
+ "smollm2": {
1173
+ "dead_cell_share": 0.0014,
1174
+ "mean_surprisal_bits_witnessed_steps": 1.941,
1175
+ "unwitnessed_step_share": 0.0167
1176
+ },
1177
+ "cl100k_base": {
1178
+ "dead_cell_share": 0.0011,
1179
+ "mean_surprisal_bits_witnessed_steps": 1.947,
1180
+ "unwitnessed_step_share": 0.0121
1181
+ },
1182
+ "o200k_base": {
1183
+ "dead_cell_share": 0.0005,
1184
+ "mean_surprisal_bits_witnessed_steps": 2.096,
1185
+ "unwitnessed_step_share": 0.0079
1186
+ },
1187
+ "llama3": {
1188
+ "dead_cell_share": 0.001,
1189
+ "mean_surprisal_bits_witnessed_steps": 1.966,
1190
+ "unwitnessed_step_share": 0.0121
1191
+ },
1192
+ "qwen25": {
1193
+ "dead_cell_share": 0.0011,
1194
+ "mean_surprisal_bits_witnessed_steps": 1.958,
1195
+ "unwitnessed_step_share": 0.0119
1196
+ },
1197
+ "deepseek-v3": {
1198
+ "dead_cell_share": 0.0008,
1199
+ "mean_surprisal_bits_witnessed_steps": 1.994,
1200
+ "unwitnessed_step_share": 0.0143
1201
+ },
1202
+ "mistral-v03": {
1203
+ "dead_cell_share": 0.0044,
1204
+ "mean_surprisal_bits_witnessed_steps": 1.722,
1205
+ "unwitnessed_step_share": 0.0549
1206
+ },
1207
+ "xlmr": {
1208
+ "dead_cell_share": 0.0024,
1209
+ "mean_surprisal_bits_witnessed_steps": 2.71,
1210
+ "unwitnessed_step_share": 0.0441
1211
+ },
1212
+ "t5-unigram": {
1213
+ "dead_cell_share": 0.0016,
1214
+ "mean_surprisal_bits_witnessed_steps": 1.973,
1215
+ "unwitnessed_step_share": 0.0214
1216
+ },
1217
+ "bert-wordpiece": {
1218
+ "dead_cell_share": 0.001,
1219
+ "mean_surprisal_bits_witnessed_steps": 1.964,
1220
+ "unwitnessed_step_share": 0.0104
1221
+ }
1222
+ },
1223
+ "diet_tokens": {
1224
+ "cells": 145262,
1225
+ "under_floor_share": 0.0005,
1226
+ "never_seen_share": 0.0002,
1227
+ "mean_entropy_bits_known_cells": 1.895
1228
+ }
1229
+ },
1230
+ "whole_pack_sample": {
1231
+ "words_total": 68850,
1232
+ "words_distinct": 2039,
1233
+ "hapax_share_of_types": 0.153,
1234
+ "mean_word_len": 5.52,
1235
+ "difficulty_fleet_consensus": {
1236
+ "top_share_q10_50_90_types": [
1237
+ 0.222,
1238
+ 0.355,
1239
+ 0.599
1240
+ ],
1241
+ "top_share_q10_50_90_tokens": [
1242
+ 0.207,
1243
+ 0.355,
1244
+ 0.585
1245
+ ],
1246
+ "mean_top_types": 0.39,
1247
+ "mean_top_tokens": 0.377,
1248
+ "entropy_bits_q10_50_90_types": [
1249
+ 1.677,
1250
+ 2.819,
1251
+ 3.437
1252
+ ],
1253
+ "closability_q10_50_90_types": [
1254
+ 0.223,
1255
+ 0.509,
1256
+ 0.89
1257
+ ],
1258
+ "band_share_types": {
1259
+ "easy": 0.684,
1260
+ "mid": 0.175,
1261
+ "trap": 0.141,
1262
+ "novel": 0.0
1263
+ },
1264
+ "band_share_tokens": {
1265
+ "easy": 0.704,
1266
+ "mid": 0.172,
1267
+ "trap": 0.123,
1268
+ "novel": 0.0
1269
+ },
1270
+ "novel_word_ends_types": 1,
1271
+ "novel_word_ends_tokens": 25
1272
+ },
1273
+ "surprise_per_tokenizer_tokens": {
1274
+ "gpt2": {
1275
+ "dead_cell_share": 0.0016,
1276
+ "mean_surprisal_bits_witnessed_steps": 1.962,
1277
+ "unwitnessed_step_share": 0.0171
1278
+ },
1279
+ "smollm2": {
1280
+ "dead_cell_share": 0.0015,
1281
+ "mean_surprisal_bits_witnessed_steps": 1.983,
1282
+ "unwitnessed_step_share": 0.0188
1283
+ },
1284
+ "cl100k_base": {
1285
+ "dead_cell_share": 0.0009,
1286
+ "mean_surprisal_bits_witnessed_steps": 1.993,
1287
+ "unwitnessed_step_share": 0.0131
1288
+ },
1289
+ "o200k_base": {
1290
+ "dead_cell_share": 0.0003,
1291
+ "mean_surprisal_bits_witnessed_steps": 2.143,
1292
+ "unwitnessed_step_share": 0.0085
1293
+ },
1294
+ "llama3": {
1295
+ "dead_cell_share": 0.0009,
1296
+ "mean_surprisal_bits_witnessed_steps": 2.016,
1297
+ "unwitnessed_step_share": 0.013
1298
+ },
1299
+ "qwen25": {
1300
+ "dead_cell_share": 0.0009,
1301
+ "mean_surprisal_bits_witnessed_steps": 2.006,
1302
+ "unwitnessed_step_share": 0.0127
1303
+ },
1304
+ "deepseek-v3": {
1305
+ "dead_cell_share": 0.0006,
1306
+ "mean_surprisal_bits_witnessed_steps": 2.037,
1307
+ "unwitnessed_step_share": 0.0157
1308
+ },
1309
+ "mistral-v03": {
1310
+ "dead_cell_share": 0.0043,
1311
+ "mean_surprisal_bits_witnessed_steps": 1.761,
1312
+ "unwitnessed_step_share": 0.0626
1313
+ },
1314
+ "xlmr": {
1315
+ "dead_cell_share": 0.0018,
1316
+ "mean_surprisal_bits_witnessed_steps": 2.726,
1317
+ "unwitnessed_step_share": 0.0448
1318
+ },
1319
+ "t5-unigram": {
1320
+ "dead_cell_share": 0.0016,
1321
+ "mean_surprisal_bits_witnessed_steps": 2.012,
1322
+ "unwitnessed_step_share": 0.0212
1323
+ },
1324
+ "bert-wordpiece": {
1325
+ "dead_cell_share": 0.0011,
1326
+ "mean_surprisal_bits_witnessed_steps": 1.994,
1327
+ "unwitnessed_step_share": 0.0116
1328
+ }
1329
+ },
1330
+ "diet_tokens": {
1331
+ "cells": 242390,
1332
+ "under_floor_share": 0.0003,
1333
+ "never_seen_share": 0.0001,
1334
+ "mean_entropy_bits_known_cells": 1.959
1335
+ }
1336
+ }
1337
+ },
1338
+ "implanted_lexicon_hits": {}
1339
+ }