{
  "dataset_groups": [
    {
      "bytes": 3977331,
      "delta_bits": 1345478.5344710723,
      "documents": 129648,
      "gap_bpb": 0.3382867894251377,
      "model_a_bpb": 0.5736822739748099,
      "model_b_bpb": 0.23539548454967224,
      "name": "epic:5005"
    },
    {
      "bytes": 3977331,
      "delta_bits": 1345478.5344710723,
      "documents": 129648,
      "gap_bpb": 0.3382867894251377,
      "model_a_bpb": 0.5736822739748099,
      "model_b_bpb": 0.23539548454967224,
      "name": "eval_only"
    },
    {
      "bytes": 3977331,
      "delta_bits": 1345478.5344710723,
      "documents": 129648,
      "gap_bpb": 0.3382867894251377,
      "model_a_bpb": 0.5736822739748099,
      "model_b_bpb": 0.23539548454967224,
      "name": "issue:6103"
    },
    {
      "bytes": 3977331,
      "delta_bits": 1345478.5344710723,
      "documents": 129648,
      "gap_bpb": 0.3382867894251377,
      "model_a_bpb": 0.5736822739748099,
      "model_b_bpb": 0.23539548454967224,
      "name": "loss:target_only"
    },
    {
      "bytes": 3977331,
      "delta_bits": 1345478.5344710723,
      "documents": 129648,
      "gap_bpb": 0.3382867894251377,
      "model_a_bpb": 0.5736822739748099,
      "model_b_bpb": 0.23539548454967224,
      "name": "ppl_circuit_coverage"
    },
    {
      "bytes": 3977331,
      "delta_bits": 1345478.5344710723,
      "documents": 129648,
      "gap_bpb": 0.3382867894251377,
      "model_a_bpb": 0.5736822739748099,
      "model_b_bpb": 0.23539548454967224,
      "name": "seed:6103"
    },
    {
      "bytes": 3977331,
      "delta_bits": 1345478.5344710723,
      "documents": 129648,
      "gap_bpb": 0.3382867894251377,
      "model_a_bpb": 0.5736822739748099,
      "model_b_bpb": 0.23539548454967224,
      "name": "source:generated_ppl_circuit_coverage_v1"
    },
    {
      "bytes": 3977331,
      "delta_bits": 1345478.5344710723,
      "documents": 129648,
      "gap_bpb": 0.3382867894251377,
      "model_a_bpb": 0.5736822739748099,
      "model_b_bpb": 0.23539548454967224,
      "name": "tokenizer:marin-community"
    },
    {
      "bytes": 3977331,
      "delta_bits": 1345478.5344710723,
      "documents": 129648,
      "gap_bpb": 0.3382867894251377,
      "model_a_bpb": 0.5736822739748099,
      "model_b_bpb": 0.23539548454967224,
      "name": "tokenizer:marin-community/marin-tokenizer"
    },
    {
      "bytes": 3909567,
      "delta_bits": 1321717.1655325263,
      "documents": 126648,
      "gap_bpb": 0.33807251942031596,
      "model_a_bpb": 0.5633475927750925,
      "model_b_bpb": 0.22527507335477653,
      "name": "examples:tokenizer_vocab_coverage"
    },
    {
      "bytes": 3909567,
      "delta_bits": 1321717.1655325263,
      "documents": 126648,
      "gap_bpb": 0.33807251942031596,
      "model_a_bpb": 0.5633475927750925,
      "model_b_bpb": 0.22527507335477653,
      "name": "family:vocab_mechanics"
    },
    {
      "bytes": 3909567,
      "delta_bits": 1321717.1655325263,
      "documents": 126648,
      "gap_bpb": 0.33807251942031596,
      "model_a_bpb": 0.5633475927750925,
      "model_b_bpb": 0.22527507335477653,
      "name": "ppl_circuit_coverage/vocab_mechanics"
    },
    {
      "bytes": 3909567,
      "delta_bits": 1321717.1655325263,
      "documents": 126648,
      "gap_bpb": 0.33807251942031596,
      "model_a_bpb": 0.5633475927750925,
      "model_b_bpb": 0.22527507335477653,
      "name": "task:token_chars"
    },
    {
      "bytes": 67764,
      "delta_bits": 23761.368938546115,
      "documents": 3000,
      "gap_bpb": 0.35064885394230144,
      "model_a_bpb": 1.1699299655800255,
      "model_b_bpb": 0.8192811116377242,
      "name": "examples:1000"
    },
    {
      "bytes": 64751,
      "delta_bits": 19613.099714707503,
      "documents": 2000,
      "gap_bpb": 0.3029003369014765,
      "model_a_bpb": 1.0656726461622585,
      "model_b_bpb": 0.7627723092607821,
      "name": "family:text_mechanics"
    },
    {
      "bytes": 64751,
      "delta_bits": 19613.099714707503,
      "documents": 2000,
      "gap_bpb": 0.3029003369014765,
      "model_a_bpb": 1.0656726461622585,
      "model_b_bpb": 0.7627723092607821,
      "name": "ppl_circuit_coverage/text_mechanics"
    },
    {
      "bytes": 58483,
      "delta_bits": 11320.221457130616,
      "documents": 1000,
      "gap_bpb": 0.19356430855343632,
      "model_a_bpb": 0.9362593456192041,
      "model_b_bpb": 0.7426950370657677,
      "name": "task:line_rhythm"
    },
    {
      "bytes": 6268,
      "delta_bits": 8292.878257576887,
      "documents": 1000,
      "gap_bpb": 1.3230501368182654,
      "model_a_bpb": 2.2731515956931228,
      "model_b_bpb": 0.9501014588748572,
      "name": "task:character_indices"
    },
    {
      "bytes": 3013,
      "delta_bits": 4148.26922383861,
      "documents": 1000,
      "gap_bpb": 1.376790316574381,
      "model_a_bpb": 3.410476161935762,
      "model_b_bpb": 2.0336858453613815,
      "name": "family:python_interpretation"
    },
    {
      "bytes": 3013,
      "delta_bits": 4148.26922383861,
      "documents": 1000,
      "gap_bpb": 1.376790316574381,
      "model_a_bpb": 3.410476161935762,
      "model_b_bpb": 2.0336858453613815,
      "name": "ppl_circuit_coverage/python_interpretation"
    },
    {
      "bytes": 3013,
      "delta_bits": 4148.26922383861,
      "documents": 1000,
      "gap_bpb": 1.376790316574381,
      "model_a_bpb": 3.410476161935762,
      "model_b_bpb": 2.0336858453613815,
      "name": "task:arithmetic_repl"
    }
  ],
  "datasets": [
    {
      "bytes": 3909567,
      "delta_bits": 1321717.1655325263,
      "documents": 126648,
      "gap_bpb": 0.33807251942031596,
      "model_a_bpb": 0.5633475927750925,
      "model_b_bpb": 0.22527507335477653,
      "name": "ppl_circuit_coverage/vocab_mechanics/token_chars"
    },
    {
      "bytes": 58483,
      "delta_bits": 11320.221457130616,
      "documents": 1000,
      "gap_bpb": 0.19356430855343632,
      "model_a_bpb": 0.9362593456192041,
      "model_b_bpb": 0.7426950370657677,
      "name": "ppl_circuit_coverage/text_mechanics/line_rhythm"
    },
    {
      "bytes": 6268,
      "delta_bits": 8292.878257576887,
      "documents": 1000,
      "gap_bpb": 1.3230501368182654,
      "model_a_bpb": 2.2731515956931228,
      "model_b_bpb": 0.9501014588748572,
      "name": "ppl_circuit_coverage/text_mechanics/character_indices"
    },
    {
      "bytes": 3013,
      "delta_bits": 4148.26922383861,
      "documents": 1000,
      "gap_bpb": 1.376790316574381,
      "model_a_bpb": 3.410476161935762,
      "model_b_bpb": 2.0336858453613815,
      "name": "ppl_circuit_coverage/python_interpretation/arithmetic_repl"
    }
  ],
  "model_a": "marin-community/marin-32b-base",
  "model_b": "Qwen/Qwen3-32B",
  "pattern_buckets": [
    {
      "bytes": 2401100,
      "delta_bits": 764853.2136891272,
      "documents": 1432112,
      "gap_bpb": 0.31854284023536183,
      "model_a_bpb": 0.5768083416224548,
      "model_b_bpb": 0.258265501387093,
      "name": "text/punctuation"
    },
    {
      "bytes": 579285,
      "delta_bits": 271035.99878886534,
      "documents": 555937,
      "gap_bpb": 0.46788022957415665,
      "model_a_bpb": 0.6397647863626839,
      "model_b_bpb": 0.17188455678852727,
      "name": "text/word"
    },
    {
      "bytes": 155800,
      "delta_bits": 176884.15735680782,
      "documents": 68470,
      "gap_bpb": 1.1353283527394598,
      "model_a_bpb": 1.404737052546774,
      "model_b_bpb": 0.2694086998073142,
      "name": "text/non_ascii_word"
    },
    {
      "bytes": 673787,
      "delta_bits": 64733.577631564025,
      "documents": 673787,
      "gap_bpb": 0.09607424546861845,
      "model_a_bpb": 0.23532359319435836,
      "model_b_bpb": 0.1392493477257399,
      "name": "whitespace/single_space"
    },
    {
      "bytes": 129648,
      "delta_bits": 44921.619721292955,
      "documents": 129648,
      "gap_bpb": 0.3464891068222645,
      "model_a_bpb": 0.6364287039102098,
      "model_b_bpb": 0.28993959708794514,
      "name": "whitespace/newline"
    },
    {
      "bytes": 29213,
      "delta_bits": 15452.205938990759,
      "documents": 8769,
      "gap_bpb": 0.5289496436172512,
      "model_a_bpb": 1.8228142074068105,
      "model_b_bpb": 1.2938645637895594,
      "name": "text/number"
    },
    {
      "bytes": 8498,
      "delta_bits": 7597.7613444359,
      "documents": 1311,
      "gap_bpb": 0.8940646439675101,
      "model_a_bpb": 1.5257764618471563,
      "model_b_bpb": 0.6317118178796463,
      "name": "text/non_ascii"
    }
  ],
  "top_documents": {
    "model_a_worse": [
      {
        "bytes": 78,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 94.87529903756109,
        "gap_bpb": 1.2163499876610397,
        "model_a_bpb": 1.2865832395441112,
        "model_b_bpb": 0.07023325188307175,
        "preview": ">>>\u2420token_chars(\"\"\"\u2420\u0445\u0430\u0440\u0430\u043a\u0442\u0435\u0440\u0438\u0441\u0442\u0438\"\"\")\u23ce['\u2420',\u2420'\u0445',\u2420'\u0430',\u2420'\u0440',\u2420'\u0430',\u2420'\u043a',\u2420'\u0442',\u2420'\u0435',\u2420'\u0440',\u2420'\u0438',\u2420'\u0441',\u2420'\u0442',\u2420'\u0438']\u23ce",
        "row_index": 110774,
        "score_byte_end": 127,
        "score_byte_start": 49,
        "shard": "vocab_mechanics_token_chars_jsonl_gz",
        "worst_bucket": "text/non_ascii_word",
        "worst_gap_bpb": 9.599535673221894,
        "worst_text": "\u0445"
      },
      {
        "bytes": 84,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 94.05128471623631,
        "gap_bpb": 1.1196581513837656,
        "model_a_bpb": 1.1899040392668776,
        "model_b_bpb": 0.07024588788311205,
        "preview": ">>>\u2420token_chars(\"\"\"\u2420\u0430\u0434\u043c\u0456\u043d\u0456\u0441\u0442\u0440\u0430\u0442\u0438\u0432\"\"\")\u23ce['\u2420',\u2420'\u0430',\u2420'\u0434',\u2420'\u043c',\u2420'\u0456',\u2420'\u043d',\u2420'\u0456',\u2420'\u0441',\u2420'\u0442',\u2420'\u0440',\u2420'\u0430',\u2420'\u0442',\u2420'\u0438',\u2420'\u0432']\u23ce",
        "row_index": 125284,
        "score_byte_end": 135,
        "score_byte_start": 51,
        "shard": "vocab_mechanics_token_chars_jsonl_gz",
        "worst_bucket": "text/non_ascii_word",
        "worst_gap_bpb": 7.839022196983278,
        "worst_text": "\u0430"
      },
      {
        "bytes": 43,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 90.34608274145425,
        "gap_bpb": 2.1010716916617267,
        "model_a_bpb": 2.1169964852123164,
        "model_b_bpb": 0.015924793550589744,
        "preview": ">>>\u2420token_chars(\"\"\"\u7f51\u520a\u4e0b\u8f7d\u6b21\u6570\"\"\")\u23ce['\u7f51',\u2420'\u520a',\u2420'\u4e0b',\u2420'\u8f7d',\u2420'\u6b21',\u2420'\u6570']\u23ce",
        "row_index": 124525,
        "score_byte_end": 85,
        "score_byte_start": 42,
        "shard": "vocab_mechanics_token_chars_jsonl_gz",
        "worst_bucket": "text/non_ascii_word",
        "worst_gap_bpb": 8.554271212628164,
        "worst_text": "\u7f51"
      },
      {
        "bytes": 62,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 85.08023853778387,
        "gap_bpb": 1.37226191189974,
        "model_a_bpb": 1.4925985499607135,
        "model_b_bpb": 0.1203366380609736,
        "preview": ">>>\u2420token_chars(\"\"\"\u2420\u010ceskosloven\"\"\")\u23ce['\u2420',\u2420'\u010c',\u2420'e',\u2420's',\u2420'k',\u2420'o',\u2420's',\u2420'l',\u2420'o',\u2420'v',\u2420'e',\u2420'n']\u23ce",
        "row_index": 122295,
        "score_byte_end": 99,
        "score_byte_start": 37,
        "shard": "vocab_mechanics_token_chars_jsonl_gz",
        "worst_bucket": "text/non_ascii_word",
        "worst_gap_bpb": 12.375292318519609,
        "worst_text": "\u010c"
      },
      {
        "bytes": 66,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 83.90319316369906,
        "gap_bpb": 1.2712605024802888,
        "model_a_bpb": 1.3493836112115287,
        "model_b_bpb": 0.07812310873123987,
        "preview": ">>>\u2420token_chars(\"\"\"\u2420\u043f\u0440\u043e\u0444\u0438\u043b\u0430\u043a\u0442\u0438\"\"\")\u23ce['\u2420',\u2420'\u043f',\u2420'\u0440',\u2420'\u043e',\u2420'\u0444',\u2420'\u0438',\u2420'\u043b',\u2420'\u0430',\u2420'\u043a',\u2420'\u0442',\u2420'\u0438']\u23ce",
        "row_index": 125681,
        "score_byte_end": 111,
        "score_byte_start": 45,
        "shard": "vocab_mechanics_token_chars_jsonl_gz",
        "worst_bucket": "text/non_ascii_word",
        "worst_gap_bpb": 9.202382328959837,
        "worst_text": "\u043f"
      },
      {
        "bytes": 78,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 82.22087509307283,
        "gap_bpb": 1.0541137832445235,
        "model_a_bpb": 1.1122362814425297,
        "model_b_bpb": 0.05812249819800639,
        "preview": ">>>\u2420token_chars(\"\"\"\u2420\u0437\u0430\u0431\u0435\u0437\u043f\u0435\u0447\u0435\u043d\u043d\u044f\"\"\")\u23ce['\u2420',\u2420'\u0437',\u2420'\u0430',\u2420'\u0431',\u2420'\u0435',\u2420'\u0437',\u2420'\u043f',\u2420'\u0435',\u2420'\u0447',\u2420'\u0435',\u2420'\u043d',\u2420'\u043d',\u2420'\u044f']\u23ce",
        "row_index": 112666,
        "score_byte_end": 127,
        "score_byte_start": 49,
        "shard": "vocab_mechanics_token_chars_jsonl_gz",
        "worst_bucket": "text/non_ascii_word",
        "worst_gap_bpb": 6.041667973138053,
        "worst_text": "\u0437"
      },
      {
        "bytes": 66,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 79.81420033230887,
        "gap_bpb": 1.2093060656410435,
        "model_a_bpb": 1.339189719213794,
        "model_b_bpb": 0.12988365357275036,
        "preview": ">>>\u2420token_chars(\"\"\"\u2420\u0445\u0430\u0440\u0430\u043a\u0442\u0435\u0440\u0438\u0437\"\"\")\u23ce['\u2420',\u2420'\u0445',\u2420'\u0430',\u2420'\u0440',\u2420'\u0430',\u2420'\u043a',\u2420'\u0442',\u2420'\u0435',\u2420'\u0440',\u2420'\u0438',\u2420'\u0437']\u23ce",
        "row_index": 119418,
        "score_byte_end": 111,
        "score_byte_start": 45,
        "shard": "vocab_mechanics_token_chars_jsonl_gz",
        "worst_bucket": "text/non_ascii_word",
        "worst_gap_bpb": 9.837921513513427,
        "worst_text": "\u0445"
      },
      {
        "bytes": 57,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 78.9278293046073,
        "gap_bpb": 1.3846987597299527,
        "model_a_bpb": 1.4593085509132895,
        "model_b_bpb": 0.07460979118333662,
        "preview": ">>>\u2420token_chars(\"\"\"\u2420ar\u015fivlendi\"\"\")\u23ce['\u2420',\u2420'a',\u2420'r',\u2420'\u015f',\u2420'i',\u2420'v',\u2420'l',\u2420'e',\u2420'n',\u2420'd',\u2420'i']\u23ce",
        "row_index": 110743,
        "score_byte_end": 93,
        "score_byte_start": 36,
        "shard": "vocab_mechanics_token_chars_jsonl_gz",
        "worst_bucket": "text/non_ascii_word",
        "worst_gap_bpb": 6.678156501924411,
        "worst_text": "\u015f"
      },
      {
        "bytes": 78,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 78.82364762679366,
        "gap_bpb": 1.010559584958893,
        "model_a_bpb": 1.0908227062870677,
        "model_b_bpb": 0.08026312132817455,
        "preview": ">>>\u2420token_chars(\"\"\"\u2420\u0432\u0438\u043a\u043e\u0440\u0438\u0441\u0442\u0430\u043d\u043d\u044f\"\"\")\u23ce['\u2420',\u2420'\u0432',\u2420'\u0438',\u2420'\u043a',\u2420'\u043e',\u2420'\u0440',\u2420'\u0438',\u2420'\u0441',\u2420'\u0442',\u2420'\u0430',\u2420'\u043d',\u2420'\u043d',\u2420'\u044f']\u23ce",
        "row_index": 112113,
        "score_byte_end": 127,
        "score_byte_start": 49,
        "shard": "vocab_mechanics_token_chars_jsonl_gz",
        "worst_bucket": "text/non_ascii_word",
        "worst_gap_bpb": 5.040427975477492,
        "worst_text": "\u0438"
      },
      {
        "bytes": 61,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 77.96187612143487,
        "gap_bpb": 1.278063542974342,
        "model_a_bpb": 1.2969570495083407,
        "model_b_bpb": 0.018893506533998728,
        "preview": ">>>\u2420token_chars(\"\"\"\u0410\u0440\u0445\u0456\u0432\u043e\u0432\u0430\u043d\u043e\"\"\")\u23ce['\u0410',\u2420'\u0440',\u2420'\u0445',\u2420'\u0456',\u2420'\u0432',\u2420'\u043e',\u2420'\u0432',\u2420'\u0430',\u2420'\u043d',\u2420'\u043e']\u23ce",
        "row_index": 116452,
        "score_byte_end": 105,
        "score_byte_start": 44,
        "shard": "vocab_mechanics_token_chars_jsonl_gz",
        "worst_bucket": "text/non_ascii_word",
        "worst_gap_bpb": 12.91348657096243,
        "worst_text": "\u0410"
      },
      {
        "bytes": 72,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 77.69026754599905,
        "gap_bpb": 1.0790314936944312,
        "model_a_bpb": 1.1994711351854905,
        "model_b_bpb": 0.12043964149105918,
        "preview": ">>>\u2420token_chars(\"\"\"\u2420\u043f\u0456\u0434\u043f\u0440\u0438\u0454\u043c\u0441\u0442\u0432\"\"\")\u23ce['\u2420',\u2420'\u043f',\u2420'\u0456',\u2420'\u0434',\u2420'\u043f',\u2420'\u0440',\u2420'\u0438',\u2420'\u0454',\u2420'\u043c',\u2420'\u0441',\u2420'\u0442',\u2420'\u0432']\u23ce",
        "row_index": 125266,
        "score_byte_end": 119,
        "score_byte_start": 47,
        "shard": "vocab_mechanics_token_chars_jsonl_gz",
        "worst_bucket": "text/non_ascii_word",
        "worst_gap_bpb": 4.781713097923367,
        "worst_text": "\u0456"
      },
      {
        "bytes": 96,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 76.60003891732954,
        "gap_bpb": 0.7979170720555161,
        "model_a_bpb": 0.8753754639324746,
        "model_b_bpb": 0.0774583918769585,
        "preview": ">>>\u2420token_chars(\"\"\"\u2420\u0432\u0438\u043a\u043e\u0440\u0438\u0441\u0442\u043e\u0432\u0443\u0432\u0430\u0442\u0438\"\"\")\u23ce['\u2420',\u2420'\u0432',\u2420'\u0438',\u2420'\u043a',\u2420'\u043e',\u2420'\u0440',\u2420'\u0438',\u2420'\u0441',\u2420'\u0442',\u2420'\u043e',\u2420'\u0432',\u2420'\u0443',\u2420'\u0432',\u2420'\u0430',\u2420'\u0442',\u2420'\u0438\u2026",
        "row_index": 123014,
        "score_byte_end": 151,
        "score_byte_start": 55,
        "shard": "vocab_mechanics_token_chars_jsonl_gz",
        "worst_bucket": "text/non_ascii_word",
        "worst_gap_bpb": 4.4629202088739355,
        "worst_text": "\u0438"
      },
      {
        "bytes": 78,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 74.94831823712263,
        "gap_bpb": 0.9608758748349056,
        "model_a_bpb": 1.0612716380580054,
        "model_b_bpb": 0.10039576322309979,
        "preview": ">>>\u2420token_chars(\"\"\"\u2420\u0432\u0456\u0434\u0431\u0443\u0432\u0430\u0454\u0442\u044c\u0441\u044f\"\"\")\u23ce['\u2420',\u2420'\u0432',\u2420'\u0456',\u2420'\u0434',\u2420'\u0431',\u2420'\u0443',\u2420'\u0432',\u2420'\u0430',\u2420'\u0454',\u2420'\u0442',\u2420'\u044c',\u2420'\u0441',\u2420'\u044f']\u23ce",
        "row_index": 123297,
        "score_byte_end": 127,
        "score_byte_start": 49,
        "shard": "vocab_mechanics_token_chars_jsonl_gz",
        "worst_bucket": "text/non_ascii_word",
        "worst_gap_bpb": 5.302612431030133,
        "worst_text": "\u0431"
      },
      {
        "bytes": 56,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 74.71036654016092,
        "gap_bpb": 1.3341136882171594,
        "model_a_bpb": 1.4791424318303448,
        "model_b_bpb": 0.14502874361318532,
        "preview": ">>>\u2420token_chars(\"\"\"\u2420eoqkrvldkf\"\"\")\u23ce['\u2420',\u2420'e',\u2420'o',\u2420'q',\u2420'k',\u2420'r',\u2420'v',\u2420'l',\u2420'd',\u2420'k',\u2420'f']\u23ce",
        "row_index": 107930,
        "score_byte_end": 91,
        "score_byte_start": 35,
        "shard": "vocab_mechanics_token_chars_jsonl_gz",
        "worst_bucket": "text/word",
        "worst_gap_bpb": 12.030910130848284,
        "worst_text": "e"
      },
      {
        "bytes": 78,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 74.0202262955497,
        "gap_bpb": 0.9489772601993552,
        "model_a_bpb": 1.0506610764615631,
        "model_b_bpb": 0.1016838162622079,
        "preview": ">>>\u2420token_chars(\"\"\"\u2420\u0438\u0441\u043f\u043e\u043b\u044c\u0437\u043e\u0432\u0430\u043d\u0438\"\"\")\u23ce['\u2420',\u2420'\u0438',\u2420'\u0441',\u2420'\u043f',\u2420'\u043e',\u2420'\u043b',\u2420'\u044c',\u2420'\u0437',\u2420'\u043e',\u2420'\u0432',\u2420'\u0430',\u2420'\u043d',\u2420'\u0438']\u23ce",
        "row_index": 117125,
        "score_byte_end": 127,
        "score_byte_start": 49,
        "shard": "vocab_mechanics_token_chars_jsonl_gz",
        "worst_bucket": "text/non_ascii_word",
        "worst_gap_bpb": 8.068352806741597,
        "worst_text": "\u0438"
      },
      {
        "bytes": 54,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 73.20790776457329,
        "gap_bpb": 1.355701995640246,
        "model_a_bpb": 1.4269961423493172,
        "model_b_bpb": 0.07129414670907097,
        "preview": ">>>\u2420token_chars(\"\"\"\u2420\u043f\u0440\u043e\u0444\u0435\u0441\u0441\u0438\"\"\")\u23ce['\u2420',\u2420'\u043f',\u2420'\u0440',\u2420'\u043e',\u2420'\u0444',\u2420'\u0435',\u2420'\u0441',\u2420'\u0441',\u2420'\u0438']\u23ce",
        "row_index": 115730,
        "score_byte_end": 95,
        "score_byte_start": 41,
        "shard": "vocab_mechanics_token_chars_jsonl_gz",
        "worst_bucket": "text/non_ascii_word",
        "worst_gap_bpb": 12.465070013992403,
        "worst_text": "\u043f"
      },
      {
        "bytes": 78,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 72.71176273013336,
        "gap_bpb": 0.932202086283761,
        "model_a_bpb": 1.0319446329400146,
        "model_b_bpb": 0.09974254665625376,
        "preview": ">>>\u2420token_chars(\"\"\"\u2420\u0434\u0435\u044f\u0442\u0435\u043b\u044c\u043d\u043e\u0441\u0442\u0438\"\"\")\u23ce['\u2420',\u2420'\u0434',\u2420'\u0435',\u2420'\u044f',\u2420'\u0442',\u2420'\u0435',\u2420'\u043b',\u2420'\u044c',\u2420'\u043d',\u2420'\u043e',\u2420'\u0441',\u2420'\u0442',\u2420'\u0438']\u23ce",
        "row_index": 111338,
        "score_byte_end": 127,
        "score_byte_start": 49,
        "shard": "vocab_mechanics_token_chars_jsonl_gz",
        "worst_bucket": "text/non_ascii_word",
        "worst_gap_bpb": 7.695808116026196,
        "worst_text": "\u0434"
      },
      {
        "bytes": 34,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 72.32439701258497,
        "gap_bpb": 2.1271881474289698,
        "model_a_bpb": 2.245386238989138,
        "model_b_bpb": 0.1181980915601686,
        "preview": ">>>\u2420token_chars(\"\"\"\u2420\ub178\ucd9c\ub4f1\ub85d\"\"\")\u23ce['\u2420',\u2420'\ub178',\u2420'\ucd9c',\u2420'\ub4f1',\u2420'\ub85d']\u23ce",
        "row_index": 117045,
        "score_byte_end": 71,
        "score_byte_start": 37,
        "shard": "vocab_mechanics_token_chars_jsonl_gz",
        "worst_bucket": "text/non_ascii_word",
        "worst_gap_bpb": 7.080218918972693,
        "worst_text": "\ub178"
      },
      {
        "bytes": 72,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 71.94794093118989,
        "gap_bpb": 0.9992769573776373,
        "model_a_bpb": 1.1496969776415866,
        "model_b_bpb": 0.15042002026394913,
        "preview": ">>>\u2420token_chars(\"\"\"\u2420\u03bd\u03b5\u03c6\u03bf\u03ba\u03ac\u03bb\u03c5\u03c8\u03b7\u03c2\"\"\")\u23ce['\u2420',\u2420'\u03bd',\u2420'\u03b5',\u2420'\u03c6',\u2420'\u03bf',\u2420'\u03ba',\u2420'\u03ac',\u2420'\u03bb',\u2420'\u03c5',\u2420'\u03c8',\u2420'\u03b7',\u2420'\u03c2']\u23ce",
        "row_index": 123707,
        "score_byte_end": 119,
        "score_byte_start": 47,
        "shard": "vocab_mechanics_token_chars_jsonl_gz",
        "worst_bucket": "text/non_ascii_word",
        "worst_gap_bpb": 7.381851035153731,
        "worst_text": "\u03bd"
      },
      {
        "bytes": 66,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 71.71899069868518,
        "gap_bpb": 1.0866513742225026,
        "model_a_bpb": 1.3252560806778522,
        "model_b_bpb": 0.2386047064553496,
        "preview": ">>>\u2420token_chars(\"\"\"\u2420\u043f\u0440\u0435\u0434\u043d\u0430\u0437\u043d\u0430\u0447\"\"\")\u23ce['\u2420',\u2420'\u043f',\u2420'\u0440',\u2420'\u0435',\u2420'\u0434',\u2420'\u043d',\u2420'\u0430',\u2420'\u0437',\u2420'\u043d',\u2420'\u0430',\u2420'\u0447']\u23ce",
        "row_index": 125016,
        "score_byte_end": 111,
        "score_byte_start": 45,
        "shard": "vocab_mechanics_token_chars_jsonl_gz",
        "worst_bucket": "text/non_ascii_word",
        "worst_gap_bpb": 10.713642725164187,
        "worst_text": "\u043f"
      },
      {
        "bytes": 47,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 71.56871885282101,
        "gap_bpb": 1.5227386989961917,
        "model_a_bpb": 1.5528843449649934,
        "model_b_bpb": 0.030145645968801697,
        "preview": ">>>\u2420token_chars(\"\"\"\u040e\u044b\u045fN\u040e\u044b\u045fN\"\"\")\u23ce['\u040e',\u2420'\u044b',\u2420'\u045f',\u2420'N',\u2420'\u040e',\u2420'\u044b',\u2420'\u045f',\u2420'N']\u23ce",
        "row_index": 125772,
        "score_byte_end": 85,
        "score_byte_start": 38,
        "shard": "vocab_mechanics_token_chars_jsonl_gz",
        "worst_bucket": "text/non_ascii_word",
        "worst_gap_bpb": 8.473788835469199,
        "worst_text": "\u040e"
      },
      {
        "bytes": 51,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 71.50812660245072,
        "gap_bpb": 1.402120129459818,
        "model_a_bpb": 1.4216784551363388,
        "model_b_bpb": 0.0195583256765209,
        "preview": ">>>\u2420token_chars(\"\"\"ilmektedir\"\"\")\u23ce['i',\u2420'l',\u2420'm',\u2420'e',\u2420'k',\u2420't',\u2420'e',\u2420'd',\u2420'i',\u2420'r']\u23ce",
        "row_index": 107466,
        "score_byte_end": 85,
        "score_byte_start": 34,
        "shard": "vocab_mechanics_token_chars_jsonl_gz",
        "worst_bucket": "text/word",
        "worst_gap_bpb": 13.051695956779408,
        "worst_text": "i"
      },
      {
        "bytes": 66,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 71.42079650494595,
        "gap_bpb": 1.082133280377969,
        "model_a_bpb": 1.2175624467770494,
        "model_b_bpb": 0.13542916639908037,
        "preview": ">>>\u2420token_chars(\"\"\"\u2420\u0432\u0438\u043a\u043e\u0440\u0438\u0441\u0442\u043e\u0432\"\"\")\u23ce['\u2420',\u2420'\u0432',\u2420'\u0438',\u2420'\u043a',\u2420'\u043e',\u2420'\u0440',\u2420'\u0438',\u2420'\u0441',\u2420'\u0442',\u2420'\u043e',\u2420'\u0432']\u23ce",
        "row_index": 106126,
        "score_byte_end": 111,
        "score_byte_start": 45,
        "shard": "vocab_mechanics_token_chars_jsonl_gz",
        "worst_bucket": "text/non_ascii_word",
        "worst_gap_bpb": 8.239717695922506,
        "worst_text": "\u0432"
      },
      {
        "bytes": 78,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 70.8959087383064,
        "gap_bpb": 0.9089219069013641,
        "model_a_bpb": 0.9638739798975112,
        "model_b_bpb": 0.05495207299614703,
        "preview": ">>>\u2420token_chars(\"\"\"\u2420\u0437\u0430\u0445\u0432\u043e\u0440\u044e\u0432\u0430\u043d\u043d\u044f\"\"\")\u23ce['\u2420',\u2420'\u0437',\u2420'\u0430',\u2420'\u0445',\u2420'\u0432',\u2420'\u043e',\u2420'\u0440',\u2420'\u044e',\u2420'\u0432',\u2420'\u0430',\u2420'\u043d',\u2420'\u043d',\u2420'\u044f']\u23ce",
        "row_index": 123169,
        "score_byte_end": 127,
        "score_byte_start": 49,
        "shard": "vocab_mechanics_token_chars_jsonl_gz",
        "worst_bucket": "text/non_ascii_word",
        "worst_gap_bpb": 4.70118084579565,
        "worst_text": "\u044e"
      },
      {
        "bytes": 34,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 70.29837796681858,
        "gap_bpb": 2.0675993519652525,
        "model_a_bpb": 2.241948645969957,
        "model_b_bpb": 0.17434929400470459,
        "preview": ">>>\u2420token_chars(\"\"\"\u2420\ub4f1\ub85d\ub300\ud589\"\"\")\u23ce['\u2420',\u2420'\ub4f1',\u2420'\ub85d',\u2420'\ub300',\u2420'\ud589']\u23ce",
        "row_index": 115154,
        "score_byte_end": 71,
        "score_byte_start": 37,
        "shard": "vocab_mechanics_token_chars_jsonl_gz",
        "worst_bucket": "text/non_ascii_word",
        "worst_gap_bpb": 7.666060620871764,
        "worst_text": "\ub4f1"
      }
    ],
    "model_b_worse": [
      {
        "bytes": 46,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -46.11854848781752,
        "gap_bpb": -1.0025771410395112,
        "model_a_bpb": 0.5565310419768084,
        "model_b_bpb": 1.5591081830163196,
        "preview": ">>>\u2420token_chars(\"\"\"\u2420wannonce\"\"\")\u23ce['\u2420',\u2420'w',\u2420'a',\u2420'n',\u2420'n',\u2420'o',\u2420'n',\u2420'c',\u2420'e']\u23ce",
        "row_index": 83966,
        "score_byte_end": 79,
        "score_byte_start": 33,
        "shard": "vocab_mechanics_token_chars_jsonl_gz",
        "worst_bucket": "text/word",
        "worst_gap_bpb": -17.551718842837232,
        "worst_text": "w"
      },
      {
        "bytes": 61,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -36.98260531953274,
        "gap_bpb": -0.6062722183529957,
        "model_a_bpb": 0.5330969873765795,
        "model_b_bpb": 1.1393692057295752,
        "preview": ">>>\u2420token_chars(\"\"\"\u2420sexkontakte\"\"\")\u23ce['\u2420',\u2420's',\u2420'e',\u2420'x',\u2420'k',\u2420'o',\u2420'n',\u2420't',\u2420'a',\u2420'k',\u2420't',\u2420'e']\u23ce",
        "row_index": 86656,
        "score_byte_end": 97,
        "score_byte_start": 36,
        "shard": "vocab_mechanics_token_chars_jsonl_gz",
        "worst_bucket": "text/word",
        "worst_gap_bpb": -7.829941599009052,
        "worst_text": "n"
      },
      {
        "bytes": 56,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -33.65525299686297,
        "gap_bpb": -0.6009866606582673,
        "model_a_bpb": 0.48952788494551447,
        "model_b_bpb": 1.090514545603782,
        "preview": ">>>\u2420token_chars(\"\"\"\u2420vivastreet\"\"\")\u23ce['\u2420',\u2420'v',\u2420'i',\u2420'v',\u2420'a',\u2420's',\u2420't',\u2420'r',\u2420'e',\u2420'e',\u2420't']\u23ce",
        "row_index": 65287,
        "score_byte_end": 91,
        "score_byte_start": 35,
        "shard": "vocab_mechanics_token_chars_jsonl_gz",
        "worst_bucket": "text/word",
        "worst_gap_bpb": -13.251871809589472,
        "worst_text": "v"
      },
      {
        "bytes": 41,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -31.586742683264536,
        "gap_bpb": -0.7704083581284034,
        "model_a_bpb": 0.5211139326521513,
        "model_b_bpb": 1.2915222907805546,
        "preview": ">>>\u2420token_chars(\"\"\"\u2420sexdate\"\"\")\u23ce['\u2420',\u2420's',\u2420'e',\u2420'x',\u2420'd',\u2420'a',\u2420't',\u2420'e']\u23ce",
        "row_index": 62411,
        "score_byte_end": 73,
        "score_byte_start": 32,
        "shard": "vocab_mechanics_token_chars_jsonl_gz",
        "worst_bucket": "text/word",
        "worst_gap_bpb": -9.797161281293645,
        "worst_text": "d"
      },
      {
        "bytes": 52,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -30.09996958803606,
        "gap_bpb": -0.5788455690006935,
        "model_a_bpb": 0.3904587188618888,
        "model_b_bpb": 0.9693042878625823,
        "preview": ">>>\u2420token_chars(\"\"\"\u21e5TokenName\"\"\")\u23ce['\\t',\u2420'T',\u2420'o',\u2420'k',\u2420'e',\u2420'n',\u2420'N',\u2420'a',\u2420'm',\u2420'e']\u23ce",
        "row_index": 40271,
        "score_byte_end": 86,
        "score_byte_start": 34,
        "shard": "vocab_mechanics_token_chars_jsonl_gz",
        "worst_bucket": "text/word",
        "worst_gap_bpb": -11.864086830524382,
        "worst_text": "N"
      },
      {
        "bytes": 61,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -28.704538416172873,
        "gap_bpb": -0.4705662035438176,
        "model_a_bpb": 0.4451376588325769,
        "model_b_bpb": 0.9157038623763945,
        "preview": ">>>\u2420token_chars(\"\"\"\u2420swingerclub\"\"\")\u23ce['\u2420',\u2420's',\u2420'w',\u2420'i',\u2420'n',\u2420'g',\u2420'e',\u2420'r',\u2420'c',\u2420'l',\u2420'u',\u2420'b']\u23ce",
        "row_index": 46457,
        "score_byte_end": 97,
        "score_byte_start": 36,
        "shard": "vocab_mechanics_token_chars_jsonl_gz",
        "worst_bucket": "text/word",
        "worst_gap_bpb": -10.56935839617078,
        "worst_text": "s"
      },
      {
        "bytes": 36,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -28.320285705530345,
        "gap_bpb": -0.7866746029313985,
        "model_a_bpb": 0.5760298547433316,
        "model_b_bpb": 1.36270445767473,
        "preview": ">>>\u2420token_chars(\"\"\"erdings\"\"\")\u23ce['e',\u2420'r',\u2420'd',\u2420'i',\u2420'n',\u2420'g',\u2420's']\u23ce",
        "row_index": 69096,
        "score_byte_end": 67,
        "score_byte_start": 31,
        "shard": "vocab_mechanics_token_chars_jsonl_gz",
        "worst_bucket": "text/word",
        "worst_gap_bpb": -15.398426917009221,
        "worst_text": "i"
      },
      {
        "bytes": 56,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -28.0963786707817,
        "gap_bpb": -0.5017210476925303,
        "model_a_bpb": 0.6098481791157294,
        "model_b_bpb": 1.1115692268082598,
        "preview": ">>>\u2420token_chars(\"\"\"\u2420sextreffen\"\"\")\u23ce['\u2420',\u2420's',\u2420'e',\u2420'x',\u2420't',\u2420'r',\u2420'e',\u2420'f',\u2420'f',\u2420'e',\u2420'n']\u23ce",
        "row_index": 97172,
        "score_byte_end": 91,
        "score_byte_start": 35,
        "shard": "vocab_mechanics_token_chars_jsonl_gz",
        "worst_bucket": "text/word",
        "worst_gap_bpb": -12.100202671807185,
        "worst_text": "x"
      },
      {
        "bytes": 41,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -26.988059834690095,
        "gap_bpb": -0.6582453618217097,
        "model_a_bpb": 0.3627697631246915,
        "model_b_bpb": 1.0210151249464012,
        "preview": ">>>\u2420token_chars(\"\"\"\u2420eskorte\"\"\")\u23ce['\u2420',\u2420'e',\u2420's',\u2420'k',\u2420'o',\u2420'r',\u2420't',\u2420'e']\u23ce",
        "row_index": 38820,
        "score_byte_end": 73,
        "score_byte_start": 32,
        "shard": "vocab_mechanics_token_chars_jsonl_gz",
        "worst_bucket": "text/word",
        "worst_gap_bpb": -8.980833762170414,
        "worst_text": "e"
      },
      {
        "bytes": 16,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -25.579744813639213,
        "gap_bpb": -1.5987340508524508,
        "model_a_bpb": 2.3710021730665622,
        "model_b_bpb": 3.969736223919013,
        "preview": "(?:['(',\u2420'?',\u2420':']\u23ce",
        "row_index": 67627,
        "score_byte_end": 19,
        "score_byte_start": 3,
        "shard": "vocab_mechanics_token_chars_jsonl_gz",
        "worst_bucket": "text/punctuation",
        "worst_gap_bpb": -3.6986961752942684,
        "worst_text": "['(',"
      },
      {
        "bytes": 76,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -23.554025044557626,
        "gap_bpb": -0.3099213821652319,
        "model_a_bpb": 0.8751303960154092,
        "model_b_bpb": 1.1850517781806413,
        "preview": ">>>\u2420token_chars(\"\"\"\u2420thuisontvangst\"\"\")\u23ce['\u2420',\u2420't',\u2420'h',\u2420'u',\u2420'i',\u2420's',\u2420'o',\u2420'n',\u2420't',\u2420'v',\u2420'a',\u2420'n',\u2420'g',\u2420's',\u2420't']\u23ce",
        "row_index": 78741,
        "score_byte_end": 115,
        "score_byte_start": 39,
        "shard": "vocab_mechanics_token_chars_jsonl_gz",
        "worst_bucket": "text/word",
        "worst_gap_bpb": -10.833787053334536,
        "worst_text": "t"
      },
      {
        "bytes": 91,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -21.98111622414017,
        "gap_bpb": -0.24155072773780406,
        "model_a_bpb": 0.566592210737716,
        "model_b_bpb": 0.8081429384755201,
        "preview": "\u2026oken_chars(\"\"\"MASConstraintMaker\"\"\")\u23ce['M',\u2420'A',\u2420'S',\u2420'C',\u2420'o',\u2420'n',\u2420's',\u2420't',\u2420'r',\u2420'a',\u2420'i',\u2420'n',\u2420't',\u2420'M',\u2420'a',\u2420'k',\u2420\u2026",
        "row_index": 59493,
        "score_byte_end": 133,
        "score_byte_start": 42,
        "shard": "vocab_mechanics_token_chars_jsonl_gz",
        "worst_bucket": "text/word",
        "worst_gap_bpb": -9.996542488265407,
        "worst_text": "o"
      },
      {
        "bytes": 41,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -21.845492292646394,
        "gap_bpb": -0.5328168851864974,
        "model_a_bpb": 0.8606758488520652,
        "model_b_bpb": 1.3934927340385626,
        "preview": ">>>\u2420token_chars(\"\"\"\u2420davidjl\"\"\")\u23ce['\u2420',\u2420'd',\u2420'a',\u2420'v',\u2420'i',\u2420'd',\u2420'j',\u2420'l']\u23ce",
        "row_index": 99172,
        "score_byte_end": 73,
        "score_byte_start": 32,
        "shard": "vocab_mechanics_token_chars_jsonl_gz",
        "worst_bucket": "text/word",
        "worst_gap_bpb": -8.041443799248313,
        "worst_text": "d"
      },
      {
        "bytes": 66,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -21.329466805441932,
        "gap_bpb": -0.3231737394763929,
        "model_a_bpb": 0.5829112801871499,
        "model_b_bpb": 0.9060850196635428,
        "preview": ">>>\u2420token_chars(\"\"\"\u2420prostituerte\"\"\")\u23ce['\u2420',\u2420'p',\u2420'r',\u2420'o',\u2420's',\u2420't',\u2420'i',\u2420't',\u2420'u',\u2420'e',\u2420'r',\u2420't',\u2420'e']\u23ce",
        "row_index": 51183,
        "score_byte_end": 103,
        "score_byte_start": 37,
        "shard": "vocab_mechanics_token_chars_jsonl_gz",
        "worst_bucket": "text/word",
        "worst_gap_bpb": -8.17830716284832,
        "worst_text": "p"
      },
      {
        "bytes": 71,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -20.994900175541645,
        "gap_bpb": -0.295702819373826,
        "model_a_bpb": 0.49251054805311906,
        "model_b_bpb": 0.788213367426945,
        "preview": "\u2026oproject\u23cetoken_chars\u2420raw\u2420block\u2420end\u23cecharacters:\u23ce['.',\u2420'd',\u2420'j',\u2420'a',\u2420'n',\u2420'g',\u2420'o',\u2420'p',\u2420'r',\u2420'o',\u2420'j',\u2420'e',\u2420'c',\u2420't']\u23ce",
        "row_index": 64019,
        "score_byte_end": 152,
        "score_byte_start": 81,
        "shard": "vocab_mechanics_token_chars_jsonl_gz",
        "worst_bucket": "text/word",
        "worst_gap_bpb": -9.587273998145974,
        "worst_text": "p"
      },
      {
        "bytes": 31,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -20.985472355344235,
        "gap_bpb": -0.6769507211401367,
        "model_a_bpb": 0.21095722553337073,
        "model_b_bpb": 0.8879079466735075,
        "preview": ">>>\u2420token_chars(\"\"\"iosity\"\"\")\u23ce['i',\u2420'o',\u2420's',\u2420'i',\u2420't',\u2420'y']\u23ce",
        "row_index": 32887,
        "score_byte_end": 61,
        "score_byte_start": 30,
        "shard": "vocab_mechanics_token_chars_jsonl_gz",
        "worst_bucket": "text/word",
        "worst_gap_bpb": -17.438647908497366,
        "worst_text": "t"
      },
      {
        "bytes": 56,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -20.091554125639906,
        "gap_bpb": -0.3587777522435697,
        "model_a_bpb": 0.950894260785739,
        "model_b_bpb": 1.3096720130293087,
        "preview": ">>>\u2420token_chars(\"\"\"useRalative\"\"\")\u23ce['u',\u2420's',\u2420'e',\u2420'R',\u2420'a',\u2420'l',\u2420'a',\u2420't',\u2420'i',\u2420'v',\u2420'e']\u23ce",
        "row_index": 88747,
        "score_byte_end": 91,
        "score_byte_start": 35,
        "shard": "vocab_mechanics_token_chars_jsonl_gz",
        "worst_bucket": "text/word",
        "worst_gap_bpb": -5.3789699762908265,
        "worst_text": "u"
      },
      {
        "bytes": 16,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -20.079761480843256,
        "gap_bpb": -1.2549850925527035,
        "model_a_bpb": 3.2724589653326626,
        "model_b_bpb": 4.527444057885366,
        "preview": "]+\"[']',\u2420'+',\u2420'\"']\u23ce",
        "row_index": 53420,
        "score_byte_end": 19,
        "score_byte_start": 3,
        "shard": "vocab_mechanics_token_chars_jsonl_gz",
        "worst_bucket": "text/punctuation",
        "worst_gap_bpb": -2.8670464122369,
        "worst_text": "[']',"
      },
      {
        "bytes": 57,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -19.703113724783826,
        "gap_bpb": -0.34566866183831274,
        "model_a_bpb": 1.069128741357851,
        "model_b_bpb": 1.4147974031961636,
        "preview": ">>>\u2420token_chars(\"\"\"\u21e5NdrFcShort\"\"\")\u23ce['\\t',\u2420'N',\u2420'd',\u2420'r',\u2420'F',\u2420'c',\u2420'S',\u2420'h',\u2420'o',\u2420'r',\u2420't']\u23ce",
        "row_index": 75603,
        "score_byte_end": 92,
        "score_byte_start": 35,
        "shard": "vocab_mechanics_token_chars_jsonl_gz",
        "worst_bucket": "text/word",
        "worst_gap_bpb": -9.22504689062114,
        "worst_text": "N"
      },
      {
        "bytes": 17,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -19.340486488872823,
        "gap_bpb": -1.1376756758160484,
        "model_a_bpb": 2.675683663679205,
        "model_b_bpb": 3.813359339495254,
        "preview": "+'\\['+',\u2420\"'\",\u2420'\\\\']\u23ce",
        "row_index": 61900,
        "score_byte_end": 20,
        "score_byte_start": 3,
        "shard": "vocab_mechanics_token_chars_jsonl_gz",
        "worst_bucket": "text/punctuation",
        "worst_gap_bpb": -1.1268467012014844,
        "worst_text": "['+',"
      },
      {
        "bytes": 36,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -19.330324977922505,
        "gap_bpb": -0.5369534716089585,
        "model_a_bpb": 0.9607383724647478,
        "model_b_bpb": 1.4976918440737064,
        "preview": ">>>\u2420token_chars(\"\"\"\u2420bakeka\"\"\")\u23ce['\u2420',\u2420'b',\u2420'a',\u2420'k',\u2420'e',\u2420'k',\u2420'a']\u23ce",
        "row_index": 82343,
        "score_byte_end": 67,
        "score_byte_start": 31,
        "shard": "vocab_mechanics_token_chars_jsonl_gz",
        "worst_bucket": "text/word",
        "worst_gap_bpb": -12.894884411497161,
        "worst_text": "k"
      },
      {
        "bytes": 33,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -18.96483970408817,
        "gap_bpb": -0.5746921122450961,
        "model_a_bpb": 1.0285854853956289,
        "model_b_bpb": 1.603277597640725,
        "preview": ">>>\u2420token_chars(\"\"\"\u2420\"\"\"\u21b5\u23ce\"\"\")\u23ce['\u2420',\u2420'\"',\u2420'\"',\u2420'\"',\u2420'\\r',\u2420'\\n']\u23ce",
        "row_index": 26212,
        "score_byte_end": 63,
        "score_byte_start": 30,
        "shard": "vocab_mechanics_token_chars_jsonl_gz",
        "worst_bucket": "text/punctuation",
        "worst_gap_bpb": -1.3190386818776414,
        "worst_text": "'\"',"
      },
      {
        "bytes": 56,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -18.94483059755185,
        "gap_bpb": -0.33830054638485446,
        "model_a_bpb": 0.6290904155876866,
        "model_b_bpb": 0.967390961972541,
        "preview": ">>>\u2420token_chars(\"\"\"\u2420'../../../\"\"\")\u23ce['\u2420',\u2420\"'\",\u2420'.',\u2420'.',\u2420'/',\u2420'.',\u2420'.',\u2420'/',\u2420'.',\u2420'.',\u2420'/']\u23ce",
        "row_index": 18793,
        "score_byte_end": 91,
        "score_byte_start": 35,
        "shard": "vocab_mechanics_token_chars_jsonl_gz",
        "worst_bucket": "text/punctuation",
        "worst_gap_bpb": -1.8868300725675171,
        "worst_text": "'.',"
      },
      {
        "bytes": 61,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -18.66268572805092,
        "gap_bpb": -0.30594566767296594,
        "model_a_bpb": 0.35170897458976935,
        "model_b_bpb": 0.6576546422627352,
        "preview": ">>>\u2420token_chars(\"\"\"\u2420thaimassage\"\"\")\u23ce['\u2420',\u2420't',\u2420'h',\u2420'a',\u2420'i',\u2420'm',\u2420'a',\u2420's',\u2420's',\u2420'a',\u2420'g',\u2420'e']\u23ce",
        "row_index": 66360,
        "score_byte_end": 97,
        "score_byte_start": 36,
        "shard": "vocab_mechanics_token_chars_jsonl_gz",
        "worst_bucket": "text/word",
        "worst_gap_bpb": -7.122219382581346,
        "worst_text": "s"
      },
      {
        "bytes": 56,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -18.50125045338528,
        "gap_bpb": -0.33037947238188,
        "model_a_bpb": 0.3705031567903077,
        "model_b_bpb": 0.7008826291721878,
        "preview": ">>>\u2420token_chars(\"\"\"\u2420datingside\"\"\")\u23ce['\u2420',\u2420'd',\u2420'a',\u2420't',\u2420'i',\u2420'n',\u2420'g',\u2420's',\u2420'i',\u2420'd',\u2420'e']\u23ce",
        "row_index": 76462,
        "score_byte_end": 91,
        "score_byte_start": 35,
        "shard": "vocab_mechanics_token_chars_jsonl_gz",
        "worst_bucket": "text/word",
        "worst_gap_bpb": -8.319856279762408,
        "worst_text": "d"
      }
    ]
  },
  "top_literals": {
    "model_a_worse": [
      {
        "bucket": "text/punctuation",
        "bytes": 203408,
        "delta_bits": 443750.1394450762,
        "documents": 101704,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"\u2420));\u23ce\"\"\")\u23ce['\u2420',\u2420')',\u2420')',\u2420';',\u2420'\\n']\u23ce",
        "gap_bpb": 2.1815766314258838,
        "model_a_bpb": 2.439141114545987,
        "model_a_token_boundaries": "|['|",
        "model_b_bpb": 0.2575644831201035,
        "model_b_token_boundaries": "|['|",
        "name": "['"
      },
      {
        "bucket": "text/punctuation",
        "bytes": 245004,
        "delta_bits": 85428.96758206986,
        "documents": 122502,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"\u2420\u0159id\"\"\")\u23ce['\u2420',\u2420'\u0159',\u2420'i',\u2420'd']\u23ce",
        "gap_bpb": 0.34868397080076186,
        "model_a_bpb": 0.6167446005513763,
        "model_a_token_boundaries": "|']\u2026|",
        "model_b_bpb": 0.2680606297506144,
        "model_b_token_boundaries": "|']\u2026|",
        "name": "']"
      },
      {
        "bucket": "text/punctuation",
        "bytes": 1132918,
        "delta_bits": 75119.43201934149,
        "documents": 566459,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"CompanyName\"\"\")\u23ce['C',\u2420'o',\u2420'm',\u2420'p',\u2420'a',\u2420'n',\u2420'y',\u2420'N',\u2420'a',\u2420'm',\u2420'e']\u23ce",
        "gap_bpb": 0.06630615103594567,
        "model_a_bpb": 0.19605052026362926,
        "model_a_token_boundaries": "|',|",
        "model_b_bpb": 0.1297443692276836,
        "model_b_token_boundaries": "|',|",
        "name": "',"
      },
      {
        "bucket": "whitespace/single_space",
        "bytes": 673787,
        "delta_bits": 64733.577631564025,
        "documents": 673787,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"')?></\"\"\")\u23ce[\"'\",\u2420')',\u2420'?',\u2420'>',\u2420'<',\u2420'/']\u23ce",
        "gap_bpb": 0.09607424546861845,
        "model_a_bpb": 0.23532359319435836,
        "model_a_token_boundaries": "|\u2420\u2026|",
        "model_b_bpb": 0.1392493477257399,
        "model_b_token_boundaries": "|\u2420\u2026|",
        "name": "\u2420"
      },
      {
        "bucket": "whitespace/newline",
        "bytes": 129648,
        "delta_bits": 44921.619721292955,
        "documents": 129648,
        "example_dataset": "ppl_circuit_coverage/python_interpretation/arithmetic_repl",
        "example_doc_preview": ">>>\u2420(((15\u2420*\u2420-9)\u2420-\u2420(16\u2420*\u242017))\u2420*\u2420((7\u2420+\u2420-2)\u2420+\u2420(7\u2420-\u2420-2)))\u23ce-5698\u23ce",
        "gap_bpb": 0.3464891068222645,
        "model_a_bpb": 0.6364287039102098,
        "model_a_token_boundaries": "|\u23ce|",
        "model_b_bpb": 0.28993959708794514,
        "model_b_token_boundaries": "|\u23ce|",
        "name": "\u23ce"
      },
      {
        "bucket": "text/punctuation",
        "bytes": 30680,
        "delta_bits": 43871.92051601834,
        "documents": 6136,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": "\u2026in\u23ce.LayoutInflater\u23cetoken_chars\u2420raw\u2420block\u2420end\u23cecharacters:\u23ce['.',\u2420'L',\u2420'a',\u2420'y',\u2420'o',\u2420'u',\u2420't',\u2420'I',\u2420'n',\u2420'f',\u2420'l',\u2420'a',\u2420\u2026",
        "gap_bpb": 1.4299843714477947,
        "model_a_bpb": 3.264430417557586,
        "model_a_token_boundaries": "|['|.',|",
        "model_b_bpb": 1.8344460461097911,
        "model_b_token_boundaries": "|['|.',|",
        "name": "['.',"
      },
      {
        "bucket": "text/punctuation",
        "bytes": 580524,
        "delta_bits": 28848.22993266414,
        "documents": 580524,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": "token_chars\u2420raw\u2420block\u2420begin\u23ce(tokens\u23cetoken_chars\u2420raw\u2420block\u2420end\u23cecharacters:\u23ce['(',\u2420't',\u2420'o',\u2420'k',\u2420'e',\u2420'n',\u2420's']\u23ce",
        "gap_bpb": 0.04969343202462627,
        "model_a_bpb": 0.06191839227490312,
        "model_a_token_boundaries": "|\u2026'|",
        "model_b_bpb": 0.012224960250276843,
        "model_b_token_boundaries": "|\u2026'|",
        "name": "'"
      },
      {
        "bucket": "text/word",
        "bytes": 59097,
        "delta_bits": 18340.44972654296,
        "documents": 59097,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"\u2420\u010dl\u00e1nek\"\"\")\u23ce['\u2420',\u2420'\u010d',\u2420'l',\u2420'\u00e1',\u2420'n',\u2420'e',\u2420'k']\u23ce",
        "gap_bpb": 0.3103448521336609,
        "model_a_bpb": 0.40055106547769465,
        "model_a_token_boundaries": "|e|",
        "model_b_bpb": 0.09020621334403371,
        "model_b_token_boundaries": "|e|",
        "name": "e"
      },
      {
        "bucket": "text/word",
        "bytes": 32328,
        "delta_bits": 16968.21898114583,
        "documents": 32328,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"cassert\"\"\")\u23ce['c',\u2420'a',\u2420's',\u2420's',\u2420'e',\u2420'r',\u2420't']\u23ce",
        "gap_bpb": 0.5248768553930286,
        "model_a_bpb": 0.6924466019064888,
        "model_a_token_boundaries": "|s|",
        "model_b_bpb": 0.16756974651346013,
        "model_b_token_boundaries": "|s|",
        "name": "s"
      },
      {
        "bucket": "text/punctuation",
        "bytes": 13255,
        "delta_bits": 16285.520083523415,
        "documents": 2651,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": "((((['(',\u2420'(',\u2420'(',\u2420'(']\u23ce",
        "gap_bpb": 1.2286322205600464,
        "model_a_bpb": 3.202311490024768,
        "model_a_token_boundaries": "|['|(',|",
        "model_b_bpb": 1.9736792694647216,
        "model_b_token_boundaries": "|['|(',|",
        "name": "['(',"
      },
      {
        "bucket": "text/word",
        "bytes": 36782,
        "delta_bits": 14687.757248876707,
        "documents": 36782,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"\u2420judiciary\"\"\")\u23ce['\u2420',\u2420'j',\u2420'u',\u2420'd',\u2420'i',\u2420'c',\u2420'i',\u2420'a',\u2420'r',\u2420'y']\u23ce",
        "gap_bpb": 0.39931915743778773,
        "model_a_bpb": 0.575581611202348,
        "model_a_token_boundaries": "|r|",
        "model_b_bpb": 0.17626245376456026,
        "model_b_token_boundaries": "|r|",
        "name": "r"
      },
      {
        "bucket": "text/word",
        "bytes": 38381,
        "delta_bits": 14643.936363067933,
        "documents": 38381,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"erdings\"\"\")\u23ce['e',\u2420'r',\u2420'd',\u2420'i',\u2420'n',\u2420'g',\u2420's']\u23ce",
        "gap_bpb": 0.38154129290711375,
        "model_a_bpb": 0.5106702290933508,
        "model_a_token_boundaries": "|i|",
        "model_b_bpb": 0.129128936186237,
        "model_b_token_boundaries": "|i|",
        "name": "i"
      },
      {
        "bucket": "text/word",
        "bytes": 39260,
        "delta_bits": 14625.337618387304,
        "documents": 39260,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": "\u2026\u2420token_chars(\"\"\"TouchableOpacity\"\"\")\u23ce['T',\u2420'o',\u2420'u',\u2420'c',\u2420'h',\u2420'a',\u2420'b',\u2420'l',\u2420'e',\u2420'O',\u2420'p',\u2420'a',\u2420'c',\u2420'i',\u2420't',\u2420'y']\u23ce",
        "gap_bpb": 0.3725251558427739,
        "model_a_bpb": 0.482476325916132,
        "model_a_token_boundaries": "|a|",
        "model_b_bpb": 0.10995117007335807,
        "model_b_token_boundaries": "|a|",
        "name": "a"
      },
      {
        "bucket": "text/word",
        "bytes": 19695,
        "delta_bits": 14133.31696824524,
        "documents": 19695,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"syscall\"\"\")\u23ce['s',\u2420'y',\u2420's',\u2420'c',\u2420'a',\u2420'l',\u2420'l']\u23ce",
        "gap_bpb": 0.7176093916346911,
        "model_a_bpb": 0.9068497335407162,
        "model_a_token_boundaries": "|c|",
        "model_b_bpb": 0.18924034190602512,
        "model_b_token_boundaries": "|c|",
        "name": "c"
      },
      {
        "bucket": "text/punctuation",
        "bytes": 9570,
        "delta_bits": 13364.918952050082,
        "documents": 1914,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": "token_chars\u2420raw\u2420block\u2420begin\u23ce-sum\u23cetoken_chars\u2420raw\u2420block\u2420end\u23cecharacters:\u23ce['-',\u2420's',\u2420'u',\u2420'm']\u23ce",
        "gap_bpb": 1.396543255177647,
        "model_a_bpb": 3.4469778686490553,
        "model_a_token_boundaries": "|['|-',|",
        "model_b_bpb": 2.0504346134714084,
        "model_b_token_boundaries": "|['|-',|",
        "name": "['-',"
      },
      {
        "bucket": "text/word",
        "bytes": 39968,
        "delta_bits": 13155.102372264246,
        "documents": 39968,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"iosity\"\"\")\u23ce['i',\u2420'o',\u2420's',\u2420'i',\u2420't',\u2420'y']\u23ce",
        "gap_bpb": 0.3291408720042095,
        "model_a_bpb": 0.4826039313739552,
        "model_a_token_boundaries": "|t|",
        "model_b_bpb": 0.1534630593697457,
        "model_b_token_boundaries": "|t|",
        "name": "t"
      },
      {
        "bucket": "text/word",
        "bytes": 32640,
        "delta_bits": 11352.752071721312,
        "documents": 32640,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"AccessorType\"\"\")\u23ce['A',\u2420'c',\u2420'c',\u2420'e',\u2420's',\u2420's',\u2420'o',\u2420'r',\u2420'T',\u2420'y',\u2420'p',\u2420'e']\u23ce",
        "gap_bpb": 0.3478171590600892,
        "model_a_bpb": 0.4458797727074253,
        "model_a_token_boundaries": "|o|",
        "model_b_bpb": 0.09806261364733608,
        "model_b_token_boundaries": "|o|",
        "name": "o"
      },
      {
        "bucket": "text/word",
        "bytes": 24477,
        "delta_bits": 10210.751842432102,
        "documents": 24477,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"\u00e6lland\"\"\")\u23ce['\u00e6',\u2420'l',\u2420'l',\u2420'a',\u2420'n',\u2420'd']\u23ce",
        "gap_bpb": 0.41715699809748347,
        "model_a_bpb": 0.515081351106472,
        "model_a_token_boundaries": "|l|",
        "model_b_bpb": 0.09792435300898858,
        "model_b_token_boundaries": "|l|",
        "name": "l"
      },
      {
        "bucket": "text/word",
        "bytes": 38393,
        "delta_bits": 10014.061621741937,
        "documents": 38393,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"\u010dn\u011b\"\"\")\u23ce['\u010d',\u2420'n',\u2420'\u011b']\u23ce",
        "gap_bpb": 0.2608304019415502,
        "model_a_bpb": 0.36973692225902516,
        "model_a_token_boundaries": "|n|",
        "model_b_bpb": 0.10890652031747493,
        "model_b_token_boundaries": "|n|",
        "name": "n"
      },
      {
        "bucket": "text/word",
        "bytes": 14304,
        "delta_bits": 9758.243555635205,
        "documents": 14304,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": "\u2026_chars(\"\"\"\u2420sourceMappingURL\"\"\")\u23ce['\u2420',\u2420's',\u2420'o',\u2420'u',\u2420'r',\u2420'c',\u2420'e',\u2420'M',\u2420'a',\u2420'p',\u2420'p',\u2420'i',\u2420'n',\u2420'g',\u2420'U',\u2420'R',\u2420'L']\u23ce",
        "gap_bpb": 0.6822038279946312,
        "model_a_bpb": 0.8931681518200193,
        "model_a_token_boundaries": "|p|",
        "model_b_bpb": 0.2109643238253881,
        "model_b_token_boundaries": "|p|",
        "name": "p"
      },
      {
        "bucket": "text/word",
        "bytes": 18275,
        "delta_bits": 8736.250719904434,
        "documents": 18275,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"\u2420b\u00e5de\"\"\")\u23ce['\u2420',\u2420'b',\u2420'\u00e5',\u2420'd',\u2420'e']\u23ce",
        "gap_bpb": 0.4780438150426503,
        "model_a_bpb": 0.6252638383571578,
        "model_a_token_boundaries": "|d|",
        "model_b_bpb": 0.14722002331450756,
        "model_b_token_boundaries": "|d|",
        "name": "d"
      },
      {
        "bucket": "text/word",
        "bytes": 14032,
        "delta_bits": 8239.438037370499,
        "documents": 14032,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"m\u0131z\u0131\"\"\")\u23ce['m',\u2420'\u0131',\u2420'z',\u2420'\u0131']\u23ce",
        "gap_bpb": 0.5871891417738383,
        "model_a_bpb": 0.7204848951562249,
        "model_a_token_boundaries": "|m|",
        "model_b_bpb": 0.13329575338238667,
        "model_b_token_boundaries": "|m|",
        "name": "m"
      },
      {
        "bucket": "text/word",
        "bytes": 16440,
        "delta_bits": 7850.063917365917,
        "documents": 16440,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"\u2420astounding\"\"\")\u23ce['\u2420',\u2420'a',\u2420's',\u2420't',\u2420'o',\u2420'u',\u2420'n',\u2420'd',\u2420'i',\u2420'n',\u2420'g']\u23ce",
        "gap_bpb": 0.47749780519257407,
        "model_a_bpb": 0.6582107630720855,
        "model_a_token_boundaries": "|u|",
        "model_b_bpb": 0.18071295787951136,
        "model_b_token_boundaries": "|u|",
        "name": "u"
      },
      {
        "bucket": "text/punctuation",
        "bytes": 5013,
        "delta_bits": 7121.913096070022,
        "documents": 1671,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": "token_chars\u2420raw\u2420block\u2420begin\u23ce\u200c\u062a\u0648\u0627\u0646\u062f\u23cetoken_chars\u2420raw\u2420block\u2420end\u23cecharacters:\u23ce['\\u200c',\u2420'\u062a',\u2420'\u0648',\u2420'\u0627',\u2420'\u0646',\u2420'\u062f']\u23ce",
        "gap_bpb": 1.420688828260527,
        "model_a_bpb": 3.5422836366840995,
        "model_a_token_boundaries": "|['|\\|",
        "model_b_bpb": 2.1215948084235725,
        "model_b_token_boundaries": "|['|\\|",
        "name": "['\\"
      },
      {
        "bucket": "text/word",
        "bytes": 12019,
        "delta_bits": 6518.80610808933,
        "documents": 12019,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"\u2420innings\"\"\")\u23ce['\u2420',\u2420'i',\u2420'n',\u2420'n',\u2420'i',\u2420'n',\u2420'g',\u2420's']\u23ce",
        "gap_bpb": 0.5423750817946027,
        "model_a_bpb": 0.6810642354169306,
        "model_a_token_boundaries": "|g|",
        "model_b_bpb": 0.13868915362232787,
        "model_b_token_boundaries": "|g|",
        "name": "g"
      },
      {
        "bucket": "text/word",
        "bytes": 4705,
        "delta_bits": 5761.948493032326,
        "documents": 4705,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": "\u2026\u2420token_chars(\"\"\"\u2420activeClassName\"\"\")\u23ce['\u2420',\u2420'a',\u2420'c',\u2420't',\u2420'i',\u2420'v',\u2420'e',\u2420'C',\u2420'l',\u2420'a',\u2420's',\u2420's',\u2420'N',\u2420'a',\u2420'm',\u2420'e']\u23ce",
        "gap_bpb": 1.2246436754585177,
        "model_a_bpb": 1.803330525069691,
        "model_a_token_boundaries": "|C|",
        "model_b_bpb": 0.5786868496111734,
        "model_b_token_boundaries": "|C|",
        "name": "C"
      },
      {
        "bucket": "text/word",
        "bytes": 5232,
        "delta_bits": 5707.282418473308,
        "documents": 5232,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"\u2420weakSelf\"\"\")\u23ce['\u2420',\u2420'w',\u2420'e',\u2420'a',\u2420'k',\u2420'S',\u2420'e',\u2420'l',\u2420'f']\u23ce",
        "gap_bpb": 1.0908414408396996,
        "model_a_bpb": 1.5605691794733552,
        "model_a_token_boundaries": "|S|",
        "model_b_bpb": 0.4697277386336555,
        "model_b_token_boundaries": "|S|",
        "name": "S"
      },
      {
        "bucket": "text/non_ascii_word",
        "bytes": 3508,
        "delta_bits": 5598.900059627392,
        "documents": 1754,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"\u0438\u0442\u0438\u0441\u044f\"\"\")\u23ce['\u0438',\u2420'\u0442',\u2420'\u0438',\u2420'\u0441',\u2420'\u044f']\u23ce",
        "gap_bpb": 1.5960376452757674,
        "model_a_bpb": 1.6870969487952827,
        "model_a_token_boundaries": "|\u0438|",
        "model_b_bpb": 0.09105930351951516,
        "model_b_token_boundaries": "|\u0438|",
        "name": "\u0438"
      },
      {
        "bucket": "text/non_ascii_word",
        "bytes": 5314,
        "delta_bits": 5383.652082374187,
        "documents": 2657,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"\u2420\u043e\u0441\u0442\u0430\u043d\u043d\u0456\"\"\")\u23ce['\u2420',\u2420'\u043e',\u2420'\u0441',\u2420'\u0442',\u2420'\u0430',\u2420'\u043d',\u2420'\u043d',\u2420'\u0456']\u23ce",
        "gap_bpb": 1.0131072793327411,
        "model_a_bpb": 1.0787996979299042,
        "model_a_token_boundaries": "|\u043e|",
        "model_b_bpb": 0.06569241859716292,
        "model_b_token_boundaries": "|\u043e|",
        "name": "\u043e"
      },
      {
        "bucket": "text/non_ascii_word",
        "bytes": 4172,
        "delta_bits": 5099.042755005243,
        "documents": 2086,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"\u0430\u0434\u0436\u0435\u043d\u043d\u044f\"\"\")\u23ce['\u0430',\u2420'\u0434',\u2420'\u0436',\u2420'\u0435',\u2420'\u043d',\u2420'\u043d',\u2420'\u044f']\u23ce",
        "gap_bpb": 1.222205837728965,
        "model_a_bpb": 1.2630164049669717,
        "model_a_token_boundaries": "|\u0430|",
        "model_b_bpb": 0.040810567238006336,
        "model_b_token_boundaries": "|\u0430|",
        "name": "\u0430"
      },
      {
        "bucket": "text/punctuation",
        "bytes": 5780,
        "delta_bits": 5029.676262070801,
        "documents": 1156,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"/\uff0f\"\"\")\u23ce['/',\u2420'\uff0f']\u23ce",
        "gap_bpb": 0.8701862045105193,
        "model_a_bpb": 3.0090666555038244,
        "model_a_token_boundaries": "|['|/',|",
        "model_b_bpb": 2.1388804509933053,
        "model_b_token_boundaries": "|['|/',|",
        "name": "['/',"
      },
      {
        "bucket": "text/word",
        "bytes": 6170,
        "delta_bits": 4924.046464756485,
        "documents": 6170,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"ificador\"\"\")\u23ce['i',\u2420'f',\u2420'i',\u2420'c',\u2420'a',\u2420'd',\u2420'o',\u2420'r']\u23ce",
        "gap_bpb": 0.798062636103158,
        "model_a_bpb": 0.9774367239974306,
        "model_a_token_boundaries": "|f|",
        "model_b_bpb": 0.1793740878942727,
        "model_b_token_boundaries": "|f|",
        "name": "f"
      },
      {
        "bucket": "text/word",
        "bytes": 7439,
        "delta_bits": 4667.356187886373,
        "documents": 7439,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"ibbean\"\"\")\u23ce['i',\u2420'b',\u2420'b',\u2420'e',\u2420'a',\u2420'n']\u23ce",
        "gap_bpb": 0.6274171512147295,
        "model_a_bpb": 0.8212507614659832,
        "model_a_token_boundaries": "|b|",
        "model_b_bpb": 0.1938336102512536,
        "model_b_token_boundaries": "|b|",
        "name": "b"
      },
      {
        "bucket": "text/non_ascii_word",
        "bytes": 3832,
        "delta_bits": 4491.672456540456,
        "documents": 1916,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"\u0435\u0440\u0456\u0433\u0430\"\"\")\u23ce['\u0435',\u2420'\u0440',\u2420'\u0456',\u2420'\u0433',\u2420'\u0430']\u23ce",
        "gap_bpb": 1.1721483446086784,
        "model_a_bpb": 1.2358789408771738,
        "model_a_token_boundaries": "|\u0435|",
        "model_b_bpb": 0.06373059626849542,
        "model_b_token_boundaries": "|\u0435|",
        "name": "\u0435"
      },
      {
        "bucket": "text/non_ascii_word",
        "bytes": 2858,
        "delta_bits": 4329.756795419887,
        "documents": 1429,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"\u0441\u043b\u0456\u0434\"\"\")\u23ce['\u0441',\u2420'\u043b',\u2420'\u0456',\u2420'\u0434']\u23ce",
        "gap_bpb": 1.5149603902798763,
        "model_a_bpb": 1.6245709339315964,
        "model_a_token_boundaries": "|\u0441|",
        "model_b_bpb": 0.10961054365171988,
        "model_b_token_boundaries": "|\u0441|",
        "name": "\u0441"
      },
      {
        "bucket": "text/word",
        "bytes": 6020,
        "delta_bits": 4315.635364263937,
        "documents": 6020,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"avadoc\"\"\")\u23ce['a',\u2420'v',\u2420'a',\u2420'd',\u2420'o',\u2420'c']\u23ce",
        "gap_bpb": 0.7168829508744081,
        "model_a_bpb": 0.9080382595379008,
        "model_a_token_boundaries": "|v|",
        "model_b_bpb": 0.19115530866349262,
        "model_b_token_boundaries": "|v|",
        "name": "v"
      },
      {
        "bucket": "text/word",
        "bytes": 9723,
        "delta_bits": 4226.120881804469,
        "documents": 9723,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"\u2420trhu\"\"\")\u23ce['\u2420',\u2420't',\u2420'r',\u2420'h',\u2420'u']\u23ce",
        "gap_bpb": 0.43465194711554755,
        "model_a_bpb": 0.5366479824936785,
        "model_a_token_boundaries": "|h|",
        "model_b_bpb": 0.101996035378131,
        "model_b_token_boundaries": "|h|",
        "name": "h"
      },
      {
        "bucket": "text/word",
        "bytes": 12000,
        "delta_bits": 4060.13512726207,
        "documents": 1000,
        "example_dataset": "ppl_circuit_coverage/text_mechanics/line_rhythm",
        "example_doc_preview": "\u2026\u2420\u2420oxpkqip\u23ce\u2420\u2420\u2420mmvxzp\u23ce\u2420bveu\u23ceerdgv\u23ce\u2420\u2420\u2420\u2420bwdwd\u23ce\u2420\u2420\u2420\u2420r\u23ceresult:\u23ce{\"line_lengths\":[11,9,5,5,9,5],\"indent_widths\":[4,3,1,0,4,4]}\u23ce",
        "gap_bpb": 0.33834459393850586,
        "model_a_bpb": 0.39860733864909487,
        "model_a_token_boundaries": "|line|_lengths|",
        "model_b_bpb": 0.06026274471058906,
        "model_b_token_boundaries": "|line|_lengths|",
        "name": "line_lengths"
      },
      {
        "bucket": "text/non_ascii_word",
        "bytes": 3258,
        "delta_bits": 4010.765165320531,
        "documents": 1629,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"\u043d\u0438\u0446\u0442\u0432\u0430\"\"\")\u23ce['\u043d',\u2420'\u0438',\u2420'\u0446',\u2420'\u0442',\u2420'\u0432',\u2420'\u0430']\u23ce",
        "gap_bpb": 1.2310513091837112,
        "model_a_bpb": 1.3367567434225802,
        "model_a_token_boundaries": "|\u043d|",
        "model_b_bpb": 0.10570543423886883,
        "model_b_token_boundaries": "|\u043d|",
        "name": "\u043d"
      },
      {
        "bucket": "text/punctuation",
        "bytes": 27665,
        "delta_bits": 3893.062989984465,
        "documents": 5533,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": "__(['_',\u2420'_',\u2420'(']\u23ce",
        "gap_bpb": 0.14072159732457853,
        "model_a_bpb": 3.5905631938449623,
        "model_a_token_boundaries": "|['_|',|",
        "model_b_bpb": 3.449841596520384,
        "model_b_token_boundaries": "|['_|',|",
        "name": "['_',"
      }
    ],
    "model_b_worse": [
      {
        "bucket": "text/punctuation",
        "bytes": 6000,
        "delta_bits": -3898.879856195666,
        "documents": 2000,
        "example_dataset": "ppl_circuit_coverage/text_mechanics/line_rhythm",
        "example_doc_preview": "line_lengths\u2420and\u2420indent_widths\u2420for\u2420raw\u2420block:\u23ce\u2420v\u23ce\u2420gunsto\u23cej\u23ceresult:\u23ce{\"line_lengths\":[2,7,1],\"indent_widths\":[1,1,0]}\u23ce",
        "gap_bpb": -0.6498133093659443,
        "model_a_bpb": 0.32552469127555606,
        "model_a_token_boundaries": "|\":[|",
        "model_b_bpb": 0.9753380006415004,
        "model_b_token_boundaries": "|\":[|",
        "name": "\":["
      },
      {
        "bucket": "text/punctuation",
        "bytes": 676,
        "delta_bits": -258.1343836581339,
        "documents": 169,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": "(^)(['(',\u2420'^',\u2420')',\u2420'(']\u23ce",
        "gap_bpb": -0.381855597127417,
        "model_a_bpb": 2.2141167305075444,
        "model_a_token_boundaries": "|\u2026'('|]\u2026|",
        "model_b_bpb": 2.5959723276349616,
        "model_b_token_boundaries": "|\u2026'('|]\u2026|",
        "name": "'(']"
      },
      {
        "bucket": "text/punctuation",
        "bytes": 3748,
        "delta_bits": -232.51456937141245,
        "documents": 937,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"\":\"\",\"\"\"\")\u23ce['\"',\u2420':',\u2420'\"',\u2420'\"',\u2420',',\u2420'\"']\u23ce",
        "gap_bpb": -0.06203697155053694,
        "model_a_bpb": 0.8918430549849552,
        "model_a_token_boundaries": "|\u2026'\"|',|",
        "model_b_bpb": 0.9538800265354922,
        "model_b_token_boundaries": "|\u2026'\"|',|",
        "name": "'\"',"
      },
      {
        "bucket": "text/punctuation",
        "bytes": 455,
        "delta_bits": -184.25406525891407,
        "documents": 91,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": "+\"\\['+',\u2420'\"',\u2420'\\\\']\u23ce",
        "gap_bpb": -0.4049539895800309,
        "model_a_bpb": 1.5920655042884215,
        "model_a_token_boundaries": "|\u2026'\\\\|']\u2026|",
        "model_b_bpb": 1.9970194938684527,
        "model_b_token_boundaries": "|\u2026'\\\\|']\u2026|",
        "name": "'\\\\']"
      },
      {
        "bucket": "text/punctuation",
        "bytes": 2000,
        "delta_bits": -123.58003937390727,
        "documents": 1000,
        "example_dataset": "ppl_circuit_coverage/text_mechanics/line_rhythm",
        "example_doc_preview": "\u2026for\u2420raw\u2420block:\u23ce\u2420s\u23ce\u2420\u2420xsfcmdja\u23ce\u2420\u2420\u2420\u2420injswjh\u23ce\u2420gh\u23ce\u2420\u2420\u2420dk\u23ceresult:\u23ce{\"line_lengths\":[2,10,11,3,5],\"indent_widths\":[1,2,4,1,3]}\u23ce",
        "gap_bpb": -0.061790019686953634,
        "model_a_bpb": 0.3314920739868864,
        "model_a_token_boundaries": "|]}\u2026|",
        "model_b_bpb": 0.39328209367384,
        "model_b_token_boundaries": "|]}\u2026|",
        "name": "]}"
      },
      {
        "bucket": "text/punctuation",
        "bytes": 780,
        "delta_bits": -105.39428651807391,
        "documents": 195,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"')?></\"\"\")\u23ce[\"'\",\u2420')',\u2420'?',\u2420'>',\u2420'<',\u2420'/']\u23ce",
        "gap_bpb": -0.1351208801513768,
        "model_a_bpb": 0.7436571605434057,
        "model_a_token_boundaries": "|\u2026'<|',|",
        "model_b_bpb": 0.8787780406947824,
        "model_b_token_boundaries": "|\u2026'<|',|",
        "name": "'<',"
      },
      {
        "bucket": "text/punctuation",
        "bytes": 556,
        "delta_bits": -88.47874888446192,
        "documents": 139,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": "\u093e:['\u093e',\u2420':']\u23ce",
        "gap_bpb": -0.15913444043967973,
        "model_a_bpb": 1.4944931537309125,
        "model_a_token_boundaries": "|\u2026':|']\u2026|",
        "model_b_bpb": 1.6536275941705922,
        "model_b_token_boundaries": "|\u2026':|']\u2026|",
        "name": "':']"
      },
      {
        "bucket": "text/punctuation",
        "bytes": 428,
        "delta_bits": -83.77018308214105,
        "documents": 107,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": "#[['#',\u2420'[']\u23ce",
        "gap_bpb": -0.19572472682743236,
        "model_a_bpb": 2.2751061373613095,
        "model_a_token_boundaries": "|\u2026'[|']\u2026|",
        "model_b_bpb": 2.4708308641887418,
        "model_b_token_boundaries": "|\u2026'[|']\u2026|",
        "name": "'[']"
      },
      {
        "bucket": "text/punctuation",
        "bytes": 1060,
        "delta_bits": -78.11621129927626,
        "documents": 212,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": "]+\"[']',\u2420'+',\u2420'\"']\u23ce",
        "gap_bpb": -0.07369453896158137,
        "model_a_bpb": 5.737658388261221,
        "model_a_token_boundaries": "|[|']|',|",
        "model_b_bpb": 5.811352927222803,
        "model_b_token_boundaries": "|[|']|',|",
        "name": "[']',"
      },
      {
        "bucket": "text/punctuation",
        "bytes": 465,
        "delta_bits": -55.30142847861962,
        "documents": 93,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"\u2420'\\''\"\"\")\u23ce['\u2420',\u2420\"'\",\u2420'\\\\',\u2420\"'\",\u2420\"'\"]\u23ce",
        "gap_bpb": -0.11892780317982714,
        "model_a_bpb": 0.9515716059407733,
        "model_a_token_boundaries": "|\u2026'\\\\|',|",
        "model_b_bpb": 1.0704994091206006,
        "model_b_token_boundaries": "|\u2026'\\\\|',|",
        "name": "'\\\\',"
      },
      {
        "bucket": "text/punctuation",
        "bytes": 1284,
        "delta_bits": -46.18396542768645,
        "documents": 321,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"','=','\"\"\")\u23ce[\"'\",\u2420',',\u2420\"'\",\u2420'=',\u2420\"'\",\u2420',',\u2420\"'\"]\u23ce",
        "gap_bpb": -0.03596882042654708,
        "model_a_bpb": 0.6860598391818765,
        "model_a_token_boundaries": "|\u2026',',|",
        "model_b_bpb": 0.7220286596084236,
        "model_b_token_boundaries": "|\u2026',',|",
        "name": "',',"
      },
      {
        "bucket": "text/punctuation",
        "bytes": 356,
        "delta_bits": -41.47305988094185,
        "documents": 89,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"\u2420!!}\"\"\")\u23ce['\u2420',\u2420'!',\u2420'!',\u2420'}']\u23ce",
        "gap_bpb": -0.11649735921612878,
        "model_a_bpb": 0.5214367185337393,
        "model_a_token_boundaries": "|\u2026'|!',|",
        "model_b_bpb": 0.6379340777498681,
        "model_b_token_boundaries": "|\u2026'|!',|",
        "name": "'!',"
      },
      {
        "bucket": "text/punctuation",
        "bytes": 920,
        "delta_bits": -34.69051310237216,
        "documents": 230,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"({...\"\"\")\u23ce['(',\u2420'{',\u2420'.',\u2420'.',\u2420'.']\u23ce",
        "gap_bpb": -0.03770707945910017,
        "model_a_bpb": 1.2979744328808283,
        "model_a_token_boundaries": "|\u2026'.|']\u2026|",
        "model_b_bpb": 1.3356815123399286,
        "model_b_token_boundaries": "|\u2026'.|']\u2026|",
        "name": "'.']"
      },
      {
        "bucket": "text/non_ascii_word",
        "bytes": 12,
        "delta_bits": -27.253598811847734,
        "documents": 4,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"\u2420\u3147\u3147\"\"\")\u23ce['\u2420',\u2420'\u3147',\u2420'\u3147']\u23ce",
        "gap_bpb": -2.2711332343206445,
        "model_a_bpb": 0.19640504719858434,
        "model_a_token_boundaries": "|\u3147|",
        "model_b_bpb": 2.467538281519229,
        "model_b_token_boundaries": "|\u3147|",
        "name": "\u3147"
      },
      {
        "bucket": "text/punctuation",
        "bytes": 472,
        "delta_bits": -25.096936771320912,
        "documents": 118,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"\u2420<->\"\"\")\u23ce['\u2420',\u2420'<',\u2420'-',\u2420'>']\u23ce",
        "gap_bpb": -0.053171476210425665,
        "model_a_bpb": 0.8423117653360839,
        "model_a_token_boundaries": "|\u2026'>|']\u2026|",
        "model_b_bpb": 0.8954832415465096,
        "model_b_token_boundaries": "|\u2026'>|']\u2026|",
        "name": "'>']"
      },
      {
        "bucket": "text/punctuation",
        "bytes": 500,
        "delta_bits": -18.75635367825223,
        "documents": 125,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"\")==\"\"\")\u23ce['\"',\u2420')',\u2420'=',\u2420'=']\u23ce",
        "gap_bpb": -0.03751270735650446,
        "model_a_bpb": 1.411456317504864,
        "model_a_token_boundaries": "|\u2026'=|']\u2026|",
        "model_b_bpb": 1.4489690248613685,
        "model_b_token_boundaries": "|\u2026'=|']\u2026|",
        "name": "'=']"
      },
      {
        "bucket": "text/number",
        "bytes": 15,
        "delta_bits": -17.923975064902553,
        "documents": 3,
        "example_dataset": "ppl_circuit_coverage/text_mechanics/line_rhythm",
        "example_doc_preview": "\u2026dths\u2420for\u2420raw\u2420block:\u23ce\u2420\u2420\u2420\u2420fkyvvqdxwzv\u23ce\u2420\u2420\u2420\u2420gmaguuvs\u23ce\u2420\u2420jvknwnb\u23ceresult:\u23ce{\"line_lengths\":[15,12,9],\"indent_widths\":[4,4,2]}\u23ce",
        "gap_bpb": -1.1949316709935036,
        "model_a_bpb": 1.8816233185938662,
        "model_a_token_boundaries": "|4|,|4|,|2|",
        "model_b_bpb": 3.0765549895873696,
        "model_b_token_boundaries": "|4|,|4|,|2|",
        "name": "4,4,2"
      },
      {
        "bucket": "text/non_ascii_word",
        "bytes": 64,
        "delta_bits": -17.85020857801302,
        "documents": 32,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"cks\u00e5\"\"\")\u23ce['c',\u2420'k',\u2420's',\u2420'\u00e5']\u23ce",
        "gap_bpb": -0.27890950903145345,
        "model_a_bpb": 0.4995915439217331,
        "model_a_token_boundaries": "|\u00e5|",
        "model_b_bpb": 0.7785010529531865,
        "model_b_token_boundaries": "|\u00e5|",
        "name": "\u00e5"
      },
      {
        "bucket": "text/non_ascii_word",
        "bytes": 45,
        "delta_bits": -16.584141181412672,
        "documents": 15,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"\u30b7\u30e3\"\"\")\u23ce['\u30b7',\u2420'\u30e3']\u23ce",
        "gap_bpb": -0.3685364706980594,
        "model_a_bpb": 1.4817964048952619,
        "model_a_token_boundaries": "|\u30b7|",
        "model_b_bpb": 1.850332875593321,
        "model_b_token_boundaries": "|\u30b7|",
        "name": "\u30b7"
      },
      {
        "bucket": "text/non_ascii",
        "bytes": 18,
        "delta_bits": -14.765797564253182,
        "documents": 3,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"\uff0c\u300a\"\"\")\u23ce['\uff0c',\u2420'\u300a']\u23ce",
        "gap_bpb": -0.8203220869029546,
        "model_a_bpb": 0.3480528038340802,
        "model_a_token_boundaries": "|\u2026'|\u300a|']\u2026|",
        "model_b_bpb": 1.1683748907370348,
        "model_b_token_boundaries": "|\u2026'|\u300a|']\u2026|",
        "name": "'\u300a']"
      },
      {
        "bucket": "text/number",
        "bytes": 25,
        "delta_bits": -14.714185739128874,
        "documents": 5,
        "example_dataset": "ppl_circuit_coverage/text_mechanics/line_rhythm",
        "example_doc_preview": "\u2026s\u2420and\u2420indent_widths\u2420for\u2420raw\u2420block:\u23ce\u2420\u2420xqnlpoxpsh\u23ce\u2420lbah\u23ce\u2420shku\u23ceresult:\u23ce{\"line_lengths\":[12,5,5],\"indent_widths\":[2,1,1]}\u23ce",
        "gap_bpb": -0.588567429565155,
        "model_a_bpb": 1.7109711541819073,
        "model_a_token_boundaries": "|2|,|1|,|1|",
        "model_b_bpb": 2.2995385837470623,
        "model_b_token_boundaries": "|2|,|1|,|1|",
        "name": "2,1,1"
      },
      {
        "bucket": "text/punctuation",
        "bytes": 1012,
        "delta_bits": -14.67558368494851,
        "documents": 253,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"={({\"\"\")\u23ce['=',\u2420'{',\u2420'(',\u2420'{']\u23ce",
        "gap_bpb": -0.014501564906075603,
        "model_a_bpb": 0.6341321307935426,
        "model_a_token_boundaries": "|\u2026'{|',|",
        "model_b_bpb": 0.6486336956996182,
        "model_b_token_boundaries": "|\u2026'{|',|",
        "name": "'{',"
      },
      {
        "bucket": "text/non_ascii_word",
        "bytes": 102,
        "delta_bits": -14.473118635565802,
        "documents": 34,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"\u3061\u3087\u3063\u3068\"\"\")\u23ce['\u3061',\u2420'\u3087',\u2420'\u3063',\u2420'\u3068']\u23ce",
        "gap_bpb": -0.14189331995652746,
        "model_a_bpb": 0.9211018235440386,
        "model_a_token_boundaries": "|\u3063|",
        "model_b_bpb": 1.062995143500566,
        "model_b_token_boundaries": "|\u3063|",
        "name": "\u3063"
      },
      {
        "bucket": "text/non_ascii_word",
        "bytes": 3,
        "delta_bits": -14.41686362641653,
        "documents": 1,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"\u2420\u3161\"\"\")\u23ce['\u2420',\u2420'\u3161']\u23ce",
        "gap_bpb": -4.80562120880551,
        "model_a_bpb": 0.4076499040646158,
        "model_a_token_boundaries": "|\u3161|\u3161|",
        "model_b_bpb": 5.2132711128701255,
        "model_b_token_boundaries": "|\u3161|",
        "name": "\u3161"
      },
      {
        "bucket": "text/non_ascii",
        "bytes": 24,
        "delta_bits": -13.09258604306656,
        "documents": 4,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"\u2420\u3001\"\"\")\u23ce['\u2420',\u2420'\u3001']\u23ce",
        "gap_bpb": -0.5455244184611067,
        "model_a_bpb": 0.8156130957270212,
        "model_a_token_boundaries": "|\u2026'|\u3001|']\u2026|",
        "model_b_bpb": 1.361137514188128,
        "model_b_token_boundaries": "|\u2026'|\u3001|']\u2026|",
        "name": "'\u3001']"
      },
      {
        "bucket": "text/non_ascii_word",
        "bytes": 12,
        "delta_bits": -13.004038618320765,
        "documents": 4,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"\u65cf\u81ea\u6cbb\"\"\")\u23ce['\u65cf',\u2420'\u81ea',\u2420'\u6cbb']\u23ce",
        "gap_bpb": -1.0836698848600637,
        "model_a_bpb": 0.535001129389643,
        "model_a_token_boundaries": "|\u65cf|",
        "model_b_bpb": 1.6186710142497065,
        "model_b_token_boundaries": "|\u65cf|",
        "name": "\u65cf"
      },
      {
        "bucket": "text/non_ascii",
        "bytes": 6,
        "delta_bits": -11.993729151396613,
        "documents": 1,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"\u2420\u300b\"\"\")\u23ce['\u2420',\u2420'\u300b']\u23ce",
        "gap_bpb": -1.998954858566102,
        "model_a_bpb": 0.447952307718889,
        "model_a_token_boundaries": "|\u2026'|\u300b|']\u2026|",
        "model_b_bpb": 2.446907166284991,
        "model_b_token_boundaries": "|\u2026'|\u300b|']\u2026|",
        "name": "'\u300b']"
      },
      {
        "bucket": "text/non_ascii",
        "bytes": 6,
        "delta_bits": -11.394379910600977,
        "documents": 1,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"\u2420\u301c\"\"\")\u23ce['\u2420',\u2420'\u301c']\u23ce",
        "gap_bpb": -1.8990633184334962,
        "model_a_bpb": 0.46885565596322953,
        "model_a_token_boundaries": "|\u2026'|\u301c|']\u2026|",
        "model_b_bpb": 2.3679189743967255,
        "model_b_token_boundaries": "|\u2026'|\u301c|']\u2026|",
        "name": "'\u301c']"
      },
      {
        "bucket": "text/non_ascii",
        "bytes": 6,
        "delta_bits": -11.155885545673812,
        "documents": 1,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"\u2420\u0e3f\"\"\")\u23ce['\u2420',\u2420'\u0e3f']\u23ce",
        "gap_bpb": -1.8593142576123018,
        "model_a_bpb": 0.26343832671833545,
        "model_a_token_boundaries": "|\u2026'|\u0e3f|\u0e3f|']\u2026|",
        "model_b_bpb": 2.1227525843306374,
        "model_b_token_boundaries": "|\u2026'|\u0e3f|']\u2026|",
        "name": "'\u0e3f']"
      },
      {
        "bucket": "text/number",
        "bytes": 10,
        "delta_bits": -10.564167658798102,
        "documents": 2,
        "example_dataset": "ppl_circuit_coverage/text_mechanics/line_rhythm",
        "example_doc_preview": "\u2026\u2420and\u2420indent_widths\u2420for\u2420raw\u2420block:\u23ce\u2420\u2420\u2420yg\u23ce\u2420ktakh\u23ce\u2420\u2420\u2420\u2420mljedgdx\u23ceresult:\u23ce{\"line_lengths\":[5,6,12],\"indent_widths\":[3,1,4]}\u23ce",
        "gap_bpb": -1.0564167658798103,
        "model_a_bpb": 2.493211039022008,
        "model_a_token_boundaries": "|3|,|1|,|4|",
        "model_b_bpb": 3.549627804901818,
        "model_b_token_boundaries": "|3|,|1|,|4|",
        "name": "3,1,4"
      },
      {
        "bucket": "text/number",
        "bytes": 15,
        "delta_bits": -10.356205710677925,
        "documents": 3,
        "example_dataset": "ppl_circuit_coverage/text_mechanics/line_rhythm",
        "example_doc_preview": "\u2026and\u2420indent_widths\u2420for\u2420raw\u2420block:\u23ce\u2420\u2420\u2420\u2420wbaqly\u23ce\u2420\u2420\u2420\u2420yde\u23ce\u2420giipae\u23ceresult:\u23ce{\"line_lengths\":[10,7,7],\"indent_widths\":[4,4,1]}\u23ce",
        "gap_bpb": -0.690413714045195,
        "model_a_bpb": 2.259694494079332,
        "model_a_token_boundaries": "|4|,|4|,|1|",
        "model_b_bpb": 2.9501082081245267,
        "model_b_token_boundaries": "|4|,|4|,|1|",
        "name": "4,4,1"
      },
      {
        "bucket": "text/number",
        "bytes": 10,
        "delta_bits": -10.19387626327648,
        "documents": 2,
        "example_dataset": "ppl_circuit_coverage/text_mechanics/line_rhythm",
        "example_doc_preview": "\u2026ent_widths\u2420for\u2420raw\u2420block:\u23ce\u2420\u2420qbhvtre\u23ce\u2420\u2420\u2420\u2420gcqecjh\u23ce\u2420\u2420\u2420hbbpqry\u23ceresult:\u23ce{\"line_lengths\":[9,11,10],\"indent_widths\":[2,4,3]}\u23ce",
        "gap_bpb": -1.019387626327648,
        "model_a_bpb": 2.0464059272438133,
        "model_a_token_boundaries": "|2|,|4|,|3|",
        "model_b_bpb": 3.0657935535714613,
        "model_b_token_boundaries": "|2|,|4|,|3|",
        "name": "2,4,3"
      },
      {
        "bucket": "text/number",
        "bytes": 7,
        "delta_bits": -10.14757450282083,
        "documents": 1,
        "example_dataset": "ppl_circuit_coverage/text_mechanics/line_rhythm",
        "example_doc_preview": "\u2026\u2420block:\u23ce\u2420\u2420qnujq\u23ce\u2420\u2420\u2420\u2420bspjqvka\u23ce\u2420\u2420\u2420\u2420knspfj\u23ce\u2420\u2420kvwvxeocspu\u23ceresult:\u23ce{\"line_lengths\":[7,12,10,13],\"indent_widths\":[2,4,4,2]}\u23ce",
        "gap_bpb": -1.4496535004029758,
        "model_a_bpb": 1.6838000328442444,
        "model_a_token_boundaries": "|2|,|4|,|4|,|2|",
        "model_b_bpb": 3.1334535332472204,
        "model_b_token_boundaries": "|2|,|4|,|4|,|2|",
        "name": "2,4,4,2"
      },
      {
        "bucket": "text/non_ascii_word",
        "bytes": 18,
        "delta_bits": -9.759982329804068,
        "documents": 6,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"\u2420\u4e07\"\"\")\u23ce['\u2420',\u2420'\u4e07']\u23ce",
        "gap_bpb": -0.5422212405446705,
        "model_a_bpb": 0.735992574091716,
        "model_a_token_boundaries": "|\u4e07|",
        "model_b_bpb": 1.2782138146363866,
        "model_b_token_boundaries": "|\u4e07|",
        "name": "\u4e07"
      },
      {
        "bucket": "text/non_ascii_word",
        "bytes": 6,
        "delta_bits": -8.994083649558187,
        "documents": 2,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"\u2420\u53c3\"\"\")\u23ce['\u2420',\u2420'\u53c3']\u23ce",
        "gap_bpb": -1.499013941593031,
        "model_a_bpb": 0.28994936337170146,
        "model_a_token_boundaries": "|\u53c3|",
        "model_b_bpb": 1.7889633049647324,
        "model_b_token_boundaries": "|\u53c3|",
        "name": "\u53c3"
      },
      {
        "bucket": "text/non_ascii_word",
        "bytes": 9,
        "delta_bits": -8.796543430872605,
        "documents": 3,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"\u2420\ub9ce\uc774\"\"\")\u23ce['\u2420',\u2420'\ub9ce',\u2420'\uc774']\u23ce",
        "gap_bpb": -0.9773937145414006,
        "model_a_bpb": 0.9816063913559147,
        "model_a_token_boundaries": "|\ub9ce|\ub9ce|",
        "model_b_bpb": 1.9590001058973152,
        "model_b_token_boundaries": "|\ub9ce|",
        "name": "\ub9ce"
      },
      {
        "bucket": "text/number",
        "bytes": 15,
        "delta_bits": -8.66283766874243,
        "documents": 3,
        "example_dataset": "ppl_circuit_coverage/text_mechanics/line_rhythm",
        "example_doc_preview": "\u2026ths\u2420and\u2420indent_widths\u2420for\u2420raw\u2420block:\u23ce\u2420\u2420\u2420homy\u23cebscwglq\u23ce\u2420\u2420oopmv\u23ceresult:\u23ce{\"line_lengths\":[7,7,7],\"indent_widths\":[3,0,2]}\u23ce",
        "gap_bpb": -0.5775225112494954,
        "model_a_bpb": 2.1386614075811523,
        "model_a_token_boundaries": "|3|,|0|,|2|",
        "model_b_bpb": 2.716183918830647,
        "model_b_token_boundaries": "|3|,|0|,|2|",
        "name": "3,0,2"
      },
      {
        "bucket": "text/non_ascii",
        "bytes": 42,
        "delta_bits": -8.54654585944059,
        "documents": 7,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"\u2420\u2588\u2588\"\"\")\u23ce['\u2420',\u2420'\u2588',\u2420'\u2588']\u23ce",
        "gap_bpb": -0.20348918712953784,
        "model_a_bpb": 0.14161296905618306,
        "model_a_token_boundaries": "|\u2026'|\u2588|',|",
        "model_b_bpb": 0.34510215618572093,
        "model_b_token_boundaries": "|\u2026'|\u2588|',|",
        "name": "'\u2588',"
      },
      {
        "bucket": "text/number",
        "bytes": 10,
        "delta_bits": -8.50900273468782,
        "documents": 2,
        "example_dataset": "ppl_circuit_coverage/text_mechanics/line_rhythm",
        "example_doc_preview": "\u2026and\u2420indent_widths\u2420for\u2420raw\u2420block:\u23ce\u2420\u2420\u2420\u2420banx\u23ce\u2420bybttpq\u23ce\u2420\u2420juehpkg\u23ceresult:\u23ce{\"line_lengths\":[8,8,9],\"indent_widths\":[4,1,2]}\u23ce",
        "gap_bpb": -0.850900273468782,
        "model_a_bpb": 2.1314321634307687,
        "model_a_token_boundaries": "|4|,|1|,|2|",
        "model_b_bpb": 2.9823324368995507,
        "model_b_token_boundaries": "|4|,|1|,|2|",
        "name": "4,1,2"
      },
      {
        "bucket": "text/non_ascii",
        "bytes": 6,
        "delta_bits": -8.461832340024255,
        "documents": 1,
        "example_dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "example_doc_preview": ">>>\u2420token_chars(\"\"\"\u2420\u25d1\"\"\")\u23ce['\u2420',\u2420'\u25d1']\u23ce",
        "gap_bpb": -1.4103053900040425,
        "model_a_bpb": 0.5767172632779228,
        "model_a_token_boundaries": "|\u2026'|\u25d1|\u25d1|']\u2026|",
        "model_b_bpb": 1.9870226532819653,
        "model_b_token_boundaries": "|\u2026'|\u25d1|']\u2026|",
        "name": "'\u25d1']"
      }
    ]
  },
  "top_segments": {
    "model_a_worse": [
      {
        "bucket": "text/non_ascii",
        "bytes": 7,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 30.015899104585408,
        "doc_preview": ">>>\u2420token_chars(\"\"\"\uff0f\uff0f\uff0f\uff0f\uff0f\uff0f\uff0f\uff0f\"\"\")\u23ce['\uff0f',\u2420'\uff0f',\u2420'\uff0f',\u2420'\uff0f',\u2420'\uff0f',\u2420'\uff0f',\u2420'\uff0f',\u2420'\uff0f']\u23ce",
        "gap_bpb": 4.287985586369344,
        "text": "['\uff0f',"
      },
      {
        "bucket": "text/non_ascii_word",
        "bytes": 3,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 26.652121135262924,
        "doc_preview": ">>>\u2420token_chars(\"\"\"\uff86\uff86\uff86\uff86\"\"\")\u23ce['\uff86',\u2420'\uff86',\u2420'\uff86',\u2420'\uff86']\u23ce",
        "gap_bpb": 8.884040378420975,
        "text": "\uff86"
      },
      {
        "bucket": "text/non_ascii",
        "bytes": 7,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 26.61334575481668,
        "doc_preview": "token_chars\u2420raw\u2420block\u2420begin\u23ce\uff09\u306f\u23cetoken_chars\u2420raw\u2420block\u2420end\u23cecharacters:\u23ce['\uff09',\u2420'\u306f']\u23ce",
        "gap_bpb": 3.801906536402383,
        "text": "['\uff09',"
      },
      {
        "bucket": "text/non_ascii_word",
        "bytes": 2,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 25.82697314192486,
        "doc_preview": ">>>\u2420token_chars(\"\"\"\u0410\u0440\u0445\u0456\u0432\u043e\u0432\u0430\u043d\u043e\"\"\")\u23ce['\u0410',\u2420'\u0440',\u2420'\u0445',\u2420'\u0456',\u2420'\u0432',\u2420'\u043e',\u2420'\u0432',\u2420'\u0430',\u2420'\u043d',\u2420'\u043e']\u23ce",
        "gap_bpb": 12.91348657096243,
        "text": "\u0410"
      },
      {
        "bucket": "text/non_ascii_word",
        "bytes": 3,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 25.66281363788449,
        "doc_preview": ">>>\u2420token_chars(\"\"\"\u7f51\u520a\u4e0b\u8f7d\u6b21\u6570\"\"\")\u23ce['\u7f51',\u2420'\u520a',\u2420'\u4e0b',\u2420'\u8f7d',\u2420'\u6b21',\u2420'\u6570']\u23ce",
        "gap_bpb": 8.554271212628164,
        "text": "\u7f51"
      },
      {
        "bucket": "text/non_ascii_word",
        "bytes": 2,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 25.172112738442735,
        "doc_preview": ">>>\u2420token_chars(\"\"\"\u045fN\"\"\")\u23ce['\u045f',\u2420'N']\u23ce",
        "gap_bpb": 12.586056369221367,
        "text": "\u045f"
      },
      {
        "bucket": "text/non_ascii_word",
        "bytes": 3,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 25.15869533995303,
        "doc_preview": ">>>\u2420token_chars(\"\"\"\u0948.\u23ce\"\"\")\u23ce['\u0948',\u2420'.',\u2420'\\n']\u23ce",
        "gap_bpb": 8.386231779984344,
        "text": "\u0948"
      },
      {
        "bucket": "text/non_ascii_word",
        "bytes": 3,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 25.04081843105164,
        "doc_preview": ">>>\u2420token_chars(\"\"\"\u0948\u0902.\u23ce\"\"\")\u23ce['\u0948',\u2420'\u0902',\u2420'.',\u2420'\\n']\u23ce",
        "gap_bpb": 8.346939477017214,
        "text": "\u0948"
      },
      {
        "bucket": "text/non_ascii_word",
        "bytes": 3,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 25.021570132767085,
        "doc_preview": ">>>\u2420token_chars(\"\"\"\u2420\uac24\ub85c\uadf8\ub85c\"\"\")\u23ce['\u2420',\u2420'\uac24',\u2420'\ub85c',\u2420'\uadf8',\u2420'\ub85c']\u23ce",
        "gap_bpb": 8.340523377589028,
        "text": "\uac24"
      },
      {
        "bucket": "text/non_ascii",
        "bytes": 7,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 24.953127917029754,
        "doc_preview": "token_chars\u2420raw\u2420block\u2420begin\u23ce\uff09\u306e\u23cetoken_chars\u2420raw\u2420block\u2420end\u23cecharacters:\u23ce['\uff09',\u2420'\u306e']\u23ce",
        "gap_bpb": 3.564732559575679,
        "text": "['\uff09',"
      },
      {
        "bucket": "text/non_ascii_word",
        "bytes": 3,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 24.949333291821127,
        "doc_preview": ">>>\u2420token_chars(\"\"\"\u2420\ub9e4\ub9e4\uac00\"\"\")\u23ce['\u2420',\u2420'\ub9e4',\u2420'\ub9e4',\u2420'\uac00']\u23ce",
        "gap_bpb": 8.316444430607042,
        "text": "\ub9e4"
      },
      {
        "bucket": "text/non_ascii_word",
        "bytes": 2,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 24.930140027984805,
        "doc_preview": ">>>\u2420token_chars(\"\"\"\u2420\u043f\u0440\u043e\u0444\u0435\u0441\u0441\u0438\"\"\")\u23ce['\u2420',\u2420'\u043f',\u2420'\u0440',\u2420'\u043e',\u2420'\u0444',\u2420'\u0435',\u2420'\u0441',\u2420'\u0441',\u2420'\u0438']\u23ce",
        "gap_bpb": 12.465070013992403,
        "text": "\u043f"
      },
      {
        "bucket": "text/non_ascii_word",
        "bytes": 2,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 24.750584637039218,
        "doc_preview": ">>>\u2420token_chars(\"\"\"\u2420\u010ceskosloven\"\"\")\u23ce['\u2420',\u2420'\u010c',\u2420'e',\u2420's',\u2420'k',\u2420'o',\u2420's',\u2420'l',\u2420'o',\u2420'v',\u2420'e',\u2420'n']\u23ce",
        "gap_bpb": 12.375292318519609,
        "text": "\u010c"
      },
      {
        "bucket": "text/non_ascii_word",
        "bytes": 3,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 24.662171795883644,
        "doc_preview": ">>>\u2420token_chars(\"\"\"\u0940)\"\"\")\u23ce['\u0940',\u2420')']\u23ce",
        "gap_bpb": 8.220723931961215,
        "text": "\u0940"
      },
      {
        "bucket": "text/non_ascii",
        "bytes": 7,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 24.54252965705382,
        "doc_preview": ">>>\u2420token_chars(\"\"\"\u30fb\u2501\u30fb\u2501\"\"\")\u23ce['\u30fb',\u2420'\u2501',\u2420'\u30fb',\u2420'\u2501']\u23ce",
        "gap_bpb": 3.5060756652934026,
        "text": "['\u30fb',"
      },
      {
        "bucket": "text/non_ascii_word",
        "bytes": 3,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 24.421225477886065,
        "doc_preview": ">>>\u2420token_chars(\"\"\"\u0948?\u23ce\"\"\")\u23ce['\u0948',\u2420'?',\u2420'\\n']\u23ce",
        "gap_bpb": 8.140408492628689,
        "text": "\u0948"
      },
      {
        "bucket": "text/non_ascii_word",
        "bytes": 3,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 24.31674808127872,
        "doc_preview": ">>>\u2420token_chars(\"\"\"\u0e4c)\"\"\")\u23ce['\u0e4c',\u2420')']\u23ce",
        "gap_bpb": 8.105582693759574,
        "text": "\u0e4c"
      },
      {
        "bucket": "text/non_ascii_word",
        "bytes": 3,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 24.13190112813137,
        "doc_preview": ">>>\u2420token_chars(\"\"\"\u91cd\u8907\u91cd\u8907\"\"\")\u23ce['\u91cd',\u2420'\u8907',\u2420'\u91cd',\u2420'\u8907']\u23ce",
        "gap_bpb": 8.043967042710458,
        "text": "\u91cd"
      },
      {
        "bucket": "text/non_ascii",
        "bytes": 7,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 24.054733544585126,
        "doc_preview": ">>>\u2420token_chars(\"\"\"\u30fb\u2501\"\"\")\u23ce['\u30fb',\u2420'\u2501']\u23ce",
        "gap_bpb": 3.4363905063693037,
        "text": "['\u30fb',"
      },
      {
        "bucket": "text/non_ascii",
        "bytes": 6,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 24.011794292178916,
        "doc_preview": "token_chars\u2420raw\u2420block\u2420begin\u23ce\u00b1\u0638\u23cetoken_chars\u2420raw\u2420block\u2420end\u23cecharacters:\u23ce['\u00b1',\u2420'\u0638']\u23ce",
        "gap_bpb": 4.001965715363153,
        "text": "['\u00b1',"
      },
      {
        "bucket": "text/punctuation",
        "bytes": 5,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 23.977094728575533,
        "doc_preview": ">>>\u2420token_chars(\"\"\"'])){\u23ce\"\"\")\u23ce[\"'\",\u2420']',\u2420')',\u2420')',\u2420'{',\u2420'\\n']\u23ce",
        "gap_bpb": 4.795418945715107,
        "text": "[\"'\","
      },
      {
        "bucket": "text/non_ascii_word",
        "bytes": 3,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 23.949547235490922,
        "doc_preview": ">>>\u2420token_chars(\"\"\"\u0947-\"\"\")\u23ce['\u0947',\u2420'-']\u23ce",
        "gap_bpb": 7.983182411830307,
        "text": "\u0947"
      },
      {
        "bucket": "text/non_ascii_word",
        "bytes": 3,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 23.880675751057005,
        "doc_preview": ">>>\u2420token_chars(\"\"\"\uae00\uc0c1\uc704\"\"\")\u23ce['\uae00',\u2420'\uc0c1',\u2420'\uc704']\u23ce",
        "gap_bpb": 7.960225250352335,
        "text": "\uae00"
      },
      {
        "bucket": "text/punctuation",
        "bytes": 5,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 23.714394669035613,
        "doc_preview": ">>>\u2420token_chars(\"\"\"});\u23ce\u23ce\u23ce\"\"\")\u23ce['}',\u2420')',\u2420';',\u2420'\\n',\u2420'\\n',\u2420'\\n']\u23ce",
        "gap_bpb": 4.742878933807122,
        "text": "['}',"
      },
      {
        "bucket": "text/non_ascii_word",
        "bytes": 2,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 23.70330385184678,
        "doc_preview": ">>>\u2420token_chars(\"\"\"\u2420\u043c\u043e\u0432\u0438\"\"\")\u23ce['\u2420',\u2420'\u043c',\u2420'\u043e',\u2420'\u0432',\u2420'\u0438']\u23ce",
        "gap_bpb": 11.85165192592339,
        "text": "\u043c"
      },
      {
        "bucket": "text/punctuation",
        "bytes": 5,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 23.688644050687216,
        "doc_preview": ">>>\u2420token_chars(\"\"\"');\u21b5\u23ce\"\"\")\u23ce[\"'\",\u2420')',\u2420';',\u2420'\\r',\u2420'\\n']\u23ce",
        "gap_bpb": 4.737728810137443,
        "text": "[\"'\","
      },
      {
        "bucket": "text/non_ascii",
        "bytes": 7,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 23.58768903075218,
        "doc_preview": "token_chars\u2420raw\u2420block\u2420begin\u23ce\u300d\u3068\u23cetoken_chars\u2420raw\u2420block\u2420end\u23cecharacters:\u23ce['\u300d',\u2420'\u3068']\u23ce",
        "gap_bpb": 3.3696698615360257,
        "text": "['\u300d',"
      },
      {
        "bucket": "text/punctuation",
        "bytes": 5,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 23.44325094926894,
        "doc_preview": ">>>\u2420token_chars(\"\"\"})\u23ce\u23ce\u23ce\"\"\")\u23ce['}',\u2420')',\u2420'\\n',\u2420'\\n',\u2420'\\n']\u23ce",
        "gap_bpb": 4.688650189853788,
        "text": "['}',"
      },
      {
        "bucket": "text/non_ascii_word",
        "bytes": 3,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 23.400374986478226,
        "doc_preview": ">>>\u2420token_chars(\"\"\"\u91b4\u91b4\"\"\")\u23ce['\u91b4',\u2420'\u91b4']\u23ce",
        "gap_bpb": 7.800124995492742,
        "text": "\u91b4"
      },
      {
        "bucket": "text/non_ascii",
        "bytes": 6,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 23.304432228919946,
        "doc_preview": "token_chars\u2420raw\u2420block\u2420begin\u23ce\u00b1\u0637\u23cetoken_chars\u2420raw\u2420block\u2420end\u23cecharacters:\u23ce['\u00b1',\u2420'\u0637']\u23ce",
        "gap_bpb": 3.8840720381533242,
        "text": "['\u00b1',"
      },
      {
        "bucket": "text/non_ascii_word",
        "bytes": 3,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 23.296791899655368,
        "doc_preview": ">>>\u2420token_chars(\"\"\"\u093e)\"\"\")\u23ce['\u093e',\u2420')']\u23ce",
        "gap_bpb": 7.765597299885123,
        "text": "\u093e"
      },
      {
        "bucket": "text/non_ascii_word",
        "bytes": 3,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 23.251272907486186,
        "doc_preview": ">>>\u2420token_chars(\"\"\"\ub9cc\uc6d0\uc785\ub2c8\ub2e4\"\"\")\u23ce['\ub9cc',\u2420'\uc6d0',\u2420'\uc785',\u2420'\ub2c8',\u2420'\ub2e4']\u23ce",
        "gap_bpb": 7.750424302495396,
        "text": "\ub9cc"
      },
      {
        "bucket": "text/non_ascii_word",
        "bytes": 3,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 23.150821280770238,
        "doc_preview": ">>>\u2420token_chars(\"\"\"\u0948?\"\"\")\u23ce['\u0948',\u2420'?']\u23ce",
        "gap_bpb": 7.716940426923412,
        "text": "\u0948"
      },
      {
        "bucket": "text/punctuation",
        "bytes": 5,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 23.14840664435212,
        "doc_preview": ">>>\u2420token_chars(\"\"\"}))\u23ce\u23ce\"\"\")\u23ce['}',\u2420')',\u2420')',\u2420'\\n',\u2420'\\n']\u23ce",
        "gap_bpb": 4.629681328870424,
        "text": "['}',"
      },
      {
        "bucket": "text/punctuation",
        "bytes": 5,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 23.127782484860706,
        "doc_preview": ">>>\u2420token_chars(\"\"\"}\")\u23ce\u23ce\"\"\")\u23ce['}',\u2420'\"',\u2420')',\u2420'\\n',\u2420'\\n']\u23ce",
        "gap_bpb": 4.625556496972141,
        "text": "['}',"
      },
      {
        "bucket": "text/non_ascii_word",
        "bytes": 2,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 23.020601520981778,
        "doc_preview": ">>>\u2420token_chars(\"\"\"\u2420\u0437\u0435\u043c\u043b\u0456\"\"\")\u23ce['\u2420',\u2420'\u0437',\u2420'\u0435',\u2420'\u043c',\u2420'\u043b',\u2420'\u0456']\u23ce",
        "gap_bpb": 11.510300760490889,
        "text": "\u0437"
      },
      {
        "bucket": "text/non_ascii_word",
        "bytes": 3,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 22.99818186261529,
        "doc_preview": ">>>\u2420token_chars(\"\"\"\u2420\ub4f1\ub85d\ub300\ud589\"\"\")\u23ce['\u2420',\u2420'\ub4f1',\u2420'\ub85d',\u2420'\ub300',\u2420'\ud589']\u23ce",
        "gap_bpb": 7.666060620871764,
        "text": "\ub4f1"
      },
      {
        "bucket": "text/non_ascii_word",
        "bytes": 2,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 22.8909045884519,
        "doc_preview": ">>>\u2420token_chars(\"\"\"\u2420\u0434\u0438\u0437\u0430\"\"\")\u23ce['\u2420',\u2420'\u0434',\u2420'\u0438',\u2420'\u0437',\u2420'\u0430']\u23ce",
        "gap_bpb": 11.44545229422595,
        "text": "\u0434"
      },
      {
        "bucket": "text/non_ascii_word",
        "bytes": 3,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 22.755346487180585,
        "doc_preview": ">>>\u2420token_chars(\"\"\"\u2420\u795e\u9a6c\u6536\u5f55\"\"\")\u23ce['\u2420',\u2420'\u795e',\u2420'\u9a6c',\u2420'\u6536',\u2420'\u5f55']\u23ce",
        "gap_bpb": 7.585115495726861,
        "text": "\u795e"
      },
      {
        "bucket": "text/non_ascii_word",
        "bytes": 2,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": 22.70859472337008,
        "doc_preview": ">>>\u2420token_chars(\"\"\"\u2420\u0437\u0434\u0430\u0442\"\"\")\u23ce['\u2420',\u2420'\u0437',\u2420'\u0434',\u2420'\u0430',\u2420'\u0442']\u23ce",
        "gap_bpb": 11.35429736168504,
        "text": "\u0437"
      }
    ],
    "model_b_worse": [
      {
        "bucket": "text/word",
        "bytes": 1,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -22.34676523973404,
        "doc_preview": ">>>\u2420token_chars(\"\"\"nestjs\"\"\")\u23ce['n',\u2420'e',\u2420's',\u2420't',\u2420'j',\u2420's']\u23ce",
        "gap_bpb": -22.34676523973404,
        "text": "j"
      },
      {
        "bucket": "text/word",
        "bytes": 1,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -19.706595244827643,
        "doc_preview": ">>>\u2420token_chars(\"\"\"AccessorType\"\"\")\u23ce['A',\u2420'c',\u2420'c',\u2420'e',\u2420's',\u2420's',\u2420'o',\u2420'r',\u2420'T',\u2420'y',\u2420'p',\u2420'e']\u23ce",
        "gap_bpb": -19.706595244827643,
        "text": "o"
      },
      {
        "bucket": "text/word",
        "bytes": 1,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -19.0426318963756,
        "doc_preview": ">>>\u2420token_chars(\"\"\"\u2420judiciary\"\"\")\u23ce['\u2420',\u2420'j',\u2420'u',\u2420'd',\u2420'i',\u2420'c',\u2420'i',\u2420'a',\u2420'r',\u2420'y']\u23ce",
        "gap_bpb": -19.0426318963756,
        "text": "r"
      },
      {
        "bucket": "text/word",
        "bytes": 1,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -18.984533480434127,
        "doc_preview": ">>>\u2420token_chars(\"\"\"buquerque\"\"\")\u23ce['b',\u2420'u',\u2420'q',\u2420'u',\u2420'e',\u2420'r',\u2420'q',\u2420'u',\u2420'e']\u23ce",
        "gap_bpb": -18.984533480434127,
        "text": "q"
      },
      {
        "bucket": "text/punctuation",
        "bytes": 5,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -18.49348087647134,
        "doc_preview": "(?:['(',\u2420'?',\u2420':']\u23ce",
        "gap_bpb": -3.6986961752942684,
        "text": "['(',"
      },
      {
        "bucket": "text/word",
        "bytes": 1,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -17.850224486408226,
        "doc_preview": "\u2026_chars(\"\"\"\u2420sourceMappingURL\"\"\")\u23ce['\u2420',\u2420's',\u2420'o',\u2420'u',\u2420'r',\u2420'c',\u2420'e',\u2420'M',\u2420'a',\u2420'p',\u2420'p',\u2420'i',\u2420'n',\u2420'g',\u2420'U',\u2420'R',\u2420'L']\u23ce",
        "gap_bpb": -17.850224486408226,
        "text": "p"
      },
      {
        "bucket": "text/word",
        "bytes": 1,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -17.551718842837232,
        "doc_preview": ">>>\u2420token_chars(\"\"\"\u2420wannonce\"\"\")\u23ce['\u2420',\u2420'w',\u2420'a',\u2420'n',\u2420'n',\u2420'o',\u2420'n',\u2420'c',\u2420'e']\u23ce",
        "gap_bpb": -17.551718842837232,
        "text": "w"
      },
      {
        "bucket": "text/word",
        "bytes": 1,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -17.438647908497366,
        "doc_preview": ">>>\u2420token_chars(\"\"\"iosity\"\"\")\u23ce['i',\u2420'o',\u2420's',\u2420'i',\u2420't',\u2420'y']\u23ce",
        "gap_bpb": -17.438647908497366,
        "text": "t"
      },
      {
        "bucket": "text/word",
        "bytes": 1,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -17.33860224145431,
        "doc_preview": ">>>\u2420token_chars(\"\"\"\u2420SubLObject\"\"\")\u23ce['\u2420',\u2420'S',\u2420'u',\u2420'b',\u2420'L',\u2420'O',\u2420'b',\u2420'j',\u2420'e',\u2420'c',\u2420't']\u23ce",
        "gap_bpb": -17.33860224145431,
        "text": "O"
      },
      {
        "bucket": "text/word",
        "bytes": 1,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -16.373938482083343,
        "doc_preview": ">>>\u2420token_chars(\"\"\"TexCoord\"\"\")\u23ce['T',\u2420'e',\u2420'x',\u2420'C',\u2420'o',\u2420'o',\u2420'r',\u2420'd']\u23ce",
        "gap_bpb": -16.373938482083343,
        "text": "C"
      },
      {
        "bucket": "text/word",
        "bytes": 1,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -16.178016878116345,
        "doc_preview": ">>>\u2420token_chars(\"\"\"\u2420XCTAssert\"\"\")\u23ce['\u2420',\u2420'X',\u2420'C',\u2420'T',\u2420'A',\u2420's',\u2420's',\u2420'e',\u2420'r',\u2420't']\u23ce",
        "gap_bpb": -16.178016878116345,
        "text": "T"
      },
      {
        "bucket": "text/word",
        "bytes": 1,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -16.17427956612611,
        "doc_preview": ">>>\u2420token_chars(\"\"\"ificador\"\"\")\u23ce['i',\u2420'f',\u2420'i',\u2420'c',\u2420'a',\u2420'd',\u2420'o',\u2420'r']\u23ce",
        "gap_bpb": -16.17427956612611,
        "text": "f"
      },
      {
        "bucket": "text/punctuation",
        "bytes": 4,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -16.15122232507723,
        "doc_preview": ">>>\u2420token_chars(\"\"\"\u2420!!}\"\"\")\u23ce['\u2420',\u2420'!',\u2420'!',\u2420'}']\u23ce",
        "gap_bpb": -4.037805581269308,
        "text": "'!',"
      },
      {
        "bucket": "text/word",
        "bytes": 1,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -16.096573289921697,
        "doc_preview": ">>>\u2420token_chars(\"\"\"\u2420innings\"\"\")\u23ce['\u2420',\u2420'i',\u2420'n',\u2420'n',\u2420'i',\u2420'n',\u2420'g',\u2420's']\u23ce",
        "gap_bpb": -16.096573289921697,
        "text": "g"
      },
      {
        "bucket": "text/word",
        "bytes": 1,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -15.854450438609566,
        "doc_preview": ">>>\u2420token_chars(\"\"\"toBeInTheDocument\"\"\")\u23ce['t',\u2420'o',\u2420'B',\u2420'e',\u2420'I',\u2420'n',\u2420'T',\u2420'h',\u2420'e',\u2420'D',\u2420'o',\u2420'c',\u2420'u',\u2420'm',\u2420'e',\u2420'\u2026",
        "gap_bpb": -15.854450438609566,
        "text": "B"
      },
      {
        "bucket": "text/word",
        "bytes": 1,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -15.653688554916677,
        "doc_preview": ">>>\u2420token_chars(\"\"\"\u21e5RTHOOK\"\"\")\u23ce['\\t',\u2420'R',\u2420'T',\u2420'H',\u2420'O',\u2420'O',\u2420'K']\u23ce",
        "gap_bpb": -15.653688554916677,
        "text": "R"
      },
      {
        "bucket": "text/word",
        "bytes": 1,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -15.398426917009221,
        "doc_preview": ">>>\u2420token_chars(\"\"\"erdings\"\"\")\u23ce['e',\u2420'r',\u2420'd',\u2420'i',\u2420'n',\u2420'g',\u2420's']\u23ce",
        "gap_bpb": -15.398426917009221,
        "text": "i"
      },
      {
        "bucket": "text/word",
        "bytes": 1,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -15.354227203754828,
        "doc_preview": ">>>\u2420token_chars(\"\"\"\u2420astounding\"\"\")\u23ce['\u2420',\u2420'a',\u2420's',\u2420't',\u2420'o',\u2420'u',\u2420'n',\u2420'd',\u2420'i',\u2420'n',\u2420'g']\u23ce",
        "gap_bpb": -15.354227203754828,
        "text": "u"
      },
      {
        "bucket": "text/word",
        "bytes": 1,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -15.231571875137298,
        "doc_preview": ">>>\u2420token_chars(\"\"\"\u2420everlasting\"\"\")\u23ce['\u2420',\u2420'e',\u2420'v',\u2420'e',\u2420'r',\u2420'l',\u2420'a',\u2420's',\u2420't',\u2420'i',\u2420'n',\u2420'g']\u23ce",
        "gap_bpb": -15.231571875137298,
        "text": "l"
      },
      {
        "bucket": "text/word",
        "bytes": 1,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -15.199615617546755,
        "doc_preview": ">>>\u2420token_chars(\"\"\"uento\"\"\")\u23ce['u',\u2420'e',\u2420'n',\u2420't',\u2420'o']\u23ce",
        "gap_bpb": -15.199615617546755,
        "text": "o"
      },
      {
        "bucket": "text/word",
        "bytes": 1,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -15.157252721999251,
        "doc_preview": ">>>\u2420token_chars(\"\"\"ibbean\"\"\")\u23ce['i',\u2420'b',\u2420'b',\u2420'e',\u2420'a',\u2420'n']\u23ce",
        "gap_bpb": -15.157252721999251,
        "text": "b"
      },
      {
        "bucket": "text/word",
        "bytes": 1,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -15.010869473739048,
        "doc_preview": "\u2026\u2420token_chars(\"\"\"TouchableOpacity\"\"\")\u23ce['T',\u2420'o',\u2420'u',\u2420'c',\u2420'h',\u2420'a',\u2420'b',\u2420'l',\u2420'e',\u2420'O',\u2420'p',\u2420'a',\u2420'c',\u2420'i',\u2420't',\u2420'y']\u23ce",
        "gap_bpb": -15.010869473739048,
        "text": "a"
      },
      {
        "bucket": "text/word",
        "bytes": 1,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -14.911586512131363,
        "doc_preview": ">>>\u2420token_chars(\"\"\"arParams\"\"\")\u23ce['a',\u2420'r',\u2420'P',\u2420'a',\u2420'r',\u2420'a',\u2420'm',\u2420's']\u23ce",
        "gap_bpb": -14.911586512131363,
        "text": "P"
      },
      {
        "bucket": "text/non_ascii_word",
        "bytes": 3,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -14.828100629957717,
        "doc_preview": ">>>\u2420token_chars(\"\"\"\u2420\u3147\u3147\"\"\")\u23ce['\u2420',\u2420'\u3147',\u2420'\u3147']\u23ce",
        "gap_bpb": -4.942700209985905,
        "text": "\u3147"
      },
      {
        "bucket": "text/word",
        "bytes": 1,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -14.812444877267035,
        "doc_preview": ">>>\u2420token_chars(\"\"\"embrance\"\"\")\u23ce['e',\u2420'm',\u2420'b',\u2420'r',\u2420'a',\u2420'n',\u2420'c',\u2420'e']\u23ce",
        "gap_bpb": -14.812444877267035,
        "text": "n"
      },
      {
        "bucket": "text/non_ascii_word",
        "bytes": 3,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -14.601341848760468,
        "doc_preview": ">>>\u2420token_chars(\"\"\"\u2420\u0e1e\u0e24\u0e29\u0e20\u0e32\u0e04\u0e21\"\"\")\u23ce['\u2420',\u2420'\u0e1e',\u2420'\u0e24',\u2420'\u0e29',\u2420'\u0e20',\u2420'\u0e32',\u2420'\u0e04',\u2420'\u0e21']\u23ce",
        "gap_bpb": -4.867113949586822,
        "text": "\u0e20"
      },
      {
        "bucket": "text/word",
        "bytes": 1,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -14.587447886345172,
        "doc_preview": ">>>\u2420token_chars(\"\"\"igInteger\"\"\")\u23ce['i',\u2420'g',\u2420'I',\u2420'n',\u2420't',\u2420'e',\u2420'g',\u2420'e',\u2420'r']\u23ce",
        "gap_bpb": -14.587447886345172,
        "text": "g"
      },
      {
        "bucket": "text/word",
        "bytes": 1,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -14.573455033905674,
        "doc_preview": ">>>\u2420token_chars(\"\"\"tridges\"\"\")\u23ce['t',\u2420'r',\u2420'i',\u2420'd',\u2420'g',\u2420'e',\u2420's']\u23ce",
        "gap_bpb": -14.573455033905674,
        "text": "g"
      },
      {
        "bucket": "text/non_ascii_word",
        "bytes": 3,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -14.41686362641653,
        "doc_preview": ">>>\u2420token_chars(\"\"\"\u2420\u3161\"\"\")\u23ce['\u2420',\u2420'\u3161']\u23ce",
        "gap_bpb": -4.80562120880551,
        "text": "\u3161"
      },
      {
        "bucket": "text/word",
        "bytes": 1,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -14.412925309703107,
        "doc_preview": "\u2026_chars(\"\"\"\u2420TouchableOpacity\"\"\")\u23ce['\u2420',\u2420'T',\u2420'o',\u2420'u',\u2420'c',\u2420'h',\u2420'a',\u2420'b',\u2420'l',\u2420'e',\u2420'O',\u2420'p',\u2420'a',\u2420'c',\u2420'i',\u2420't',\u2420'y']\u23ce",
        "gap_bpb": -14.412925309703107,
        "text": "a"
      },
      {
        "bucket": "text/word",
        "bytes": 1,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -14.372839335466017,
        "doc_preview": ">>>\u2420token_chars(\"\"\"ISyntaxException\"\"\")\u23ce['I',\u2420'S',\u2420'y',\u2420'n',\u2420't',\u2420'a',\u2420'x',\u2420'E',\u2420'x',\u2420'c',\u2420'e',\u2420'p',\u2420't',\u2420'i',\u2420'o',\u2420'n\u2026",
        "gap_bpb": -14.372839335466017,
        "text": "I"
      },
      {
        "bucket": "text/non_ascii_word",
        "bytes": 2,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -14.341172341945299,
        "doc_preview": ">>>\u2420token_chars(\"\"\"ket\u00f8y\"\"\")\u23ce['k',\u2420'e',\u2420't',\u2420'\u00f8',\u2420'y']\u23ce",
        "gap_bpb": -7.1705861709726495,
        "text": "\u00f8"
      },
      {
        "bucket": "text/punctuation",
        "bytes": 5,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -14.335232061184499,
        "doc_preview": "]+\"[']',\u2420'+',\u2420'\"']\u23ce",
        "gap_bpb": -2.8670464122369,
        "text": "[']',"
      },
      {
        "bucket": "text/punctuation",
        "bytes": 4,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -14.277906975981557,
        "doc_preview": ">>>\u2420token_chars(\"\"\"')?></\"\"\")\u23ce[\"'\",\u2420')',\u2420'?',\u2420'>',\u2420'<',\u2420'/']\u23ce",
        "gap_bpb": -3.569476743995389,
        "text": "'<',"
      },
      {
        "bucket": "text/word",
        "bytes": 1,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -14.260321749930835,
        "doc_preview": ">>>\u2420token_chars(\"\"\"m\u0131z\u0131\"\"\")\u23ce['m',\u2420'\u0131',\u2420'z',\u2420'\u0131']\u23ce",
        "gap_bpb": -14.260321749930835,
        "text": "m"
      },
      {
        "bucket": "text/word",
        "bytes": 1,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -14.215481401301979,
        "doc_preview": ">>>\u2420token_chars(\"\"\"\u2420DJs\"\"\")\u23ce['\u2420',\u2420'D',\u2420'J',\u2420's']\u23ce",
        "gap_bpb": -14.215481401301979,
        "text": "J"
      },
      {
        "bucket": "text/non_ascii_word",
        "bytes": 3,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -14.210298059183266,
        "doc_preview": ">>>\u2420token_chars(\"\"\"\u0e15\u0e23\u0e07\"\"\")\u23ce['\u0e15',\u2420'\u0e23',\u2420'\u0e07']\u23ce",
        "gap_bpb": -4.736766019727756,
        "text": "\u0e07"
      },
      {
        "bucket": "text/word",
        "bytes": 1,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -14.072371885796622,
        "doc_preview": ">>>\u2420token_chars(\"\"\"ordial\"\"\")\u23ce['o',\u2420'r',\u2420'd',\u2420'i',\u2420'a',\u2420'l']\u23ce",
        "gap_bpb": -14.072371885796622,
        "text": "a"
      },
      {
        "bucket": "text/word",
        "bytes": 1,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -14.016524045573268,
        "doc_preview": ">>>\u2420token_chars(\"\"\"addafi\"\"\")\u23ce['a',\u2420'd',\u2420'd',\u2420'a',\u2420'f',\u2420'i']\u23ce",
        "gap_bpb": -14.016524045573268,
        "text": "f"
      },
      {
        "bucket": "text/word",
        "bytes": 1,
        "dataset": "ppl_circuit_coverage/vocab_mechanics/token_chars",
        "delta_bits": -13.989825372861999,
        "doc_preview": ">>>\u2420token_chars(\"\"\"caffold\"\"\")\u23ce['c',\u2420'a',\u2420'f',\u2420'f',\u2420'o',\u2420'l',\u2420'd']\u23ce",
        "gap_bpb": -13.989825372861999,
        "text": "o"
      }
    ]
  }
}