AbstractPhil commited on
Commit
8c2fbc7
·
verified ·
1 Parent(s): 59cd3d8

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

Browse files
mini-beatrix-3/caption_pack/lexicon_difficulty.json CHANGED
@@ -1,7 +1,7 @@
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",
@@ -463,550 +463,1425 @@
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
  }
@@ -1014,326 +1889,330 @@
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
  }
 
1
  {
2
+ "utc": "2026-10-06T04:13:34Z",
3
  "pack": "/root/pack_v2",
4
+ "rows_per_source": 30000,
5
  "seed": 20261005,
6
  "fleet": [
7
  "gpt2",
 
463
  }
464
  },
465
  "sources": {
466
+ "dff": {
467
+ "words_total": 6582510,
468
+ "words_distinct": 5298,
469
+ "hapax_share_of_types": 0.169,
470
+ "mean_word_len": 5.36,
471
+ "difficulty_fleet_consensus": {
472
+ "top_share_q10_50_90_types": [
473
+ 0.207,
474
+ 0.355,
475
+ 0.62
476
+ ],
477
+ "top_share_q10_50_90_tokens": [
478
+ 0.211,
479
+ 0.351,
480
+ 0.552
481
+ ],
482
+ "mean_top_types": 0.39,
483
+ "mean_top_tokens": 0.37,
484
+ "entropy_bits_q10_50_90_types": [
485
+ 1.665,
486
+ 2.808,
487
+ 3.457
488
+ ],
489
+ "closability_q10_50_90_types": [
490
+ 0.209,
491
+ 0.52,
492
+ 0.916
493
+ ],
494
+ "band_share_types": {
495
+ "easy": 0.673,
496
+ "mid": 0.17,
497
+ "trap": 0.156,
498
+ "novel": 0.001
499
+ },
500
+ "band_share_tokens": {
501
+ "easy": 0.711,
502
+ "mid": 0.184,
503
+ "trap": 0.105,
504
+ "novel": 0.0
505
+ },
506
+ "novel_word_ends_types": 5,
507
+ "novel_word_ends_tokens": 7
508
+ },
509
+ "surprise_per_tokenizer_tokens": {
510
+ "gpt2": {
511
+ "dead_cell_share": 0.0018,
512
+ "mean_surprisal_bits_witnessed_steps": 1.975,
513
+ "unwitnessed_step_share": 0.0186
514
+ },
515
+ "smollm2": {
516
+ "dead_cell_share": 0.0019,
517
+ "mean_surprisal_bits_witnessed_steps": 2.021,
518
+ "unwitnessed_step_share": 0.0203
519
+ },
520
+ "cl100k_base": {
521
+ "dead_cell_share": 0.0008,
522
+ "mean_surprisal_bits_witnessed_steps": 2.026,
523
+ "unwitnessed_step_share": 0.0132
524
+ },
525
+ "o200k_base": {
526
+ "dead_cell_share": 0.0004,
527
+ "mean_surprisal_bits_witnessed_steps": 2.173,
528
+ "unwitnessed_step_share": 0.0091
529
+ },
530
+ "llama3": {
531
+ "dead_cell_share": 0.0007,
532
+ "mean_surprisal_bits_witnessed_steps": 2.042,
533
+ "unwitnessed_step_share": 0.013
534
+ },
535
+ "qwen25": {
536
+ "dead_cell_share": 0.0008,
537
+ "mean_surprisal_bits_witnessed_steps": 2.035,
538
+ "unwitnessed_step_share": 0.013
539
+ },
540
+ "deepseek-v3": {
541
+ "dead_cell_share": 0.0011,
542
+ "mean_surprisal_bits_witnessed_steps": 2.066,
543
+ "unwitnessed_step_share": 0.016
544
+ },
545
+ "mistral-v03": {
546
+ "dead_cell_share": 0.0087,
547
+ "mean_surprisal_bits_witnessed_steps": 1.794,
548
+ "unwitnessed_step_share": 0.0671
549
+ },
550
+ "xlmr": {
551
+ "dead_cell_share": 0.0023,
552
+ "mean_surprisal_bits_witnessed_steps": 2.751,
553
+ "unwitnessed_step_share": 0.0585
554
+ },
555
+ "t5-unigram": {
556
+ "dead_cell_share": 0.0018,
557
+ "mean_surprisal_bits_witnessed_steps": 2.012,
558
+ "unwitnessed_step_share": 0.0257
559
+ },
560
+ "bert-wordpiece": {
561
+ "dead_cell_share": 0.001,
562
+ "mean_surprisal_bits_witnessed_steps": 2.023,
563
+ "unwitnessed_step_share": 0.0158
564
+ }
565
+ },
566
+ "diet_tokens": {
567
+ "cells": 22140311,
568
+ "under_floor_share": 0.0007,
569
+ "never_seen_share": 0.0,
570
+ "mean_entropy_bits_known_cells": 1.968
571
+ },
572
+ "rows_sampled": 30274,
573
+ "rows_total": 121732,
574
+ "bytes_per_row": 1517.3,
575
+ "words_per_row": 217.4,
576
+ "registers": {
577
+ "tags": {
578
+ "words_total": 937621,
579
+ "words_distinct": 1189,
580
+ "hapax_share_of_types": 0.189,
581
+ "mean_word_len": 5.4,
582
+ "difficulty_fleet_consensus": {
583
+ "top_share_q10_50_90_types": [
584
+ 0.21,
585
+ 0.351,
586
+ 0.59
587
+ ],
588
+ "top_share_q10_50_90_tokens": [
589
+ 0.224,
590
+ 0.355,
591
+ 0.615
592
+ ],
593
+ "mean_top_types": 0.377,
594
+ "mean_top_tokens": 0.385,
595
+ "entropy_bits_q10_50_90_types": [
596
+ 1.837,
597
+ 2.868,
598
+ 3.479
599
+ ],
600
+ "closability_q10_50_90_types": [
601
+ 0.161,
602
+ 0.423,
603
+ 0.86
604
+ ],
605
+ "band_share_types": {
606
+ "easy": 0.704,
607
+ "mid": 0.16,
608
+ "trap": 0.136,
609
+ "novel": 0.0
610
+ },
611
+ "band_share_tokens": {
612
+ "easy": 0.716,
613
+ "mid": 0.16,
614
+ "trap": 0.124,
615
+ "novel": 0.0
616
+ },
617
+ "novel_word_ends_types": 0,
618
+ "novel_word_ends_tokens": 0
619
+ },
620
+ "surprise_per_tokenizer_tokens": {
621
+ "gpt2": {
622
+ "dead_cell_share": 0.0021,
623
+ "mean_surprisal_bits_witnessed_steps": 2.054,
624
+ "unwitnessed_step_share": 0.0164
625
+ },
626
+ "smollm2": {
627
+ "dead_cell_share": 0.0027,
628
+ "mean_surprisal_bits_witnessed_steps": 2.124,
629
+ "unwitnessed_step_share": 0.0209
630
+ },
631
+ "cl100k_base": {
632
+ "dead_cell_share": 0.0018,
633
+ "mean_surprisal_bits_witnessed_steps": 2.148,
634
+ "unwitnessed_step_share": 0.016
635
+ },
636
+ "o200k_base": {
637
+ "dead_cell_share": 0.0008,
638
+ "mean_surprisal_bits_witnessed_steps": 2.303,
639
+ "unwitnessed_step_share": 0.0083
640
+ },
641
+ "llama3": {
642
+ "dead_cell_share": 0.0018,
643
+ "mean_surprisal_bits_witnessed_steps": 2.169,
644
+ "unwitnessed_step_share": 0.0159
645
+ },
646
+ "qwen25": {
647
+ "dead_cell_share": 0.0018,
648
+ "mean_surprisal_bits_witnessed_steps": 2.157,
649
+ "unwitnessed_step_share": 0.0157
650
+ },
651
+ "deepseek-v3": {
652
+ "dead_cell_share": 0.0023,
653
+ "mean_surprisal_bits_witnessed_steps": 2.205,
654
+ "unwitnessed_step_share": 0.0166
655
+ },
656
+ "mistral-v03": {
657
+ "dead_cell_share": 0.0089,
658
+ "mean_surprisal_bits_witnessed_steps": 1.881,
659
+ "unwitnessed_step_share": 0.0782
660
+ },
661
+ "xlmr": {
662
+ "dead_cell_share": 0.0009,
663
+ "mean_surprisal_bits_witnessed_steps": 2.888,
664
+ "unwitnessed_step_share": 0.0669
665
+ },
666
+ "t5-unigram": {
667
+ "dead_cell_share": 0.0022,
668
+ "mean_surprisal_bits_witnessed_steps": 2.109,
669
+ "unwitnessed_step_share": 0.0232
670
+ },
671
+ "bert-wordpiece": {
672
+ "dead_cell_share": 0.0017,
673
+ "mean_surprisal_bits_witnessed_steps": 2.045,
674
+ "unwitnessed_step_share": 0.0319
675
+ }
676
+ },
677
+ "diet_tokens": {
678
+ "cells": 3188381,
679
+ "under_floor_share": 0.0012,
680
+ "never_seen_share": 0.0,
681
+ "mean_entropy_bits_known_cells": 2.031
682
+ }
683
+ },
684
+ "english": {
685
+ "words_total": 4155139,
686
+ "words_distinct": 4958,
687
+ "hapax_share_of_types": 0.188,
688
+ "mean_word_len": 5.3,
689
+ "difficulty_fleet_consensus": {
690
+ "top_share_q10_50_90_types": [
691
+ 0.207,
692
+ 0.355,
693
+ 0.615
694
+ ],
695
+ "top_share_q10_50_90_tokens": [
696
+ 0.19,
697
+ 0.341,
698
+ 0.552
699
+ ],
700
+ "mean_top_types": 0.39,
701
+ "mean_top_tokens": 0.366,
702
+ "entropy_bits_q10_50_90_types": [
703
+ 1.665,
704
+ 2.808,
705
+ 3.457
706
+ ],
707
+ "closability_q10_50_90_types": [
708
+ 0.219,
709
+ 0.545,
710
+ 0.918
711
+ ],
712
+ "band_share_types": {
713
+ "easy": 0.674,
714
+ "mid": 0.171,
715
+ "trap": 0.154,
716
+ "novel": 0.001
717
+ },
718
+ "band_share_tokens": {
719
+ "easy": 0.712,
720
+ "mid": 0.187,
721
+ "trap": 0.102,
722
+ "novel": 0.0
723
+ },
724
+ "novel_word_ends_types": 5,
725
+ "novel_word_ends_tokens": 5
726
+ },
727
+ "surprise_per_tokenizer_tokens": {
728
+ "gpt2": {
729
+ "dead_cell_share": 0.0018,
730
+ "mean_surprisal_bits_witnessed_steps": 1.958,
731
+ "unwitnessed_step_share": 0.0189
732
+ },
733
+ "smollm2": {
734
+ "dead_cell_share": 0.0019,
735
+ "mean_surprisal_bits_witnessed_steps": 1.996,
736
+ "unwitnessed_step_share": 0.0205
737
+ },
738
+ "cl100k_base": {
739
+ "dead_cell_share": 0.0008,
740
+ "mean_surprisal_bits_witnessed_steps": 2.004,
741
+ "unwitnessed_step_share": 0.0127
742
+ },
743
+ "o200k_base": {
744
+ "dead_cell_share": 0.0003,
745
+ "mean_surprisal_bits_witnessed_steps": 2.143,
746
+ "unwitnessed_step_share": 0.0092
747
+ },
748
+ "llama3": {
749
+ "dead_cell_share": 0.0005,
750
+ "mean_surprisal_bits_witnessed_steps": 2.02,
751
+ "unwitnessed_step_share": 0.0124
752
+ },
753
+ "qwen25": {
754
+ "dead_cell_share": 0.0008,
755
+ "mean_surprisal_bits_witnessed_steps": 2.013,
756
+ "unwitnessed_step_share": 0.0126
757
+ },
758
+ "deepseek-v3": {
759
+ "dead_cell_share": 0.001,
760
+ "mean_surprisal_bits_witnessed_steps": 2.035,
761
+ "unwitnessed_step_share": 0.0158
762
+ },
763
+ "mistral-v03": {
764
+ "dead_cell_share": 0.0096,
765
+ "mean_surprisal_bits_witnessed_steps": 1.774,
766
+ "unwitnessed_step_share": 0.0664
767
+ },
768
+ "xlmr": {
769
+ "dead_cell_share": 0.0029,
770
+ "mean_surprisal_bits_witnessed_steps": 2.726,
771
+ "unwitnessed_step_share": 0.0571
772
+ },
773
+ "t5-unigram": {
774
+ "dead_cell_share": 0.0019,
775
+ "mean_surprisal_bits_witnessed_steps": 1.986,
776
+ "unwitnessed_step_share": 0.0265
777
+ },
778
+ "bert-wordpiece": {
779
+ "dead_cell_share": 0.001,
780
+ "mean_surprisal_bits_witnessed_steps": 2.02,
781
+ "unwitnessed_step_share": 0.013
782
+ }
783
+ },
784
+ "diet_tokens": {
785
+ "cells": 13730230,
786
+ "under_floor_share": 0.0006,
787
+ "never_seen_share": 0.0,
788
+ "mean_entropy_bits_known_cells": 1.927
789
+ }
790
+ }
791
+ }
792
+ },
793
+ "scf": {
794
+ "words_total": 8481824,
795
+ "words_distinct": 6363,
796
+ "hapax_share_of_types": 0.2,
797
+ "mean_word_len": 5.44,
798
+ "difficulty_fleet_consensus": {
799
+ "top_share_q10_50_90_types": [
800
+ 0.208,
801
+ 0.355,
802
+ 0.597
803
+ ],
804
+ "top_share_q10_50_90_tokens": [
805
+ 0.199,
806
+ 0.352,
807
+ 0.552
808
+ ],
809
+ "mean_top_types": 0.386,
810
+ "mean_top_tokens": 0.368,
811
+ "entropy_bits_q10_50_90_types": [
812
+ 1.741,
813
+ 2.814,
814
+ 3.458
815
+ ],
816
+ "closability_q10_50_90_types": [
817
+ 0.213,
818
+ 0.514,
819
+ 0.912
820
+ ],
821
+ "band_share_types": {
822
+ "easy": 0.676,
823
+ "mid": 0.177,
824
+ "trap": 0.146,
825
+ "novel": 0.001
826
+ },
827
+ "band_share_tokens": {
828
+ "easy": 0.708,
829
+ "mid": 0.182,
830
+ "trap": 0.109,
831
+ "novel": 0.001
832
+ },
833
+ "novel_word_ends_types": 4,
834
+ "novel_word_ends_tokens": 4987
835
+ },
836
+ "surprise_per_tokenizer_tokens": {
837
+ "gpt2": {
838
+ "dead_cell_share": 0.0015,
839
+ "mean_surprisal_bits_witnessed_steps": 2.016,
840
+ "unwitnessed_step_share": 0.017
841
+ },
842
+ "smollm2": {
843
+ "dead_cell_share": 0.0014,
844
+ "mean_surprisal_bits_witnessed_steps": 2.048,
845
+ "unwitnessed_step_share": 0.0195
846
+ },
847
+ "cl100k_base": {
848
+ "dead_cell_share": 0.0008,
849
+ "mean_surprisal_bits_witnessed_steps": 2.056,
850
+ "unwitnessed_step_share": 0.0125
851
+ },
852
+ "o200k_base": {
853
+ "dead_cell_share": 0.0002,
854
+ "mean_surprisal_bits_witnessed_steps": 2.206,
855
+ "unwitnessed_step_share": 0.0085
856
+ },
857
+ "llama3": {
858
+ "dead_cell_share": 0.0007,
859
+ "mean_surprisal_bits_witnessed_steps": 2.08,
860
+ "unwitnessed_step_share": 0.0124
861
+ },
862
+ "qwen25": {
863
+ "dead_cell_share": 0.0008,
864
+ "mean_surprisal_bits_witnessed_steps": 2.069,
865
+ "unwitnessed_step_share": 0.0121
866
+ },
867
+ "deepseek-v3": {
868
+ "dead_cell_share": 0.0005,
869
+ "mean_surprisal_bits_witnessed_steps": 2.097,
870
+ "unwitnessed_step_share": 0.0158
871
+ },
872
+ "mistral-v03": {
873
+ "dead_cell_share": 0.0042,
874
+ "mean_surprisal_bits_witnessed_steps": 1.824,
875
+ "unwitnessed_step_share": 0.0661
876
+ },
877
+ "xlmr": {
878
+ "dead_cell_share": 0.0016,
879
+ "mean_surprisal_bits_witnessed_steps": 2.744,
880
+ "unwitnessed_step_share": 0.0457
881
+ },
882
+ "t5-unigram": {
883
+ "dead_cell_share": 0.0016,
884
+ "mean_surprisal_bits_witnessed_steps": 2.056,
885
+ "unwitnessed_step_share": 0.0211
886
+ },
887
+ "bert-wordpiece": {
888
+ "dead_cell_share": 0.001,
889
+ "mean_surprisal_bits_witnessed_steps": 2.04,
890
+ "unwitnessed_step_share": 0.0107
891
+ }
892
+ },
893
+ "diet_tokens": {
894
+ "cells": 29161727,
895
+ "under_floor_share": 0.0003,
896
+ "never_seen_share": 0.0002,
897
+ "mean_entropy_bits_known_cells": 1.982
898
+ },
899
+ "rows_sampled": 42080,
900
+ "rows_total": 42257,
901
+ "bytes_per_row": 1437.5,
902
+ "words_per_row": 201.6,
903
+ "registers": {
904
+ "tags": {
905
+ "words_total": 1257465,
906
+ "words_distinct": 1218,
907
+ "hapax_share_of_types": 0.214,
908
+ "mean_word_len": 5.62,
909
+ "difficulty_fleet_consensus": {
910
+ "top_share_q10_50_90_types": [
911
+ 0.211,
912
+ 0.355,
913
+ 0.574
914
+ ],
915
+ "top_share_q10_50_90_tokens": [
916
+ 0.219,
917
+ 0.355,
918
+ 0.522
919
+ ],
920
+ "mean_top_types": 0.378,
921
+ "mean_top_tokens": 0.372,
922
+ "entropy_bits_q10_50_90_types": [
923
+ 1.849,
924
+ 2.838,
925
+ 3.458
926
+ ],
927
+ "closability_q10_50_90_types": [
928
+ 0.171,
929
+ 0.424,
930
+ 0.86
931
+ ],
932
+ "band_share_types": {
933
+ "easy": 0.699,
934
+ "mid": 0.172,
935
+ "trap": 0.13,
936
+ "novel": 0.0
937
+ },
938
+ "band_share_tokens": {
939
+ "easy": 0.701,
940
+ "mid": 0.208,
941
+ "trap": 0.091,
942
+ "novel": 0.0
943
+ },
944
+ "novel_word_ends_types": 0,
945
+ "novel_word_ends_tokens": 0
946
+ },
947
+ "surprise_per_tokenizer_tokens": {
948
+ "gpt2": {
949
+ "dead_cell_share": 0.0022,
950
+ "mean_surprisal_bits_witnessed_steps": 2.078,
951
+ "unwitnessed_step_share": 0.0219
952
+ },
953
+ "smollm2": {
954
+ "dead_cell_share": 0.0023,
955
+ "mean_surprisal_bits_witnessed_steps": 2.102,
956
+ "unwitnessed_step_share": 0.0262
957
+ },
958
+ "cl100k_base": {
959
+ "dead_cell_share": 0.0007,
960
+ "mean_surprisal_bits_witnessed_steps": 2.16,
961
+ "unwitnessed_step_share": 0.0173
962
+ },
963
+ "o200k_base": {
964
+ "dead_cell_share": 0.0,
965
+ "mean_surprisal_bits_witnessed_steps": 2.267,
966
+ "unwitnessed_step_share": 0.0129
967
+ },
968
+ "llama3": {
969
+ "dead_cell_share": 0.0007,
970
+ "mean_surprisal_bits_witnessed_steps": 2.195,
971
+ "unwitnessed_step_share": 0.0173
972
+ },
973
+ "qwen25": {
974
+ "dead_cell_share": 0.0007,
975
+ "mean_surprisal_bits_witnessed_steps": 2.176,
976
+ "unwitnessed_step_share": 0.0165
977
+ },
978
+ "deepseek-v3": {
979
+ "dead_cell_share": 0.0002,
980
+ "mean_surprisal_bits_witnessed_steps": 2.175,
981
+ "unwitnessed_step_share": 0.0212
982
+ },
983
+ "mistral-v03": {
984
+ "dead_cell_share": 0.0063,
985
+ "mean_surprisal_bits_witnessed_steps": 1.884,
986
+ "unwitnessed_step_share": 0.0834
987
+ },
988
+ "xlmr": {
989
+ "dead_cell_share": 0.0003,
990
+ "mean_surprisal_bits_witnessed_steps": 2.79,
991
+ "unwitnessed_step_share": 0.0488
992
+ },
993
+ "t5-unigram": {
994
+ "dead_cell_share": 0.0031,
995
+ "mean_surprisal_bits_witnessed_steps": 2.107,
996
+ "unwitnessed_step_share": 0.0225
997
+ },
998
+ "bert-wordpiece": {
999
+ "dead_cell_share": 0.0028,
1000
+ "mean_surprisal_bits_witnessed_steps": 2.062,
1001
+ "unwitnessed_step_share": 0.0218
1002
+ }
1003
+ },
1004
+ "diet_tokens": {
1005
+ "cells": 4556370,
1006
+ "under_floor_share": 0.0001,
1007
+ "never_seen_share": 0.0,
1008
+ "mean_entropy_bits_known_cells": 2.044
1009
+ }
1010
+ },
1011
+ "english": {
1012
+ "words_total": 4858976,
1013
+ "words_distinct": 6111,
1014
+ "hapax_share_of_types": 0.235,
1015
+ "mean_word_len": 5.34,
1016
+ "difficulty_fleet_consensus": {
1017
+ "top_share_q10_50_90_types": [
1018
+ 0.208,
1019
+ 0.355,
1020
+ 0.597
1021
+ ],
1022
+ "top_share_q10_50_90_tokens": [
1023
+ 0.19,
1024
+ 0.343,
1025
+ 0.559
1026
+ ],
1027
+ "mean_top_types": 0.386,
1028
+ "mean_top_tokens": 0.364,
1029
+ "entropy_bits_q10_50_90_types": [
1030
+ 1.726,
1031
+ 2.814,
1032
+ 3.459
1033
+ ],
1034
+ "closability_q10_50_90_types": [
1035
+ 0.218,
1036
+ 0.521,
1037
+ 0.914
1038
+ ],
1039
+ "band_share_types": {
1040
+ "easy": 0.677,
1041
+ "mid": 0.177,
1042
+ "trap": 0.145,
1043
+ "novel": 0.001
1044
+ },
1045
+ "band_share_tokens": {
1046
+ "easy": 0.714,
1047
+ "mid": 0.171,
1048
+ "trap": 0.114,
1049
+ "novel": 0.001
1050
+ },
1051
+ "novel_word_ends_types": 4,
1052
+ "novel_word_ends_tokens": 4984
1053
+ },
1054
+ "surprise_per_tokenizer_tokens": {
1055
+ "gpt2": {
1056
+ "dead_cell_share": 0.0015,
1057
+ "mean_surprisal_bits_witnessed_steps": 1.983,
1058
+ "unwitnessed_step_share": 0.0159
1059
+ },
1060
+ "smollm2": {
1061
+ "dead_cell_share": 0.0014,
1062
+ "mean_surprisal_bits_witnessed_steps": 2.016,
1063
+ "unwitnessed_step_share": 0.0178
1064
+ },
1065
+ "cl100k_base": {
1066
+ "dead_cell_share": 0.001,
1067
+ "mean_surprisal_bits_witnessed_steps": 2.019,
1068
+ "unwitnessed_step_share": 0.0115
1069
+ },
1070
+ "o200k_base": {
1071
+ "dead_cell_share": 0.0004,
1072
+ "mean_surprisal_bits_witnessed_steps": 2.167,
1073
+ "unwitnessed_step_share": 0.0079
1074
+ },
1075
+ "llama3": {
1076
+ "dead_cell_share": 0.0008,
1077
+ "mean_surprisal_bits_witnessed_steps": 2.04,
1078
+ "unwitnessed_step_share": 0.0114
1079
+ },
1080
+ "qwen25": {
1081
+ "dead_cell_share": 0.001,
1082
+ "mean_surprisal_bits_witnessed_steps": 2.031,
1083
+ "unwitnessed_step_share": 0.0112
1084
+ },
1085
+ "deepseek-v3": {
1086
+ "dead_cell_share": 0.0008,
1087
+ "mean_surprisal_bits_witnessed_steps": 2.058,
1088
+ "unwitnessed_step_share": 0.0145
1089
+ },
1090
+ "mistral-v03": {
1091
+ "dead_cell_share": 0.0043,
1092
+ "mean_surprisal_bits_witnessed_steps": 1.8,
1093
+ "unwitnessed_step_share": 0.0587
1094
+ },
1095
+ "xlmr": {
1096
+ "dead_cell_share": 0.0019,
1097
+ "mean_surprisal_bits_witnessed_steps": 2.714,
1098
+ "unwitnessed_step_share": 0.045
1099
+ },
1100
+ "t5-unigram": {
1101
+ "dead_cell_share": 0.0017,
1102
+ "mean_surprisal_bits_witnessed_steps": 2.017,
1103
+ "unwitnessed_step_share": 0.0215
1104
+ },
1105
+ "bert-wordpiece": {
1106
+ "dead_cell_share": 0.0008,
1107
+ "mean_surprisal_bits_witnessed_steps": 2.015,
1108
+ "unwitnessed_step_share": 0.009
1109
+ }
1110
+ },
1111
+ "diet_tokens": {
1112
+ "cells": 16235587,
1113
+ "under_floor_share": 0.0004,
1114
+ "never_seen_share": 0.0003,
1115
+ "mean_entropy_bits_known_cells": 1.922
1116
+ }
1117
+ }
1118
+ }
1119
+ },
1120
+ "cc12m": {
1121
+ "words_total": 2543745,
1122
+ "words_distinct": 35439,
1123
+ "hapax_share_of_types": 0.276,
1124
+ "mean_word_len": 5.56,
1125
+ "difficulty_fleet_consensus": {
1126
+ "top_share_q10_50_90_types": [
1127
+ 0.207,
1128
+ 0.355,
1129
+ 0.631
1130
+ ],
1131
+ "top_share_q10_50_90_tokens": [
1132
+ 0.19,
1133
+ 0.351,
1134
+ 0.552
1135
+ ],
1136
+ "mean_top_types": 0.388,
1137
+ "mean_top_tokens": 0.364,
1138
+ "entropy_bits_q10_50_90_types": [
1139
+ 1.579,
1140
+ 2.805,
1141
+ 3.462
1142
+ ],
1143
+ "closability_q10_50_90_types": [
1144
+ 0.151,
1145
+ 0.466,
1146
+ 0.905
1147
+ ],
1148
+ "band_share_types": {
1149
+ "easy": 0.664,
1150
+ "mid": 0.171,
1151
+ "trap": 0.157,
1152
+ "novel": 0.007
1153
+ },
1154
+ "band_share_tokens": {
1155
+ "easy": 0.691,
1156
+ "mid": 0.208,
1157
+ "trap": 0.101,
1158
+ "novel": 0.0
1159
+ },
1160
+ "novel_word_ends_types": 252,
1161
+ "novel_word_ends_tokens": 587
1162
+ },
1163
+ "surprise_per_tokenizer_tokens": {
1164
+ "gpt2": {
1165
+ "dead_cell_share": 0.003,
1166
+ "mean_surprisal_bits_witnessed_steps": 1.945,
1167
+ "unwitnessed_step_share": 0.0212
1168
+ },
1169
+ "smollm2": {
1170
+ "dead_cell_share": 0.0028,
1171
+ "mean_surprisal_bits_witnessed_steps": 1.959,
1172
+ "unwitnessed_step_share": 0.0185
1173
+ },
1174
+ "cl100k_base": {
1175
+ "dead_cell_share": 0.0016,
1176
+ "mean_surprisal_bits_witnessed_steps": 1.985,
1177
+ "unwitnessed_step_share": 0.0162
1178
+ },
1179
+ "o200k_base": {
1180
+ "dead_cell_share": 0.0007,
1181
+ "mean_surprisal_bits_witnessed_steps": 2.119,
1182
+ "unwitnessed_step_share": 0.0095
1183
+ },
1184
+ "llama3": {
1185
+ "dead_cell_share": 0.0014,
1186
+ "mean_surprisal_bits_witnessed_steps": 2.001,
1187
+ "unwitnessed_step_share": 0.0157
1188
+ },
1189
+ "qwen25": {
1190
+ "dead_cell_share": 0.0016,
1191
+ "mean_surprisal_bits_witnessed_steps": 1.995,
1192
+ "unwitnessed_step_share": 0.016
1193
+ },
1194
+ "deepseek-v3": {
1195
+ "dead_cell_share": 0.0017,
1196
+ "mean_surprisal_bits_witnessed_steps": 2.008,
1197
+ "unwitnessed_step_share": 0.0156
1198
+ },
1199
+ "mistral-v03": {
1200
+ "dead_cell_share": 0.0101,
1201
+ "mean_surprisal_bits_witnessed_steps": 1.783,
1202
+ "unwitnessed_step_share": 0.0664
1203
+ },
1204
+ "xlmr": {
1205
+ "dead_cell_share": 0.0029,
1206
+ "mean_surprisal_bits_witnessed_steps": 2.684,
1207
+ "unwitnessed_step_share": 0.0444
1208
+ },
1209
+ "t5-unigram": {
1210
+ "dead_cell_share": 0.0032,
1211
+ "mean_surprisal_bits_witnessed_steps": 1.981,
1212
+ "unwitnessed_step_share": 0.0249
1213
+ },
1214
+ "bert-wordpiece": {
1215
+ "dead_cell_share": 0.0019,
1216
+ "mean_surprisal_bits_witnessed_steps": 2.0,
1217
+ "unwitnessed_step_share": 0.0154
1218
+ }
1219
+ },
1220
+ "diet_tokens": {
1221
+ "cells": 9061436,
1222
+ "under_floor_share": 0.0012,
1223
+ "never_seen_share": 0.0003,
1224
+ "mean_entropy_bits_known_cells": 1.914
1225
+ },
1226
+ "rows_sampled": 27604,
1227
+ "rows_total": 1362918,
1228
+ "bytes_per_row": 693.7,
1229
+ "words_per_row": 92.2,
1230
+ "registers": {
1231
+ "tags": null,
1232
+ "english": {
1233
+ "words_total": 2460933,
1234
+ "words_distinct": 35439,
1235
+ "hapax_share_of_types": 0.276,
1236
+ "mean_word_len": 5.57,
1237
+ "difficulty_fleet_consensus": {
1238
+ "top_share_q10_50_90_types": [
1239
+ 0.207,
1240
+ 0.355,
1241
+ 0.631
1242
+ ],
1243
+ "top_share_q10_50_90_tokens": [
1244
+ 0.19,
1245
+ 0.352,
1246
+ 0.536
1247
+ ],
1248
+ "mean_top_types": 0.388,
1249
+ "mean_top_tokens": 0.366,
1250
+ "entropy_bits_q10_50_90_types": [
1251
+ 1.579,
1252
+ 2.805,
1253
+ 3.462
1254
+ ],
1255
+ "closability_q10_50_90_types": [
1256
+ 0.151,
1257
+ 0.466,
1258
+ 0.905
1259
+ ],
1260
+ "band_share_types": {
1261
+ "easy": 0.664,
1262
+ "mid": 0.171,
1263
+ "trap": 0.157,
1264
+ "novel": 0.007
1265
+ },
1266
+ "band_share_tokens": {
1267
+ "easy": 0.692,
1268
+ "mid": 0.215,
1269
+ "trap": 0.093,
1270
+ "novel": 0.0
1271
+ },
1272
+ "novel_word_ends_types": 252,
1273
+ "novel_word_ends_tokens": 587
1274
+ },
1275
+ "surprise_per_tokenizer_tokens": {
1276
+ "gpt2": {
1277
+ "dead_cell_share": 0.0031,
1278
+ "mean_surprisal_bits_witnessed_steps": 1.96,
1279
+ "unwitnessed_step_share": 0.0219
1280
+ },
1281
+ "smollm2": {
1282
+ "dead_cell_share": 0.0029,
1283
+ "mean_surprisal_bits_witnessed_steps": 1.978,
1284
+ "unwitnessed_step_share": 0.019
1285
+ },
1286
+ "cl100k_base": {
1287
+ "dead_cell_share": 0.0017,
1288
+ "mean_surprisal_bits_witnessed_steps": 2.006,
1289
+ "unwitnessed_step_share": 0.0167
1290
+ },
1291
+ "o200k_base": {
1292
+ "dead_cell_share": 0.0008,
1293
+ "mean_surprisal_bits_witnessed_steps": 2.141,
1294
+ "unwitnessed_step_share": 0.0097
1295
+ },
1296
+ "llama3": {
1297
+ "dead_cell_share": 0.0014,
1298
+ "mean_surprisal_bits_witnessed_steps": 2.022,
1299
+ "unwitnessed_step_share": 0.0162
1300
+ },
1301
+ "qwen25": {
1302
+ "dead_cell_share": 0.0016,
1303
+ "mean_surprisal_bits_witnessed_steps": 2.015,
1304
+ "unwitnessed_step_share": 0.0165
1305
+ },
1306
+ "deepseek-v3": {
1307
+ "dead_cell_share": 0.0017,
1308
+ "mean_surprisal_bits_witnessed_steps": 2.026,
1309
+ "unwitnessed_step_share": 0.0161
1310
+ },
1311
+ "mistral-v03": {
1312
+ "dead_cell_share": 0.0104,
1313
+ "mean_surprisal_bits_witnessed_steps": 1.791,
1314
+ "unwitnessed_step_share": 0.0684
1315
+ },
1316
+ "xlmr": {
1317
+ "dead_cell_share": 0.003,
1318
+ "mean_surprisal_bits_witnessed_steps": 2.698,
1319
+ "unwitnessed_step_share": 0.0458
1320
+ },
1321
+ "t5-unigram": {
1322
+ "dead_cell_share": 0.0033,
1323
+ "mean_surprisal_bits_witnessed_steps": 1.997,
1324
+ "unwitnessed_step_share": 0.0256
1325
+ },
1326
+ "bert-wordpiece": {
1327
+ "dead_cell_share": 0.002,
1328
+ "mean_surprisal_bits_witnessed_steps": 2.016,
1329
+ "unwitnessed_step_share": 0.0159
1330
+ }
1331
+ },
1332
+ "diet_tokens": {
1333
+ "cells": 8785396,
1334
+ "under_floor_share": 0.0012,
1335
+ "never_seen_share": 0.0003,
1336
+ "mean_entropy_bits_known_cells": 1.903
1337
+ }
1338
+ }
1339
+ }
1340
+ },
1341
  "coco": {
1342
+ "words_total": 508840,
1343
+ "words_distinct": 5941,
1344
+ "hapax_share_of_types": 0.027,
1345
+ "mean_word_len": 6.11,
1346
  "difficulty_fleet_consensus": {
1347
  "top_share_q10_50_90_types": [
1348
+ 0.211,
1349
  0.355,
1350
+ 0.609
1351
  ],
1352
  "top_share_q10_50_90_tokens": [
1353
  0.23,
1354
  0.371,
1355
  0.797
1356
  ],
1357
+ "mean_top_types": 0.386,
1358
+ "mean_top_tokens": 0.441,
1359
  "entropy_bits_q10_50_90_types": [
1360
+ 1.667,
1361
+ 2.851,
1362
+ 3.444
1363
  ],
1364
  "closability_q10_50_90_types": [
1365
+ 0.23,
1366
+ 0.558,
1367
+ 0.905
1368
  ],
1369
  "band_share_types": {
1370
+ "easy": 0.684,
1371
+ "mid": 0.173,
1372
+ "trap": 0.143,
1373
+ "novel": 0.001
1374
  },
1375
  "band_share_tokens": {
1376
+ "easy": 0.652,
1377
+ "mid": 0.119,
1378
+ "trap": 0.23,
1379
  "novel": 0.0
1380
  },
1381
+ "novel_word_ends_types": 5,
1382
+ "novel_word_ends_tokens": 10
1383
  },
1384
  "surprise_per_tokenizer_tokens": {
1385
  "gpt2": {
1386
+ "dead_cell_share": 0.0014,
1387
+ "mean_surprisal_bits_witnessed_steps": 1.601,
1388
+ "unwitnessed_step_share": 0.0109
1389
  },
1390
  "smollm2": {
1391
+ "dead_cell_share": 0.0014,
1392
+ "mean_surprisal_bits_witnessed_steps": 1.578,
1393
+ "unwitnessed_step_share": 0.0087
1394
  },
1395
  "cl100k_base": {
1396
+ "dead_cell_share": 0.0015,
1397
+ "mean_surprisal_bits_witnessed_steps": 1.594,
1398
+ "unwitnessed_step_share": 0.0091
1399
  },
1400
  "o200k_base": {
1401
+ "dead_cell_share": 0.0008,
1402
+ "mean_surprisal_bits_witnessed_steps": 1.733,
1403
+ "unwitnessed_step_share": 0.0045
1404
  },
1405
  "llama3": {
1406
+ "dead_cell_share": 0.0014,
1407
+ "mean_surprisal_bits_witnessed_steps": 1.605,
1408
+ "unwitnessed_step_share": 0.009
1409
  },
1410
  "qwen25": {
1411
+ "dead_cell_share": 0.0014,
1412
+ "mean_surprisal_bits_witnessed_steps": 1.602,
1413
+ "unwitnessed_step_share": 0.0091
1414
  },
1415
  "deepseek-v3": {
1416
+ "dead_cell_share": 0.001,
1417
+ "mean_surprisal_bits_witnessed_steps": 1.658,
1418
+ "unwitnessed_step_share": 0.0085
1419
  },
1420
  "mistral-v03": {
1421
+ "dead_cell_share": 0.0049,
1422
+ "mean_surprisal_bits_witnessed_steps": 1.379,
1423
+ "unwitnessed_step_share": 0.0325
1424
  },
1425
  "xlmr": {
1426
+ "dead_cell_share": 0.0042,
1427
+ "mean_surprisal_bits_witnessed_steps": 2.557,
1428
+ "unwitnessed_step_share": 0.0336
1429
  },
1430
  "t5-unigram": {
1431
+ "dead_cell_share": 0.0016,
1432
  "mean_surprisal_bits_witnessed_steps": 1.733,
1433
+ "unwitnessed_step_share": 0.0129
1434
  },
1435
  "bert-wordpiece": {
1436
+ "dead_cell_share": 0.0011,
1437
+ "mean_surprisal_bits_witnessed_steps": 1.685,
1438
+ "unwitnessed_step_share": 0.0093
1439
  }
1440
  },
1441
  "diet_tokens": {
1442
+ "cells": 2092837,
1443
+ "under_floor_share": 0.0009,
1444
+ "never_seen_share": 0.0,
1445
+ "mean_entropy_bits_known_cells": 1.794
1446
  },
1447
+ "rows_sampled": 19227,
1448
  "rows_total": 19341,
1449
+ "bytes_per_row": 248.9,
1450
  "words_per_row": 26.5,
1451
  "registers": {
1452
  "tags": null,
1453
  "english": {
1454
+ "words_total": 470386,
1455
+ "words_distinct": 5941,
1456
+ "hapax_share_of_types": 0.027,
1457
+ "mean_word_len": 5.96,
1458
  "difficulty_fleet_consensus": {
1459
  "top_share_q10_50_90_types": [
1460
+ 0.211,
1461
  0.355,
1462
+ 0.609
1463
  ],
1464
  "top_share_q10_50_90_tokens": [
1465
+ 0.226,
1466
  0.371,
1467
  0.797
1468
  ],
1469
+ "mean_top_types": 0.386,
1470
+ "mean_top_tokens": 0.439,
1471
  "entropy_bits_q10_50_90_types": [
1472
+ 1.667,
1473
+ 2.851,
1474
+ 3.444
1475
  ],
1476
  "closability_q10_50_90_types": [
1477
+ 0.23,
1478
+ 0.558,
1479
+ 0.905
1480
  ],
1481
  "band_share_types": {
1482
+ "easy": 0.684,
1483
+ "mid": 0.173,
1484
+ "trap": 0.143,
1485
+ "novel": 0.001
1486
  },
1487
  "band_share_tokens": {
1488
+ "easy": 0.664,
1489
+ "mid": 0.129,
1490
+ "trap": 0.207,
1491
  "novel": 0.0
1492
  },
1493
+ "novel_word_ends_types": 5,
1494
+ "novel_word_ends_tokens": 10
1495
  },
1496
  "surprise_per_tokenizer_tokens": {
1497
  "gpt2": {
1498
+ "dead_cell_share": 0.0016,
1499
+ "mean_surprisal_bits_witnessed_steps": 1.654,
1500
+ "unwitnessed_step_share": 0.0124
1501
  },
1502
  "smollm2": {
1503
+ "dead_cell_share": 0.0016,
1504
+ "mean_surprisal_bits_witnessed_steps": 1.647,
1505
+ "unwitnessed_step_share": 0.0099
1506
  },
1507
  "cl100k_base": {
1508
+ "dead_cell_share": 0.0016,
1509
+ "mean_surprisal_bits_witnessed_steps": 1.66,
1510
+ "unwitnessed_step_share": 0.0104
1511
  },
1512
  "o200k_base": {
1513
+ "dead_cell_share": 0.0009,
1514
+ "mean_surprisal_bits_witnessed_steps": 1.807,
1515
+ "unwitnessed_step_share": 0.0051
1516
  },
1517
  "llama3": {
1518
+ "dead_cell_share": 0.0016,
1519
+ "mean_surprisal_bits_witnessed_steps": 1.672,
1520
+ "unwitnessed_step_share": 0.0103
1521
  },
1522
  "qwen25": {
1523
+ "dead_cell_share": 0.0016,
1524
+ "mean_surprisal_bits_witnessed_steps": 1.667,
1525
+ "unwitnessed_step_share": 0.0104
1526
  },
1527
  "deepseek-v3": {
1528
+ "dead_cell_share": 0.0012,
1529
+ "mean_surprisal_bits_witnessed_steps": 1.732,
1530
+ "unwitnessed_step_share": 0.0097
1531
  },
1532
  "mistral-v03": {
1533
+ "dead_cell_share": 0.0055,
1534
+ "mean_surprisal_bits_witnessed_steps": 1.421,
1535
+ "unwitnessed_step_share": 0.037
1536
  },
1537
  "xlmr": {
1538
+ "dead_cell_share": 0.0047,
1539
+ "mean_surprisal_bits_witnessed_steps": 2.645,
1540
+ "unwitnessed_step_share": 0.0382
1541
  },
1542
  "t5-unigram": {
1543
+ "dead_cell_share": 0.0018,
1544
+ "mean_surprisal_bits_witnessed_steps": 1.797,
1545
+ "unwitnessed_step_share": 0.0147
1546
  },
1547
  "bert-wordpiece": {
1548
+ "dead_cell_share": 0.0012,
1549
+ "mean_surprisal_bits_witnessed_steps": 1.747,
1550
+ "unwitnessed_step_share": 0.0106
1551
  }
1552
  },
1553
  "diet_tokens": {
1554
+ "cells": 1862113,
1555
+ "under_floor_share": 0.001,
1556
+ "never_seen_share": 0.0,
1557
+ "mean_entropy_bits_known_cells": 1.775
1558
  }
1559
  }
1560
  }
1561
  },
1562
+ "danbooru": {
1563
+ "words_total": 7009965,
1564
+ "words_distinct": 40110,
1565
+ "hapax_share_of_types": 0.184,
1566
+ "mean_word_len": 5.46,
1567
  "difficulty_fleet_consensus": {
1568
  "top_share_q10_50_90_types": [
1569
+ 0.205,
1570
  0.355,
1571
+ 0.638
1572
  ],
1573
  "top_share_q10_50_90_tokens": [
1574
+ 0.19,
1575
+ 0.341,
1576
+ 0.548
1577
  ],
1578
  "mean_top_types": 0.388,
1579
+ "mean_top_tokens": 0.359,
1580
  "entropy_bits_q10_50_90_types": [
1581
+ 1.564,
1582
+ 2.803,
1583
+ 3.457
1584
  ],
1585
  "closability_q10_50_90_types": [
1586
+ 0.096,
1587
+ 0.365,
1588
+ 0.86
1589
  ],
1590
  "band_share_types": {
1591
+ "easy": 0.667,
1592
+ "mid": 0.158,
1593
+ "trap": 0.165,
1594
+ "novel": 0.01
1595
  },
1596
  "band_share_tokens": {
1597
+ "easy": 0.732,
1598
+ "mid": 0.168,
1599
+ "trap": 0.1,
1600
  "novel": 0.0
1601
  },
1602
+ "novel_word_ends_types": 382,
1603
+ "novel_word_ends_tokens": 2904
1604
  },
1605
  "surprise_per_tokenizer_tokens": {
1606
  "gpt2": {
1607
+ "dead_cell_share": 0.0106,
1608
+ "mean_surprisal_bits_witnessed_steps": 2.071,
1609
+ "unwitnessed_step_share": 0.0613
1610
  },
1611
  "smollm2": {
1612
+ "dead_cell_share": 0.0126,
1613
+ "mean_surprisal_bits_witnessed_steps": 2.102,
1614
+ "unwitnessed_step_share": 0.0672
1615
  },
1616
  "cl100k_base": {
1617
+ "dead_cell_share": 0.0072,
1618
+ "mean_surprisal_bits_witnessed_steps": 2.127,
1619
+ "unwitnessed_step_share": 0.0545
1620
  },
1621
  "o200k_base": {
1622
+ "dead_cell_share": 0.0026,
1623
+ "mean_surprisal_bits_witnessed_steps": 2.295,
1624
+ "unwitnessed_step_share": 0.0369
1625
  },
1626
  "llama3": {
1627
+ "dead_cell_share": 0.0058,
1628
+ "mean_surprisal_bits_witnessed_steps": 2.146,
1629
+ "unwitnessed_step_share": 0.0516
1630
  },
1631
  "qwen25": {
1632
+ "dead_cell_share": 0.0071,
1633
+ "mean_surprisal_bits_witnessed_steps": 2.137,
1634
+ "unwitnessed_step_share": 0.0538
1635
  },
1636
  "deepseek-v3": {
1637
+ "dead_cell_share": 0.0062,
1638
+ "mean_surprisal_bits_witnessed_steps": 2.157,
1639
+ "unwitnessed_step_share": 0.0531
1640
  },
1641
  "mistral-v03": {
1642
+ "dead_cell_share": 0.0222,
1643
+ "mean_surprisal_bits_witnessed_steps": 1.878,
1644
+ "unwitnessed_step_share": 0.1253
1645
  },
1646
  "xlmr": {
1647
+ "dead_cell_share": 0.0062,
1648
+ "mean_surprisal_bits_witnessed_steps": 2.833,
1649
+ "unwitnessed_step_share": 0.0869
1650
  },
1651
  "t5-unigram": {
1652
+ "dead_cell_share": 0.0104,
1653
+ "mean_surprisal_bits_witnessed_steps": 2.128,
1654
+ "unwitnessed_step_share": 0.075
1655
  },
1656
  "bert-wordpiece": {
1657
+ "dead_cell_share": 0.0075,
1658
+ "mean_surprisal_bits_witnessed_steps": 2.109,
1659
+ "unwitnessed_step_share": 0.0496
1660
  }
1661
  },
1662
  "diet_tokens": {
1663
+ "cells": 24230571,
1664
+ "under_floor_share": 0.0051,
1665
+ "never_seen_share": 0.0007,
1666
+ "mean_entropy_bits_known_cells": 1.934
1667
  },
1668
+ "rows_sampled": 30160,
1669
+ "rows_total": 606017,
1670
+ "bytes_per_row": 1657.0,
1671
+ "words_per_row": 232.4,
1672
  "registers": {
1673
  "tags": {
1674
+ "words_total": 1681110,
1675
+ "words_distinct": 23869,
1676
+ "hapax_share_of_types": 0.377,
1677
+ "mean_word_len": 5.59,
1678
  "difficulty_fleet_consensus": {
1679
  "top_share_q10_50_90_types": [
1680
+ 0.201,
1681
+ 0.342,
1682
+ 0.626
1683
  ],
1684
  "top_share_q10_50_90_tokens": [
1685
  0.211,
1686
  0.355,
1687
+ 0.583
1688
  ],
1689
+ "mean_top_types": 0.379,
1690
+ "mean_top_tokens": 0.377,
1691
  "entropy_bits_q10_50_90_types": [
1692
+ 1.687,
1693
+ 2.819,
1694
+ 3.487
1695
  ],
1696
  "closability_q10_50_90_types": [
1697
+ 0.086,
1698
+ 0.32,
1699
+ 0.741
1700
  ],
1701
  "band_share_types": {
1702
+ "easy": 0.693,
1703
+ "mid": 0.149,
1704
+ "trap": 0.153,
1705
+ "novel": 0.005
1706
  },
1707
  "band_share_tokens": {
1708
+ "easy": 0.7,
1709
+ "mid": 0.178,
1710
+ "trap": 0.121,
1711
+ "novel": 0.001
1712
  },
1713
+ "novel_word_ends_types": 123,
1714
+ "novel_word_ends_tokens": 1093
1715
  },
1716
  "surprise_per_tokenizer_tokens": {
1717
  "gpt2": {
1718
+ "dead_cell_share": 0.0175,
1719
+ "mean_surprisal_bits_witnessed_steps": 2.175,
1720
+ "unwitnessed_step_share": 0.097
1721
  },
1722
  "smollm2": {
1723
+ "dead_cell_share": 0.0211,
1724
+ "mean_surprisal_bits_witnessed_steps": 2.215,
1725
+ "unwitnessed_step_share": 0.11
1726
  },
1727
  "cl100k_base": {
1728
+ "dead_cell_share": 0.0127,
1729
+ "mean_surprisal_bits_witnessed_steps": 2.265,
1730
+ "unwitnessed_step_share": 0.0881
1731
  },
1732
  "o200k_base": {
1733
+ "dead_cell_share": 0.0055,
1734
+ "mean_surprisal_bits_witnessed_steps": 2.441,
1735
+ "unwitnessed_step_share": 0.0596
1736
  },
1737
  "llama3": {
1738
+ "dead_cell_share": 0.0105,
1739
+ "mean_surprisal_bits_witnessed_steps": 2.291,
1740
+ "unwitnessed_step_share": 0.0826
1741
  },
1742
  "qwen25": {
1743
+ "dead_cell_share": 0.0125,
1744
+ "mean_surprisal_bits_witnessed_steps": 2.276,
1745
+ "unwitnessed_step_share": 0.0867
1746
  },
1747
  "deepseek-v3": {
1748
+ "dead_cell_share": 0.0118,
1749
+ "mean_surprisal_bits_witnessed_steps": 2.294,
1750
+ "unwitnessed_step_share": 0.0871
1751
  },
1752
  "mistral-v03": {
1753
+ "dead_cell_share": 0.0337,
1754
+ "mean_surprisal_bits_witnessed_steps": 1.966,
1755
+ "unwitnessed_step_share": 0.1765
1756
  },
1757
  "xlmr": {
1758
+ "dead_cell_share": 0.0094,
1759
+ "mean_surprisal_bits_witnessed_steps": 2.936,
1760
+ "unwitnessed_step_share": 0.115
1761
  },
1762
  "t5-unigram": {
1763
+ "dead_cell_share": 0.0176,
1764
+ "mean_surprisal_bits_witnessed_steps": 2.239,
1765
+ "unwitnessed_step_share": 0.1144
1766
  },
1767
  "bert-wordpiece": {
1768
+ "dead_cell_share": 0.0133,
1769
+ "mean_surprisal_bits_witnessed_steps": 2.202,
1770
+ "unwitnessed_step_share": 0.0836
1771
  }
1772
  },
1773
  "diet_tokens": {
1774
+ "cells": 6031919,
1775
+ "under_floor_share": 0.0085,
1776
+ "never_seen_share": 0.0012,
1777
+ "mean_entropy_bits_known_cells": 2.005
1778
  }
1779
  },
1780
  "english": {
1781
+ "words_total": 5147935,
1782
+ "words_distinct": 31940,
1783
+ "hapax_share_of_types": 0.081,
1784
+ "mean_word_len": 5.41,
1785
  "difficulty_fleet_consensus": {
1786
  "top_share_q10_50_90_types": [
1787
+ 0.205,
1788
  0.355,
1789
+ 0.638
1790
  ],
1791
  "top_share_q10_50_90_tokens": [
1792
  0.19,
1793
+ 0.333,
1794
+ 0.537
1795
  ],
1796
  "mean_top_types": 0.389,
1797
+ "mean_top_tokens": 0.354,
1798
  "entropy_bits_q10_50_90_types": [
1799
+ 1.541,
1800
+ 2.803,
1801
+ 3.457
1802
  ],
1803
  "closability_q10_50_90_types": [
1804
+ 0.1,
1805
+ 0.393,
1806
+ 0.86
1807
  ],
1808
  "band_share_types": {
1809
+ "easy": 0.663,
1810
+ "mid": 0.161,
1811
+ "trap": 0.166,
1812
+ "novel": 0.01
1813
  },
1814
  "band_share_tokens": {
1815
+ "easy": 0.735,
1816
  "mid": 0.168,
1817
+ "trap": 0.096,
1818
+ "novel": 0.0
1819
  },
1820
+ "novel_word_ends_types": 312,
1821
+ "novel_word_ends_tokens": 1811
1822
  },
1823
  "surprise_per_tokenizer_tokens": {
1824
  "gpt2": {
1825
+ "dead_cell_share": 0.0086,
1826
+ "mean_surprisal_bits_witnessed_steps": 2.039,
1827
+ "unwitnessed_step_share": 0.0511
1828
  },
1829
  "smollm2": {
1830
+ "dead_cell_share": 0.0102,
1831
+ "mean_surprisal_bits_witnessed_steps": 2.069,
1832
+ "unwitnessed_step_share": 0.0548
1833
  },
1834
  "cl100k_base": {
1835
+ "dead_cell_share": 0.0056,
1836
+ "mean_surprisal_bits_witnessed_steps": 2.09,
1837
+ "unwitnessed_step_share": 0.0447
1838
  },
1839
  "o200k_base": {
1840
+ "dead_cell_share": 0.0017,
1841
+ "mean_surprisal_bits_witnessed_steps": 2.255,
1842
+ "unwitnessed_step_share": 0.0303
1843
  },
1844
  "llama3": {
1845
+ "dead_cell_share": 0.0044,
1846
+ "mean_surprisal_bits_witnessed_steps": 2.107,
1847
+ "unwitnessed_step_share": 0.0426
1848
  },
1849
  "qwen25": {
1850
+ "dead_cell_share": 0.0055,
1851
+ "mean_surprisal_bits_witnessed_steps": 2.1,
1852
+ "unwitnessed_step_share": 0.0443
1853
  },
1854
  "deepseek-v3": {
1855
+ "dead_cell_share": 0.0046,
1856
+ "mean_surprisal_bits_witnessed_steps": 2.118,
1857
+ "unwitnessed_step_share": 0.0432
1858
  },
1859
  "mistral-v03": {
1860
+ "dead_cell_share": 0.0191,
1861
+ "mean_surprisal_bits_witnessed_steps": 1.845,
1862
+ "unwitnessed_step_share": 0.112
1863
  },
1864
  "xlmr": {
1865
+ "dead_cell_share": 0.0053,
1866
+ "mean_surprisal_bits_witnessed_steps": 2.806,
1867
+ "unwitnessed_step_share": 0.0803
1868
  },
1869
  "t5-unigram": {
1870
+ "dead_cell_share": 0.0084,
1871
+ "mean_surprisal_bits_witnessed_steps": 2.098,
1872
+ "unwitnessed_step_share": 0.064
1873
  },
1874
  "bert-wordpiece": {
1875
+ "dead_cell_share": 0.0058,
1876
+ "mean_surprisal_bits_witnessed_steps": 2.074,
1877
+ "unwitnessed_step_share": 0.0396
1878
  }
1879
  },
1880
  "diet_tokens": {
1881
+ "cells": 17547574,
1882
+ "under_floor_share": 0.0041,
1883
+ "never_seen_share": 0.0006,
1884
+ "mean_entropy_bits_known_cells": 1.907
1885
  }
1886
  }
1887
  }
 
1889
  },
1890
  "registers": {
1891
  "tags": {
1892
+ "words_total": 3876196,
1893
+ "words_distinct": 23874,
1894
+ "hapax_share_of_types": 0.377,
1895
+ "mean_word_len": 5.55,
1896
  "difficulty_fleet_consensus": {
1897
  "top_share_q10_50_90_types": [
1898
+ 0.201,
1899
+ 0.342,
1900
+ 0.626
1901
  ],
1902
  "top_share_q10_50_90_tokens": [
1903
+ 0.219,
1904
  0.355,
1905
+ 0.583
1906
  ],
1907
+ "mean_top_types": 0.379,
1908
+ "mean_top_tokens": 0.377,
1909
  "entropy_bits_q10_50_90_types": [
1910
+ 1.687,
1911
+ 2.819,
1912
+ 3.487
1913
  ],
1914
  "closability_q10_50_90_types": [
1915
+ 0.086,
1916
+ 0.32,
1917
+ 0.741
1918
  ],
1919
  "band_share_types": {
1920
+ "easy": 0.693,
1921
+ "mid": 0.149,
1922
+ "trap": 0.153,
1923
+ "novel": 0.005
1924
  },
1925
  "band_share_tokens": {
1926
+ "easy": 0.704,
1927
+ "mid": 0.183,
1928
+ "trap": 0.112,
1929
  "novel": 0.0
1930
  },
1931
+ "novel_word_ends_types": 123,
1932
+ "novel_word_ends_tokens": 1093
1933
  },
1934
  "surprise_per_tokenizer_tokens": {
1935
  "gpt2": {
1936
+ "dead_cell_share": 0.0089,
1937
+ "mean_surprisal_bits_witnessed_steps": 2.113,
1938
+ "unwitnessed_step_share": 0.0537
1939
  },
1940
  "smollm2": {
1941
+ "dead_cell_share": 0.0106,
1942
+ "mean_surprisal_bits_witnessed_steps": 2.154,
1943
+ "unwitnessed_step_share": 0.0618
1944
  },
1945
  "cl100k_base": {
1946
+ "dead_cell_share": 0.0062,
1947
+ "mean_surprisal_bits_witnessed_steps": 2.201,
1948
+ "unwitnessed_step_share": 0.0481
1949
  },
1950
  "o200k_base": {
1951
+ "dead_cell_share": 0.0026,
1952
+ "mean_surprisal_bits_witnessed_steps": 2.35,
1953
+ "unwitnessed_step_share": 0.0324
1954
  },
1955
  "llama3": {
1956
+ "dead_cell_share": 0.0052,
1957
+ "mean_surprisal_bits_witnessed_steps": 2.229,
1958
+ "unwitnessed_step_share": 0.0457
1959
  },
1960
  "qwen25": {
1961
+ "dead_cell_share": 0.0061,
1962
+ "mean_surprisal_bits_witnessed_steps": 2.213,
1963
+ "unwitnessed_step_share": 0.0472
1964
  },
1965
  "deepseek-v3": {
1966
+ "dead_cell_share": 0.0058,
1967
+ "mean_surprisal_bits_witnessed_steps": 2.232,
1968
+ "unwitnessed_step_share": 0.0491
1969
  },
1970
  "mistral-v03": {
1971
+ "dead_cell_share": 0.0189,
1972
+ "mean_surprisal_bits_witnessed_steps": 1.917,
1973
+ "unwitnessed_step_share": 0.1231
1974
  },
1975
  "xlmr": {
1976
+ "dead_cell_share": 0.0045,
1977
+ "mean_surprisal_bits_witnessed_steps": 2.875,
1978
+ "unwitnessed_step_share": 0.082
1979
  },
1980
  "t5-unigram": {
1981
+ "dead_cell_share": 0.0092,
1982
+ "mean_surprisal_bits_witnessed_steps": 2.162,
1983
+ "unwitnessed_step_share": 0.0631
1984
  },
1985
  "bert-wordpiece": {
1986
+ "dead_cell_share": 0.0071,
1987
+ "mean_surprisal_bits_witnessed_steps": 2.118,
1988
+ "unwitnessed_step_share": 0.0512
1989
  }
1990
  },
1991
  "diet_tokens": {
1992
+ "cells": 13776670,
1993
+ "under_floor_share": 0.004,
1994
+ "never_seen_share": 0.0005,
1995
+ "mean_entropy_bits_known_cells": 2.024
1996
  }
1997
  },
1998
  "english": {
1999
+ "words_total": 17093369,
2000
+ "words_distinct": 54480,
2001
+ "hapax_share_of_types": 0.182,
2002
+ "mean_word_len": 5.4,
2003
  "difficulty_fleet_consensus": {
2004
  "top_share_q10_50_90_types": [
2005
+ 0.207,
2006
  0.355,
2007
+ 0.638
2008
  ],
2009
  "top_share_q10_50_90_tokens": [
2010
  0.19,
2011
+ 0.342,
2012
+ 0.552
2013
  ],
2014
  "mean_top_types": 0.39,
2015
+ "mean_top_tokens": 0.364,
2016
  "entropy_bits_q10_50_90_types": [
2017
+ 1.541,
2018
+ 2.789,
2019
+ 3.457
2020
  ],
2021
  "closability_q10_50_90_types": [
2022
+ 0.112,
2023
+ 0.413,
2024
+ 0.87
2025
  ],
2026
  "band_share_types": {
2027
+ "easy": 0.66,
2028
+ "mid": 0.166,
2029
+ "trap": 0.164,
2030
+ "novel": 0.01
2031
  },
2032
  "band_share_tokens": {
2033
+ "easy": 0.715,
2034
+ "mid": 0.179,
2035
+ "trap": 0.105,
2036
+ "novel": 0.0
2037
  },
2038
+ "novel_word_ends_types": 554,
2039
+ "novel_word_ends_tokens": 7397
2040
  },
2041
  "surprise_per_tokenizer_tokens": {
2042
  "gpt2": {
2043
+ "dead_cell_share": 0.0039,
2044
+ "mean_surprisal_bits_witnessed_steps": 1.979,
2045
+ "unwitnessed_step_share": 0.0281
2046
  },
2047
  "smollm2": {
2048
+ "dead_cell_share": 0.0044,
2049
+ "mean_surprisal_bits_witnessed_steps": 2.008,
2050
+ "unwitnessed_step_share": 0.0295
2051
  },
2052
  "cl100k_base": {
2053
+ "dead_cell_share": 0.0025,
2054
+ "mean_surprisal_bits_witnessed_steps": 2.022,
2055
+ "unwitnessed_step_share": 0.0226
2056
  },
2057
  "o200k_base": {
2058
+ "dead_cell_share": 0.0008,
2059
+ "mean_surprisal_bits_witnessed_steps": 2.171,
2060
+ "unwitnessed_step_share": 0.0152
2061
  },
2062
  "llama3": {
2063
+ "dead_cell_share": 0.0019,
2064
+ "mean_surprisal_bits_witnessed_steps": 2.04,
2065
+ "unwitnessed_step_share": 0.0218
2066
  },
2067
  "qwen25": {
2068
+ "dead_cell_share": 0.0024,
2069
+ "mean_surprisal_bits_witnessed_steps": 2.032,
2070
+ "unwitnessed_step_share": 0.0223
2071
  },
2072
  "deepseek-v3": {
2073
+ "dead_cell_share": 0.0021,
2074
+ "mean_surprisal_bits_witnessed_steps": 2.054,
2075
+ "unwitnessed_step_share": 0.0235
2076
  },
2077
  "mistral-v03": {
2078
+ "dead_cell_share": 0.011,
2079
+ "mean_surprisal_bits_witnessed_steps": 1.792,
2080
+ "unwitnessed_step_share": 0.0774
2081
  },
2082
  "xlmr": {
2083
+ "dead_cell_share": 0.0034,
2084
+ "mean_surprisal_bits_witnessed_steps": 2.739,
2085
+ "unwitnessed_step_share": 0.0584
2086
  },
2087
  "t5-unigram": {
2088
+ "dead_cell_share": 0.004,
2089
+ "mean_surprisal_bits_witnessed_steps": 2.023,
2090
+ "unwitnessed_step_share": 0.0359
2091
  },
2092
  "bert-wordpiece": {
2093
+ "dead_cell_share": 0.0026,
2094
+ "mean_surprisal_bits_witnessed_steps": 2.025,
2095
+ "unwitnessed_step_share": 0.0203
2096
  }
2097
  },
2098
  "diet_tokens": {
2099
+ "cells": 58160900,
2100
+ "under_floor_share": 0.0017,
2101
+ "never_seen_share": 0.0003,
2102
+ "mean_entropy_bits_known_cells": 1.911
2103
  }
2104
  },
2105
  "whole_pack_sample": {
2106
+ "words_total": 25126884,
2107
+ "words_distinct": 61394,
2108
+ "hapax_share_of_types": 0.226,
2109
+ "mean_word_len": 5.45,
2110
  "difficulty_fleet_consensus": {
2111
  "top_share_q10_50_90_types": [
2112
+ 0.206,
2113
  0.355,
2114
+ 0.638
2115
  ],
2116
  "top_share_q10_50_90_tokens": [
2117
+ 0.194,
2118
+ 0.352,
2119
+ 0.552
2120
  ],
2121
+ "mean_top_types": 0.389,
2122
+ "mean_top_tokens": 0.367,
2123
  "entropy_bits_q10_50_90_types": [
2124
+ 1.541,
2125
+ 2.791,
2126
+ 3.459
2127
  ],
2128
  "closability_q10_50_90_types": [
2129
+ 0.103,
2130
+ 0.4,
2131
+ 0.86
2132
  ],
2133
  "band_share_types": {
2134
+ "easy": 0.663,
2135
+ "mid": 0.164,
2136
+ "trap": 0.163,
2137
+ "novel": 0.01
2138
  },
2139
  "band_share_tokens": {
2140
+ "easy": 0.712,
2141
+ "mid": 0.18,
2142
+ "trap": 0.107,
2143
  "novel": 0.0
2144
  },
2145
+ "novel_word_ends_types": 617,
2146
+ "novel_word_ends_tokens": 8495
2147
  },
2148
  "surprise_per_tokenizer_tokens": {
2149
  "gpt2": {
2150
+ "dead_cell_share": 0.0043,
2151
+ "mean_surprisal_bits_witnessed_steps": 2.002,
2152
+ "unwitnessed_step_share": 0.0301
2153
  },
2154
  "smollm2": {
2155
+ "dead_cell_share": 0.0048,
2156
+ "mean_surprisal_bits_witnessed_steps": 2.034,
2157
+ "unwitnessed_step_share": 0.0327
2158
  },
2159
  "cl100k_base": {
2160
+ "dead_cell_share": 0.0027,
2161
+ "mean_surprisal_bits_witnessed_steps": 2.048,
2162
+ "unwitnessed_step_share": 0.0247
2163
  },
2164
  "o200k_base": {
2165
+ "dead_cell_share": 0.001,
2166
+ "mean_surprisal_bits_witnessed_steps": 2.2,
2167
+ "unwitnessed_step_share": 0.0166
2168
  },
2169
  "llama3": {
2170
+ "dead_cell_share": 0.0022,
2171
+ "mean_surprisal_bits_witnessed_steps": 2.068,
2172
+ "unwitnessed_step_share": 0.0238
2173
  },
2174
  "qwen25": {
2175
+ "dead_cell_share": 0.0027,
2176
+ "mean_surprisal_bits_witnessed_steps": 2.059,
2177
+ "unwitnessed_step_share": 0.0243
2178
  },
2179
  "deepseek-v3": {
2180
+ "dead_cell_share": 0.0024,
2181
+ "mean_surprisal_bits_witnessed_steps": 2.084,
2182
+ "unwitnessed_step_share": 0.0261
2183
  },
2184
  "mistral-v03": {
2185
+ "dead_cell_share": 0.011,
2186
+ "mean_surprisal_bits_witnessed_steps": 1.814,
2187
+ "unwitnessed_step_share": 0.0821
2188
  },
2189
  "xlmr": {
2190
+ "dead_cell_share": 0.0032,
2191
+ "mean_surprisal_bits_witnessed_steps": 2.758,
2192
+ "unwitnessed_step_share": 0.06
2193
  },
2194
  "t5-unigram": {
2195
+ "dead_cell_share": 0.0043,
2196
+ "mean_surprisal_bits_witnessed_steps": 2.048,
2197
+ "unwitnessed_step_share": 0.0375
2198
  },
2199
  "bert-wordpiece": {
2200
+ "dead_cell_share": 0.0029,
2201
+ "mean_surprisal_bits_witnessed_steps": 2.041,
2202
+ "unwitnessed_step_share": 0.0233
2203
  }
2204
  },
2205
  "diet_tokens": {
2206
+ "cells": 86686882,
2207
+ "under_floor_share": 0.0018,
2208
+ "never_seen_share": 0.0003,
2209
+ "mean_entropy_bits_known_cells": 1.953
2210
  }
2211
  }
2212
  },
2213
+ "implanted_lexicon_hits": {
2214
+ "scf:passag": 1,
2215
+ "cc12m:cades": 1,
2216
+ "danbooru:raye": 1
2217
+ }
2218
  }