{
 "final_results": {
  "baseline_published": {
   "checkpoint": "/Users/jamartin/OE_lemmatiser_retrain/baseline_project/experiments/checkpoints/improved_v2/best_model.pt",
   "universe": "original",
   "dev": {
    "n": 10506,
    "greedy": {
     "lemma": 68.68,
     "lemma_notation_insensitive": 72.99,
     "pos": 82.12
    },
    "beam10": {
     "lemma": 69.06,
     "lemma_notation_insensitive": 73.46,
     "pos": 82.12
    },
    "oracle_in_beam10": 93.3,
    "breakdown_greedy": {
     "by_pos": {
      "NOUN": {
       "n": 3314,
       "acc": 62.67
      },
      "VERB": {
       "n": 3128,
       "acc": 73.95
      },
      "ADJ": {
       "n": 1561,
       "acc": 65.02
      },
      "PROPN": {
       "n": 1369,
       "acc": 77.87
      },
      "X": {
       "n": 418,
       "acc": 86.36
      },
      "ADV": {
       "n": 331,
       "acc": 50.76
      },
      "DET": {
       "n": 122,
       "acc": 57.38
      },
      "NUM": {
       "n": 97,
       "acc": 61.86
      },
      "ADP": {
       "n": 78,
       "acc": 51.28
      },
      "PRON": {
       "n": 64,
       "acc": 54.69
      }
     },
     "seen_form": {
      "n": 2149,
      "acc": 37.88
     },
     "unseen_form": {
      "n": 8357,
      "acc": 76.59
     },
     "gold_consistent": {
      "n": 170,
      "acc": 95.29
     },
     "gold_ambiguous": {
      "n": 125,
      "acc": 69.6
     },
     "gold_conflicting": {
      "n": 1854,
      "acc": 30.47
     }
    }
   },
   "test": {
    "n": 10507,
    "greedy": {
     "lemma": 67.58,
     "lemma_notation_insensitive": 72.13,
     "pos": 82.75
    },
    "beam10": {
     "lemma": 68.21,
     "lemma_notation_insensitive": 72.76,
     "pos": 82.75
    },
    "oracle_in_beam10": 92.98,
    "breakdown_greedy": {
     "by_pos": {
      "NOUN": {
       "n": 3230,
       "acc": 61.11
      },
      "VERB": {
       "n": 3126,
       "acc": 71.94
      },
      "ADJ": {
       "n": 1559,
       "acc": 62.92
      },
      "PROPN": {
       "n": 1395,
       "acc": 78.49
      },
      "X": {
       "n": 463,
       "acc": 85.53
      },
      "ADV": {
       "n": 356,
       "acc": 52.25
      },
      "DET": {
       "n": 136,
       "acc": 59.56
      },
      "PRON": {
       "n": 76,
       "acc": 51.32
      },
      "NUM": {
       "n": 73,
       "acc": 71.23
      },
      "ADP": {
       "n": 71,
       "acc": 50.7
      }
     },
     "seen_form": {
      "n": 2140,
      "acc": 36.73
     },
     "unseen_form": {
      "n": 8367,
      "acc": 75.48
     },
     "gold_consistent": {
      "n": 171,
      "acc": 97.66
     },
     "gold_ambiguous": {
      "n": 145,
      "acc": 64.14
     },
     "gold_conflicting": {
      "n": 1824,
      "acc": 28.84
     }
    }
   }
  },
  "C0_replication": {
   "checkpoint": "/Users/jamartin/OE_lemmatiser_retrain/experiments/C0_replication/best_model.pt",
   "universe": "original",
   "training_summary": {
    "name": "C0_replication",
    "aug": "synthetic",
    "aug_prob": 0.2,
    "extra_train": {},
    "universe": "original",
    "corpus_rows": 84050,
    "total_train_rows": 84050,
    "epochs": 80,
    "batch_size": 72,
    "lr": 0.00015,
    "label_smoothing": 0.15,
    "pos_weight": 0.2,
    "dropout": 0.15,
    "cycles": 4,
    "patience": 20,
    "seed": 42,
    "pretrained_encoder": true,
    "steps_per_epoch": 1168,
    "total_steps": 93440,
    "parameters": 26437278,
    "device": "mps",
    "started": "2026-08-25T13:21:00",
    "best_dev_lemma": 0.6872263468494194,
    "best_epoch": 75,
    "epochs_run": 80,
    "wall_seconds": 20520.1,
    "finished": "2026-08-25T19:03:00"
   },
   "dev": {
    "n": 10506,
    "greedy": {
     "lemma": 68.72,
     "lemma_notation_insensitive": 73.07,
     "pos": 82.38
    },
    "beam10": {
     "lemma": 69.12,
     "lemma_notation_insensitive": 73.58,
     "pos": 82.38
    },
    "oracle_in_beam10": 93.19,
    "breakdown_greedy": {
     "by_pos": {
      "NOUN": {
       "n": 3314,
       "acc": 62.73
      },
      "VERB": {
       "n": 3128,
       "acc": 73.95
      },
      "ADJ": {
       "n": 1561,
       "acc": 66.11
      },
      "PROPN": {
       "n": 1369,
       "acc": 77.57
      },
      "X": {
       "n": 418,
       "acc": 85.17
      },
      "ADV": {
       "n": 331,
       "acc": 48.04
      },
      "DET": {
       "n": 122,
       "acc": 59.84
      },
      "NUM": {
       "n": 97,
       "acc": 61.86
      },
      "ADP": {
       "n": 78,
       "acc": 51.28
      },
      "PRON": {
       "n": 64,
       "acc": 59.38
      }
     },
     "seen_form": {
      "n": 2149,
      "acc": 38.3
     },
     "unseen_form": {
      "n": 8357,
      "acc": 76.55
     },
     "gold_consistent": {
      "n": 170,
      "acc": 95.88
     },
     "gold_ambiguous": {
      "n": 125,
      "acc": 68.0
     },
     "gold_conflicting": {
      "n": 1854,
      "acc": 31.01
     }
    }
   },
   "test": {
    "n": 10507,
    "greedy": {
     "lemma": 67.41,
     "lemma_notation_insensitive": 72.16,
     "pos": 82.86
    },
    "beam10": {
     "lemma": 68.08,
     "lemma_notation_insensitive": 72.88,
     "pos": 82.86
    },
    "oracle_in_beam10": 93.39,
    "breakdown_greedy": {
     "by_pos": {
      "NOUN": {
       "n": 3230,
       "acc": 59.81
      },
      "VERB": {
       "n": 3126,
       "acc": 71.91
      },
      "ADJ": {
       "n": 1559,
       "acc": 64.34
      },
      "PROPN": {
       "n": 1395,
       "acc": 78.71
      },
      "X": {
       "n": 463,
       "acc": 85.75
      },
      "ADV": {
       "n": 356,
       "acc": 51.12
      },
      "DET": {
       "n": 136,
       "acc": 60.29
      },
      "PRON": {
       "n": 76,
       "acc": 50.0
      },
      "NUM": {
       "n": 73,
       "acc": 72.6
      },
      "ADP": {
       "n": 71,
       "acc": 53.52
      }
     },
     "seen_form": {
      "n": 2140,
      "acc": 36.5
     },
     "unseen_form": {
      "n": 8367,
      "acc": 75.32
     },
     "gold_consistent": {
      "n": 171,
      "acc": 99.42
     },
     "gold_ambiguous": {
      "n": 145,
      "acc": 66.21
     },
     "gold_conflicting": {
      "n": 1824,
      "acc": 28.23
     }
    }
   }
  },
  "C1_variant": {
   "checkpoint": "/Users/jamartin/OE_lemmatiser_retrain/experiments/C1_variant/best_model.pt",
   "universe": "original",
   "training_summary": {
    "name": "C1_variant",
    "aug": "attested",
    "aug_prob": 0.2,
    "extra_train": {},
    "universe": "original",
    "corpus_rows": 84050,
    "total_train_rows": 84050,
    "epochs": 80,
    "batch_size": 72,
    "lr": 0.00015,
    "label_smoothing": 0.15,
    "pos_weight": 0.2,
    "dropout": 0.15,
    "cycles": 4,
    "patience": 20,
    "seed": 42,
    "pretrained_encoder": true,
    "steps_per_epoch": 1168,
    "total_steps": 93440,
    "parameters": 26437278,
    "device": "mps",
    "started": "2026-08-25T13:21:24",
    "best_dev_lemma": 0.687797449076718,
    "best_epoch": 78,
    "epochs_run": 80,
    "wall_seconds": 20553.6,
    "finished": "2026-08-25T19:03:58",
    "augmenter_stats": {
     "attested_substitutions": 795715,
     "synthetic_fallbacks": 548603
    }
   },
   "dev": {
    "n": 10506,
    "greedy": {
     "lemma": 68.78,
     "lemma_notation_insensitive": 73.28,
     "pos": 82.53
    },
    "beam10": {
     "lemma": 69.08,
     "lemma_notation_insensitive": 73.58,
     "pos": 82.53
    },
    "oracle_in_beam10": 93.27,
    "breakdown_greedy": {
     "by_pos": {
      "NOUN": {
       "n": 3314,
       "acc": 62.82
      },
      "VERB": {
       "n": 3128,
       "acc": 73.63
      },
      "ADJ": {
       "n": 1561,
       "acc": 65.53
      },
      "PROPN": {
       "n": 1369,
       "acc": 77.65
      },
      "X": {
       "n": 418,
       "acc": 87.56
      },
      "ADV": {
       "n": 331,
       "acc": 51.36
      },
      "DET": {
       "n": 122,
       "acc": 59.84
      },
      "NUM": {
       "n": 97,
       "acc": 61.86
      },
      "ADP": {
       "n": 78,
       "acc": 48.72
      },
      "PRON": {
       "n": 64,
       "acc": 60.94
      }
     },
     "seen_form": {
      "n": 2149,
      "acc": 37.46
     },
     "unseen_form": {
      "n": 8357,
      "acc": 76.83
     },
     "gold_consistent": {
      "n": 170,
      "acc": 95.29
     },
     "gold_ambiguous": {
      "n": 125,
      "acc": 68.8
     },
     "gold_conflicting": {
      "n": 1854,
      "acc": 30.04
     }
    }
   },
   "test": {
    "n": 10507,
    "greedy": {
     "lemma": 68.08,
     "lemma_notation_insensitive": 72.78,
     "pos": 82.68
    },
    "beam10": {
     "lemma": 68.18,
     "lemma_notation_insensitive": 72.97,
     "pos": 82.68
    },
    "oracle_in_beam10": 93.26,
    "breakdown_greedy": {
     "by_pos": {
      "NOUN": {
       "n": 3230,
       "acc": 61.52
      },
      "VERB": {
       "n": 3126,
       "acc": 72.26
      },
      "ADJ": {
       "n": 1559,
       "acc": 64.21
      },
      "PROPN": {
       "n": 1395,
       "acc": 78.35
      },
      "X": {
       "n": 463,
       "acc": 87.04
      },
      "ADV": {
       "n": 356,
       "acc": 51.69
      },
      "DET": {
       "n": 136,
       "acc": 63.97
      },
      "PRON": {
       "n": 76,
       "acc": 47.37
      },
      "NUM": {
       "n": 73,
       "acc": 69.86
      },
      "ADP": {
       "n": 71,
       "acc": 56.34
      }
     },
     "seen_form": {
      "n": 2140,
      "acc": 36.4
     },
     "unseen_form": {
      "n": 8367,
      "acc": 76.18
     },
     "gold_consistent": {
      "n": 171,
      "acc": 99.42
     },
     "gold_ambiguous": {
      "n": 145,
      "acc": 64.83
     },
     "gold_conflicting": {
      "n": 1824,
      "acc": 28.23
     }
    }
   }
  },
  "C2_lexicon": {
   "checkpoint": "/Users/jamartin/OE_lemmatiser_retrain/experiments/C2_lexicon/best_model.pt",
   "universe": "original",
   "training_summary": {
    "name": "C2_lexicon",
    "aug": "synthetic",
    "aug_prob": 0.2,
    "extra_train": {
     "lexicon_safe.tsv": 122076
    },
    "universe": "original",
    "corpus_rows": 84050,
    "total_train_rows": 206126,
    "epochs": 80,
    "batch_size": 72,
    "lr": 0.00015,
    "label_smoothing": 0.15,
    "pos_weight": 0.2,
    "dropout": 0.15,
    "cycles": 4,
    "patience": 20,
    "seed": 42,
    "pretrained_encoder": true,
    "steps_per_epoch": 2863,
    "total_steps": 229040,
    "parameters": 26437278,
    "device": "mps",
    "started": "2026-08-25T19:03:10",
    "best_dev_lemma": 0.4799162383399962,
    "best_epoch": 24,
    "epochs_run": 44,
    "wall_seconds": 23287.0,
    "finished": "2026-08-26T01:31:17"
   },
   "dev": {
    "n": 10506,
    "greedy": {
     "lemma": 47.99,
     "lemma_notation_insensitive": 61.19,
     "pos": 78.78
    },
    "beam10": {
     "lemma": 48.72,
     "lemma_notation_insensitive": 61.85,
     "pos": 78.78
    },
    "oracle_in_beam10": 90.75,
    "breakdown_greedy": {
     "by_pos": {
      "NOUN": {
       "n": 3314,
       "acc": 44.84
      },
      "VERB": {
       "n": 3128,
       "acc": 45.33
      },
      "ADJ": {
       "n": 1561,
       "acc": 42.66
      },
      "PROPN": {
       "n": 1369,
       "acc": 63.92
      },
      "X": {
       "n": 418,
       "acc": 77.75
      },
      "ADV": {
       "n": 331,
       "acc": 38.37
      },
      "DET": {
       "n": 122,
       "acc": 47.54
      },
      "NUM": {
       "n": 97,
       "acc": 29.9
      },
      "ADP": {
       "n": 78,
       "acc": 33.33
      },
      "PRON": {
       "n": 64,
       "acc": 34.38
      }
     },
     "seen_form": {
      "n": 2149,
      "acc": 32.11
     },
     "unseen_form": {
      "n": 8357,
      "acc": 52.08
     },
     "gold_consistent": {
      "n": 170,
      "acc": 71.18
     },
     "gold_ambiguous": {
      "n": 125,
      "acc": 40.8
     },
     "gold_conflicting": {
      "n": 1854,
      "acc": 27.94
     }
    }
   },
   "test": {
    "n": 10507,
    "greedy": {
     "lemma": 47.55,
     "lemma_notation_insensitive": 60.25,
     "pos": 78.92
    },
    "beam10": {
     "lemma": 48.42,
     "lemma_notation_insensitive": 60.95,
     "pos": 78.92
    },
    "oracle_in_beam10": 90.61,
    "breakdown_greedy": {
     "by_pos": {
      "NOUN": {
       "n": 3230,
       "acc": 45.39
      },
      "VERB": {
       "n": 3126,
       "acc": 42.87
      },
      "ADJ": {
       "n": 1559,
       "acc": 43.94
      },
      "PROPN": {
       "n": 1395,
       "acc": 63.23
      },
      "X": {
       "n": 463,
       "acc": 79.05
      },
      "ADV": {
       "n": 356,
       "acc": 31.46
      },
      "DET": {
       "n": 136,
       "acc": 47.79
      },
      "PRON": {
       "n": 76,
       "acc": 30.26
      },
      "NUM": {
       "n": 73,
       "acc": 28.77
      },
      "ADP": {
       "n": 71,
       "acc": 38.03
      }
     },
     "seen_form": {
      "n": 2140,
      "acc": 32.2
     },
     "unseen_form": {
      "n": 8367,
      "acc": 51.48
     },
     "gold_consistent": {
      "n": 171,
      "acc": 76.02
     },
     "gold_ambiguous": {
      "n": 145,
      "acc": 45.52
     },
     "gold_conflicting": {
      "n": 1824,
      "acc": 27.03
     }
    }
   }
  },
  "C4_nomacron": {
   "checkpoint": "/Users/jamartin/OE_lemmatiser_retrain/experiments/C4_nomacron/best_model.pt",
   "universe": "nomacron",
   "training_summary": {
    "name": "C4_nomacron",
    "aug": "synthetic",
    "aug_prob": 0.2,
    "extra_train": {},
    "universe": "nomacron",
    "corpus_rows": 84050,
    "total_train_rows": 84050,
    "epochs": 80,
    "batch_size": 72,
    "lr": 0.00015,
    "label_smoothing": 0.15,
    "pos_weight": 0.2,
    "dropout": 0.15,
    "cycles": 4,
    "patience": 20,
    "seed": 42,
    "pretrained_encoder": true,
    "steps_per_epoch": 1168,
    "total_steps": 93440,
    "parameters": 26437278,
    "device": "mps",
    "started": "2026-08-25T19:04:00",
    "best_dev_lemma": 0.7271083190557777,
    "best_epoch": 80,
    "epochs_run": 80,
    "wall_seconds": 19213.2,
    "finished": "2026-08-26T00:24:13"
   },
   "dev": {
    "n": 10506,
    "greedy": {
     "lemma": 72.71,
     "lemma_notation_insensitive": 73.34,
     "pos": 82.68
    },
    "beam10": {
     "lemma": 73.19,
     "lemma_notation_insensitive": 73.8,
     "pos": 82.68
    },
    "oracle_in_beam10": 94.2,
    "breakdown_greedy": {
     "by_pos": {
      "NOUN": {
       "n": 3314,
       "acc": 67.98
      },
      "VERB": {
       "n": 3128,
       "acc": 78.48
      },
      "ADJ": {
       "n": 1561,
       "acc": 69.31
      },
      "PROPN": {
       "n": 1369,
       "acc": 78.74
      },
      "X": {
       "n": 418,
       "acc": 86.36
      },
      "ADV": {
       "n": 331,
       "acc": 56.8
      },
      "DET": {
       "n": 122,
       "acc": 61.48
      },
      "NUM": {
       "n": 97,
       "acc": 61.86
      },
      "ADP": {
       "n": 78,
       "acc": 52.56
      },
      "PRON": {
       "n": 64,
       "acc": 57.81
      }
     },
     "seen_form": {
      "n": 2149,
      "acc": 47.93
     },
     "unseen_form": {
      "n": 8357,
      "acc": 79.08
     },
     "gold_consistent": {
      "n": 471,
      "acc": 97.66
     },
     "gold_ambiguous": {
      "n": 180,
      "acc": 69.44
     },
     "gold_conflicting": {
      "n": 1498,
      "acc": 29.71
     }
    }
   },
   "test": {
    "n": 10507,
    "greedy": {
     "lemma": 71.77,
     "lemma_notation_insensitive": 72.46,
     "pos": 82.9
    },
    "beam10": {
     "lemma": 72.22,
     "lemma_notation_insensitive": 72.91,
     "pos": 82.9
    },
    "oracle_in_beam10": 93.9,
    "breakdown_greedy": {
     "by_pos": {
      "NOUN": {
       "n": 3230,
       "acc": 66.97
      },
      "VERB": {
       "n": 3126,
       "acc": 77.19
      },
      "ADJ": {
       "n": 1559,
       "acc": 66.97
      },
      "PROPN": {
       "n": 1395,
       "acc": 78.71
      },
      "X": {
       "n": 463,
       "acc": 84.02
      },
      "ADV": {
       "n": 356,
       "acc": 54.78
      },
      "DET": {
       "n": 136,
       "acc": 64.71
      },
      "PRON": {
       "n": 76,
       "acc": 59.21
      },
      "NUM": {
       "n": 73,
       "acc": 75.34
      },
      "ADP": {
       "n": 71,
       "acc": 54.93
      }
     },
     "seen_form": {
      "n": 2140,
      "acc": 48.13
     },
     "unseen_form": {
      "n": 8367,
      "acc": 77.82
     },
     "gold_consistent": {
      "n": 502,
      "acc": 99.0
     },
     "gold_ambiguous": {
      "n": 196,
      "acc": 66.84
     },
     "gold_conflicting": {
      "n": 1442,
      "acc": 27.88
     }
    }
   }
  },
  "C5_normalised": {
   "checkpoint": "/Users/jamartin/OE_lemmatiser_retrain/experiments/C5_normalised/best_model.pt",
   "universe": "normalised",
   "training_summary": {
    "name": "C5_normalised",
    "aug": "synthetic",
    "aug_prob": 0.2,
    "extra_train": {},
    "universe": "normalised",
    "corpus_rows": 84050,
    "total_train_rows": 84050,
    "epochs": 80,
    "batch_size": 72,
    "lr": 0.00015,
    "label_smoothing": 0.15,
    "pos_weight": 0.2,
    "dropout": 0.15,
    "cycles": 4,
    "patience": 20,
    "seed": 42,
    "pretrained_encoder": true,
    "steps_per_epoch": 1168,
    "total_steps": 93440,
    "parameters": 26437278,
    "device": "mps",
    "started": "2026-08-26T00:25:12",
    "best_dev_lemma": 0.7424328954882924,
    "best_epoch": 80,
    "epochs_run": 80,
    "wall_seconds": 13185.7,
    "finished": "2026-08-26T04:04:57"
   },
   "dev": {
    "n": 10506,
    "greedy": {
     "lemma": 74.24,
     "lemma_notation_insensitive": 74.24,
     "pos": 82.72
    },
    "beam10": {
     "lemma": 74.96,
     "lemma_notation_insensitive": 74.96,
     "pos": 82.72
    },
    "oracle_in_beam10": 94.48,
    "breakdown_greedy": {
     "by_pos": {
      "NOUN": {
       "n": 3314,
       "acc": 69.61
      },
      "VERB": {
       "n": 3128,
       "acc": 79.41
      },
      "ADJ": {
       "n": 1561,
       "acc": 71.56
      },
      "PROPN": {
       "n": 1369,
       "acc": 81.67
      },
      "X": {
       "n": 418,
       "acc": 86.6
      },
      "ADV": {
       "n": 331,
       "acc": 56.5
      },
      "DET": {
       "n": 122,
       "acc": 58.2
      },
      "NUM": {
       "n": 97,
       "acc": 67.01
      },
      "ADP": {
       "n": 78,
       "acc": 50.0
      },
      "PRON": {
       "n": 64,
       "acc": 60.94
      }
     },
     "seen_form": {
      "n": 2149,
      "acc": 52.21
     },
     "unseen_form": {
      "n": 8357,
      "acc": 79.91
     },
     "gold_consistent": {
      "n": 675,
      "acc": 98.07
     },
     "gold_ambiguous": {
      "n": 190,
      "acc": 65.26
     },
     "gold_conflicting": {
      "n": 1284,
      "acc": 26.17
     }
    }
   },
   "test": {
    "n": 10507,
    "greedy": {
     "lemma": 73.86,
     "lemma_notation_insensitive": 73.86,
     "pos": 82.79
    },
    "beam10": {
     "lemma": 74.65,
     "lemma_notation_insensitive": 74.65,
     "pos": 82.79
    },
    "oracle_in_beam10": 94.43,
    "breakdown_greedy": {
     "by_pos": {
      "NOUN": {
       "n": 3230,
       "acc": 69.6
      },
      "VERB": {
       "n": 3126,
       "acc": 77.48
      },
      "ADJ": {
       "n": 1559,
       "acc": 70.69
      },
      "PROPN": {
       "n": 1395,
       "acc": 80.86
      },
      "X": {
       "n": 463,
       "acc": 85.96
      },
      "ADV": {
       "n": 356,
       "acc": 58.71
      },
      "DET": {
       "n": 136,
       "acc": 66.18
      },
      "PRON": {
       "n": 76,
       "acc": 60.53
      },
      "NUM": {
       "n": 73,
       "acc": 80.82
      },
      "ADP": {
       "n": 71,
       "acc": 61.97
      }
     },
     "seen_form": {
      "n": 2140,
      "acc": 53.97
     },
     "unseen_form": {
      "n": 8367,
      "acc": 78.94
     },
     "gold_consistent": {
      "n": 722,
      "acc": 99.17
     },
     "gold_ambiguous": {
      "n": 210,
      "acc": 66.19
     },
     "gold_conflicting": {
      "n": 1208,
      "acc": 24.83
     }
    }
   }
  },
  "C7_variant_normalised": {
   "checkpoint": "/Users/jamartin/OE_lemmatiser_retrain/experiments/C7_variant_normalised/best_model.pt",
   "universe": "normalised",
   "training_summary": {
    "name": "C7_variant_normalised",
    "aug": "attested",
    "aug_prob": 0.2,
    "extra_train": {},
    "universe": "normalised",
    "corpus_rows": 84050,
    "total_train_rows": 84050,
    "epochs": 80,
    "batch_size": 72,
    "lr": 0.00015,
    "label_smoothing": 0.15,
    "pos_weight": 0.2,
    "dropout": 0.15,
    "cycles": 4,
    "patience": 20,
    "seed": 42,
    "pretrained_encoder": true,
    "steps_per_epoch": 1168,
    "total_steps": 93440,
    "parameters": 26437278,
    "device": "mps",
    "started": "2026-08-26T04:05:14",
    "best_dev_lemma": 0.7459546925566343,
    "best_epoch": 78,
    "epochs_run": 80,
    "wall_seconds": 19508.4,
    "finished": "2026-08-26T09:30:22",
    "augmenter_stats": {
     "attested_substitutions": 821284,
     "synthetic_fallbacks": 522723
    }
   },
   "dev": {
    "n": 10506,
    "greedy": {
     "lemma": 74.6,
     "lemma_notation_insensitive": 74.6,
     "pos": 82.54
    },
    "beam10": {
     "lemma": 75.02,
     "lemma_notation_insensitive": 75.02,
     "pos": 82.54
    },
    "oracle_in_beam10": 94.58,
    "breakdown_greedy": {
     "by_pos": {
      "NOUN": {
       "n": 3314,
       "acc": 70.43
      },
      "VERB": {
       "n": 3128,
       "acc": 78.71
      },
      "ADJ": {
       "n": 1561,
       "acc": 71.56
      },
      "PROPN": {
       "n": 1369,
       "acc": 81.96
      },
      "X": {
       "n": 418,
       "acc": 88.04
      },
      "ADV": {
       "n": 331,
       "acc": 61.03
      },
      "DET": {
       "n": 122,
       "acc": 59.84
      },
      "NUM": {
       "n": 97,
       "acc": 69.07
      },
      "ADP": {
       "n": 78,
       "acc": 51.28
      },
      "PRON": {
       "n": 64,
       "acc": 65.62
      }
     },
     "seen_form": {
      "n": 2149,
      "acc": 52.63
     },
     "unseen_form": {
      "n": 8357,
      "acc": 80.24
     },
     "gold_consistent": {
      "n": 675,
      "acc": 98.52
     },
     "gold_ambiguous": {
      "n": 190,
      "acc": 66.84
     },
     "gold_conflicting": {
      "n": 1284,
      "acc": 26.4
     }
    }
   },
   "test": {
    "n": 10507,
    "greedy": {
     "lemma": 74.3,
     "lemma_notation_insensitive": 74.3,
     "pos": 82.77
    },
    "beam10": {
     "lemma": 74.73,
     "lemma_notation_insensitive": 74.73,
     "pos": 82.77
    },
    "oracle_in_beam10": 94.42,
    "breakdown_greedy": {
     "by_pos": {
      "NOUN": {
       "n": 3230,
       "acc": 70.0
      },
      "VERB": {
       "n": 3126,
       "acc": 78.53
      },
      "ADJ": {
       "n": 1559,
       "acc": 70.69
      },
      "PROPN": {
       "n": 1395,
       "acc": 81.94
      },
      "X": {
       "n": 463,
       "acc": 86.18
      },
      "ADV": {
       "n": 356,
       "acc": 57.58
      },
      "DET": {
       "n": 136,
       "acc": 64.71
      },
      "PRON": {
       "n": 76,
       "acc": 60.53
      },
      "NUM": {
       "n": 73,
       "acc": 75.34
      },
      "ADP": {
       "n": 71,
       "acc": 53.52
      }
     },
     "seen_form": {
      "n": 2140,
      "acc": 53.18
     },
     "unseen_form": {
      "n": 8367,
      "acc": 79.71
     },
     "gold_consistent": {
      "n": 722,
      "acc": 98.61
     },
     "gold_ambiguous": {
      "n": 210,
      "acc": 64.76
     },
     "gold_conflicting": {
      "n": 1208,
      "acc": 24.01
     }
    }
   }
  },
  "C6_combined_normalised": {
   "checkpoint": "/Users/jamartin/OE_lemmatiser_retrain/experiments/C6_combined_normalised/best_model.pt",
   "universe": "normalised",
   "training_summary": {
    "name": "C6_combined_normalised",
    "aug": "attested",
    "aug_prob": 0.2,
    "extra_train": {
     "lexicon_safe_normalised.tsv": 121783
    },
    "universe": "normalised",
    "corpus_rows": 84050,
    "total_train_rows": 205833,
    "epochs": 80,
    "batch_size": 72,
    "lr": 0.00015,
    "label_smoothing": 0.15,
    "pos_weight": 0.2,
    "dropout": 0.15,
    "cycles": 4,
    "patience": 20,
    "seed": 42,
    "pretrained_encoder": true,
    "steps_per_epoch": 2859,
    "total_steps": 228720,
    "parameters": 26437278,
    "device": "mps",
    "started": "2026-08-26T09:30:37",
    "best_dev_lemma": 0.7013135351227869,
    "best_epoch": 54,
    "epochs_run": 74,
    "wall_seconds": 21494.5,
    "finished": "2026-08-26T15:28:52",
    "augmenter_stats": {
     "attested_substitutions": 1704504,
     "synthetic_fallbacks": 1339887
    }
   },
   "dev": {
    "n": 10506,
    "greedy": {
     "lemma": 70.13,
     "lemma_notation_insensitive": 70.13,
     "pos": 81.48
    },
    "beam10": {
     "lemma": 70.61,
     "lemma_notation_insensitive": 70.61,
     "pos": 81.48
    },
    "oracle_in_beam10": 94.84,
    "breakdown_greedy": {
     "by_pos": {
      "NOUN": {
       "n": 3314,
       "acc": 70.64
      },
      "VERB": {
       "n": 3128,
       "acc": 68.77
      },
      "ADJ": {
       "n": 1561,
       "acc": 70.85
      },
      "PROPN": {
       "n": 1369,
       "acc": 75.68
      },
      "X": {
       "n": 418,
       "acc": 77.99
      },
      "ADV": {
       "n": 331,
       "acc": 57.4
      },
      "DET": {
       "n": 122,
       "acc": 54.1
      },
      "NUM": {
       "n": 97,
       "acc": 67.01
      },
      "ADP": {
       "n": 78,
       "acc": 50.0
      },
      "PRON": {
       "n": 64,
       "acc": 57.81
      }
     },
     "seen_form": {
      "n": 2149,
      "acc": 53.19
     },
     "unseen_form": {
      "n": 8357,
      "acc": 74.49
     },
     "gold_consistent": {
      "n": 675,
      "acc": 93.33
     },
     "gold_ambiguous": {
      "n": 190,
      "acc": 64.74
     },
     "gold_conflicting": {
      "n": 1284,
      "acc": 30.37
     }
    }
   },
   "test": {
    "n": 10507,
    "greedy": {
     "lemma": 69.59,
     "lemma_notation_insensitive": 69.59,
     "pos": 81.67
    },
    "beam10": {
     "lemma": 69.87,
     "lemma_notation_insensitive": 69.87,
     "pos": 81.67
    },
    "oracle_in_beam10": 94.83,
    "breakdown_greedy": {
     "by_pos": {
      "NOUN": {
       "n": 3230,
       "acc": 70.74
      },
      "VERB": {
       "n": 3126,
       "acc": 67.98
      },
      "ADJ": {
       "n": 1559,
       "acc": 68.44
      },
      "PROPN": {
       "n": 1395,
       "acc": 74.19
      },
      "X": {
       "n": 463,
       "acc": 81.43
      },
      "ADV": {
       "n": 356,
       "acc": 53.65
      },
      "DET": {
       "n": 136,
       "acc": 60.29
      },
      "PRON": {
       "n": 76,
       "acc": 57.89
      },
      "NUM": {
       "n": 73,
       "acc": 72.6
      },
      "ADP": {
       "n": 71,
       "acc": 56.34
      }
     },
     "seen_form": {
      "n": 2140,
      "acc": 52.99
     },
     "unseen_form": {
      "n": 8367,
      "acc": 73.84
     },
     "gold_consistent": {
      "n": 722,
      "acc": 93.21
     },
     "gold_ambiguous": {
      "n": 210,
      "acc": 62.38
     },
     "gold_conflicting": {
      "n": 1208,
      "acc": 27.32
     }
    }
   }
  },
  "C8_conflictfree_normalised": {
   "checkpoint": "/Users/jamartin/OE_lemmatiser_retrain/experiments/C8_conflictfree_normalised/best_model.pt",
   "universe": "normalised",
   "training_summary": {
    "name": "C8_conflictfree_normalised",
    "aug": "attested",
    "aug_prob": 0.2,
    "extra_train": {
     "lexicon_conflictfree_normalised.tsv": 97556
    },
    "universe": "normalised",
    "corpus_rows": 84050,
    "total_train_rows": 181606,
    "epochs": 80,
    "batch_size": 72,
    "lr": 0.00015,
    "label_smoothing": 0.15,
    "pos_weight": 0.2,
    "dropout": 0.15,
    "cycles": 4,
    "patience": 20,
    "seed": 42,
    "pretrained_encoder": true,
    "steps_per_epoch": 2523,
    "total_steps": 201840,
    "parameters": 26437278,
    "device": "mps",
    "started": "2026-08-26T15:29:54",
    "best_dev_lemma": 0.7472872644203312,
    "best_epoch": 76,
    "epochs_run": 80,
    "wall_seconds": 24465.5,
    "finished": "2026-08-26T22:17:40",
    "augmenter_stats": {
     "attested_substitutions": 1649398,
     "synthetic_fallbacks": 1255065
    }
   },
   "dev": {
    "n": 10506,
    "greedy": {
     "lemma": 74.73,
     "lemma_notation_insensitive": 74.73,
     "pos": 82.65
    },
    "beam10": {
     "lemma": 75.02,
     "lemma_notation_insensitive": 75.02,
     "pos": 82.65
    },
    "oracle_in_beam10": 94.99,
    "breakdown_greedy": {
     "by_pos": {
      "NOUN": {
       "n": 3314,
       "acc": 73.48
      },
      "VERB": {
       "n": 3128,
       "acc": 76.02
      },
      "ADJ": {
       "n": 1561,
       "acc": 73.93
      },
      "PROPN": {
       "n": 1369,
       "acc": 79.84
      },
      "X": {
       "n": 418,
       "acc": 83.49
      },
      "ADV": {
       "n": 331,
       "acc": 62.54
      },
      "DET": {
       "n": 122,
       "acc": 59.84
      },
      "NUM": {
       "n": 97,
       "acc": 77.32
      },
      "ADP": {
       "n": 78,
       "acc": 50.0
      },
      "PRON": {
       "n": 64,
       "acc": 57.81
      }
     },
     "seen_form": {
      "n": 2149,
      "acc": 53.89
     },
     "unseen_form": {
      "n": 8357,
      "acc": 80.09
     },
     "gold_consistent": {
      "n": 675,
      "acc": 97.63
     },
     "gold_ambiguous": {
      "n": 190,
      "acc": 68.42
     },
     "gold_conflicting": {
      "n": 1284,
      "acc": 28.74
     }
    }
   },
   "test": {
    "n": 10507,
    "greedy": {
     "lemma": 73.61,
     "lemma_notation_insensitive": 73.61,
     "pos": 82.82
    },
    "beam10": {
     "lemma": 73.96,
     "lemma_notation_insensitive": 73.96,
     "pos": 82.82
    },
    "oracle_in_beam10": 94.73,
    "breakdown_greedy": {
     "by_pos": {
      "NOUN": {
       "n": 3230,
       "acc": 73.44
      },
      "VERB": {
       "n": 3126,
       "acc": 74.15
      },
      "ADJ": {
       "n": 1559,
       "acc": 71.46
      },
      "PROPN": {
       "n": 1395,
       "acc": 78.06
      },
      "X": {
       "n": 463,
       "acc": 79.7
      },
      "ADV": {
       "n": 356,
       "acc": 61.52
      },
      "DET": {
       "n": 136,
       "acc": 65.44
      },
      "PRON": {
       "n": 76,
       "acc": 64.47
      },
      "NUM": {
       "n": 73,
       "acc": 82.19
      },
      "ADP": {
       "n": 71,
       "acc": 59.15
      }
     },
     "seen_form": {
      "n": 2140,
      "acc": 55.47
     },
     "unseen_form": {
      "n": 8367,
      "acc": 78.25
     },
     "gold_consistent": {
      "n": 722,
      "acc": 98.34
     },
     "gold_ambiguous": {
      "n": 210,
      "acc": 67.14
     },
     "gold_conflicting": {
      "n": 1208,
      "acc": 27.81
     }
    }
   }
  },
  "C9_unseenonly_normalised": {
   "checkpoint": "/Users/jamartin/OE_lemmatiser_retrain/experiments/C9_unseenonly_normalised/best_model.pt",
   "universe": "normalised",
   "training_summary": {
    "name": "C9_unseenonly_normalised",
    "aug": "attested",
    "aug_prob": 0.2,
    "extra_train": {
     "lexicon_unseenonly_normalised.tsv": 74456
    },
    "universe": "normalised",
    "corpus_rows": 84050,
    "total_train_rows": 158506,
    "epochs": 80,
    "batch_size": 72,
    "lr": 0.00015,
    "label_smoothing": 0.15,
    "pos_weight": 0.2,
    "dropout": 0.15,
    "cycles": 4,
    "patience": 20,
    "seed": 42,
    "pretrained_encoder": true,
    "steps_per_epoch": 2202,
    "total_steps": 176160,
    "parameters": 26437278,
    "device": "mps",
    "started": "2026-08-26T22:17:57",
    "best_dev_lemma": 0.7411955073291453,
    "best_epoch": 79,
    "epochs_run": 80,
    "wall_seconds": 20286.0,
    "finished": "2026-08-27T03:56:03",
    "augmenter_stats": {
     "attested_substitutions": 1397656,
     "synthetic_fallbacks": 1135016
    }
   },
   "dev": {
    "n": 10506,
    "greedy": {
     "lemma": 74.12,
     "lemma_notation_insensitive": 74.12,
     "pos": 83.36
    },
    "beam10": {
     "lemma": 74.45,
     "lemma_notation_insensitive": 74.45,
     "pos": 83.36
    },
    "oracle_in_beam10": 95.12,
    "breakdown_greedy": {
     "by_pos": {
      "NOUN": {
       "n": 3314,
       "acc": 72.72
      },
      "VERB": {
       "n": 3128,
       "acc": 75.96
      },
      "ADJ": {
       "n": 1561,
       "acc": 73.29
      },
      "PROPN": {
       "n": 1369,
       "acc": 78.67
      },
      "X": {
       "n": 418,
       "acc": 82.78
      },
      "ADV": {
       "n": 331,
       "acc": 59.21
      },
      "DET": {
       "n": 122,
       "acc": 62.3
      },
      "NUM": {
       "n": 97,
       "acc": 73.2
      },
      "ADP": {
       "n": 78,
       "acc": 53.85
      },
      "PRON": {
       "n": 64,
       "acc": 57.81
      }
     },
     "seen_form": {
      "n": 2149,
      "acc": 53.37
     },
     "unseen_form": {
      "n": 8357,
      "acc": 79.45
     },
     "gold_consistent": {
      "n": 675,
      "acc": 95.85
     },
     "gold_ambiguous": {
      "n": 190,
      "acc": 71.58
     },
     "gold_conflicting": {
      "n": 1284,
      "acc": 28.35
     }
    }
   },
   "test": {
    "n": 10507,
    "greedy": {
     "lemma": 73.46,
     "lemma_notation_insensitive": 73.46,
     "pos": 83.14
    },
    "beam10": {
     "lemma": 73.81,
     "lemma_notation_insensitive": 73.81,
     "pos": 83.14
    },
    "oracle_in_beam10": 95.1,
    "breakdown_greedy": {
     "by_pos": {
      "NOUN": {
       "n": 3230,
       "acc": 73.0
      },
      "VERB": {
       "n": 3126,
       "acc": 74.12
      },
      "ADJ": {
       "n": 1559,
       "acc": 72.55
      },
      "PROPN": {
       "n": 1395,
       "acc": 77.71
      },
      "X": {
       "n": 463,
       "acc": 81.21
      },
      "ADV": {
       "n": 356,
       "acc": 57.58
      },
      "DET": {
       "n": 136,
       "acc": 65.44
      },
      "PRON": {
       "n": 76,
       "acc": 63.16
      },
      "NUM": {
       "n": 73,
       "acc": 79.45
      },
      "ADP": {
       "n": 71,
       "acc": 54.93
      }
     },
     "seen_form": {
      "n": 2140,
      "acc": 54.67
     },
     "unseen_form": {
      "n": 8367,
      "acc": 78.26
     },
     "gold_consistent": {
      "n": 722,
      "acc": 97.37
     },
     "gold_ambiguous": {
      "n": 210,
      "acc": 64.76
     },
     "gold_conflicting": {
      "n": 1208,
      "acc": 27.4
     }
    }
   }
  },
  "C10_revised": {
   "checkpoint": "/Users/jamartin/OE_lemmatiser_retrain/experiments/C10_revised/best_model.pt",
   "universe": "revised",
   "training_summary": {
    "name": "C10_revised",
    "aug": "attested",
    "aug_prob": 0.2,
    "extra_train": {},
    "universe": "revised",
    "corpus_rows": 84050,
    "total_train_rows": 84050,
    "epochs": 80,
    "batch_size": 72,
    "lr": 0.00015,
    "label_smoothing": 0.15,
    "pos_weight": 0.2,
    "dropout": 0.15,
    "cycles": 4,
    "patience": 20,
    "seed": 42,
    "pretrained_encoder": true,
    "steps_per_epoch": 1168,
    "total_steps": 93440,
    "parameters": 26437278,
    "device": "mps",
    "started": "2026-08-27T14:36:44",
    "best_dev_lemma": 0.7718446601941747,
    "best_epoch": 78,
    "epochs_run": 80,
    "wall_seconds": 13511.2,
    "finished": "2026-08-27T18:21:55",
    "augmenter_stats": {
     "attested_substitutions": 856496,
     "synthetic_fallbacks": 488813
    }
   },
   "dev": {
    "n": 10506,
    "greedy": {
     "lemma": 77.18,
     "lemma_notation_insensitive": 77.18,
     "pos": 82.34
    },
    "beam10": {
     "lemma": 77.43,
     "lemma_notation_insensitive": 77.43,
     "pos": 82.34
    },
    "oracle_in_beam10": 95.44,
    "breakdown_greedy": {
     "by_pos": {
      "NOUN": {
       "n": 3314,
       "acc": 72.6
      },
      "VERB": {
       "n": 3128,
       "acc": 80.79
      },
      "ADJ": {
       "n": 1561,
       "acc": 76.55
      },
      "PROPN": {
       "n": 1369,
       "acc": 82.83
      },
      "X": {
       "n": 418,
       "acc": 86.36
      },
      "ADV": {
       "n": 331,
       "acc": 64.05
      },
      "DET": {
       "n": 122,
       "acc": 70.49
      },
      "NUM": {
       "n": 97,
       "acc": 77.32
      },
      "ADP": {
       "n": 78,
       "acc": 65.38
      },
      "PRON": {
       "n": 64,
       "acc": 73.44
      }
     },
     "seen_form": {
      "n": 2149,
      "acc": 63.15
     },
     "unseen_form": {
      "n": 8357,
      "acc": 80.79
     },
     "gold_consistent": {
      "n": 1017,
      "acc": 98.03
     },
     "gold_ambiguous": {
      "n": 218,
      "acc": 71.1
     },
     "gold_conflicting": {
      "n": 914,
      "acc": 22.43
     }
    }
   },
   "test": {
    "n": 10507,
    "greedy": {
     "lemma": 76.9,
     "lemma_notation_insensitive": 76.9,
     "pos": 82.96
    },
    "beam10": {
     "lemma": 77.17,
     "lemma_notation_insensitive": 77.17,
     "pos": 82.96
    },
    "oracle_in_beam10": 95.17,
    "breakdown_greedy": {
     "by_pos": {
      "NOUN": {
       "n": 3230,
       "acc": 72.79
      },
      "VERB": {
       "n": 3126,
       "acc": 80.29
      },
      "ADJ": {
       "n": 1559,
       "acc": 74.92
      },
      "PROPN": {
       "n": 1395,
       "acc": 82.58
      },
      "X": {
       "n": 463,
       "acc": 86.39
      },
      "ADV": {
       "n": 356,
       "acc": 62.36
      },
      "DET": {
       "n": 136,
       "acc": 72.06
      },
      "PRON": {
       "n": 76,
       "acc": 68.42
      },
      "NUM": {
       "n": 73,
       "acc": 80.82
      },
      "ADP": {
       "n": 71,
       "acc": 74.65
      }
     },
     "seen_form": {
      "n": 2140,
      "acc": 63.6
     },
     "unseen_form": {
      "n": 8367,
      "acc": 80.3
     },
     "gold_consistent": {
      "n": 1006,
      "acc": 97.71
     },
     "gold_ambiguous": {
      "n": 223,
      "acc": 71.3
     },
     "gold_conflicting": {
      "n": 911,
      "acc": 24.04
     }
    }
   }
  },
  "C11_revised_apparatus": {
   "checkpoint": "/Users/jamartin/OE_lemmatiser_retrain/experiments/C11_revised_apparatus/best_model.pt",
   "universe": "revised2",
   "training_summary": {
    "name": "C11_revised_apparatus",
    "aug": "attested",
    "aug_prob": 0.2,
    "extra_train": {},
    "universe": "revised2",
    "corpus_rows": 84050,
    "total_train_rows": 84050,
    "epochs": 80,
    "batch_size": 72,
    "lr": 0.00015,
    "label_smoothing": 0.15,
    "pos_weight": 0.2,
    "dropout": 0.15,
    "cycles": 4,
    "patience": 20,
    "seed": 42,
    "pretrained_encoder": true,
    "steps_per_epoch": 1168,
    "total_steps": 93440,
    "parameters": 26437278,
    "device": "mps",
    "started": "2026-08-27T20:14:07",
    "best_dev_lemma": 0.7724157624214735,
    "best_epoch": 77,
    "epochs_run": 80,
    "wall_seconds": 9385.9,
    "finished": "2026-08-27T22:50:33",
    "augmenter_stats": {
     "attested_substitutions": 851163,
     "synthetic_fallbacks": 494556
    }
   },
   "dev": {
    "n": 10506,
    "greedy": {
     "lemma": 77.24,
     "lemma_notation_insensitive": 77.24,
     "pos": 82.31
    },
    "beam10": {
     "lemma": 77.45,
     "lemma_notation_insensitive": 77.45,
     "pos": 82.31
    },
    "oracle_in_beam10": 95.36,
    "breakdown_greedy": {
     "by_pos": {
      "NOUN": {
       "n": 3314,
       "acc": 72.72
      },
      "VERB": {
       "n": 3128,
       "acc": 79.73
      },
      "ADJ": {
       "n": 1561,
       "acc": 77.0
      },
      "PROPN": {
       "n": 1369,
       "acc": 84.22
      },
      "X": {
       "n": 418,
       "acc": 88.76
      },
      "ADV": {
       "n": 331,
       "acc": 64.05
      },
      "DET": {
       "n": 122,
       "acc": 73.77
      },
      "NUM": {
       "n": 97,
       "acc": 70.1
      },
      "ADP": {
       "n": 78,
       "acc": 69.23
      },
      "PRON": {
       "n": 64,
       "acc": 73.44
      }
     },
     "seen_form": {
      "n": 2149,
      "acc": 62.59
     },
     "unseen_form": {
      "n": 8357,
      "acc": 81.01
     },
     "gold_consistent": {
      "n": 959,
      "acc": 98.44
     },
     "gold_ambiguous": {
      "n": 223,
      "acc": 68.61
     },
     "gold_conflicting": {
      "n": 967,
      "acc": 25.65
     }
    }
   },
   "test": {
    "n": 10507,
    "greedy": {
     "lemma": 76.63,
     "lemma_notation_insensitive": 76.63,
     "pos": 82.65
    },
    "beam10": {
     "lemma": 76.82,
     "lemma_notation_insensitive": 76.82,
     "pos": 82.65
    },
    "oracle_in_beam10": 95.2,
    "breakdown_greedy": {
     "by_pos": {
      "NOUN": {
       "n": 3230,
       "acc": 72.51
      },
      "VERB": {
       "n": 3126,
       "acc": 78.63
      },
      "ADJ": {
       "n": 1559,
       "acc": 76.14
      },
      "PROPN": {
       "n": 1395,
       "acc": 82.94
      },
      "X": {
       "n": 463,
       "acc": 86.18
      },
      "ADV": {
       "n": 356,
       "acc": 62.92
      },
      "DET": {
       "n": 136,
       "acc": 75.74
      },
      "PRON": {
       "n": 76,
       "acc": 71.05
      },
      "NUM": {
       "n": 73,
       "acc": 80.82
      },
      "ADP": {
       "n": 71,
       "acc": 74.65
      }
     },
     "seen_form": {
      "n": 2140,
      "acc": 62.62
     },
     "unseen_form": {
      "n": 8367,
      "acc": 80.21
     },
     "gold_consistent": {
      "n": 969,
      "acc": 97.42
     },
     "gold_ambiguous": {
      "n": 229,
      "acc": 68.56
     },
     "gold_conflicting": {
      "n": 942,
      "acc": 25.37
     }
    }
   }
  },
  "C12_prefix_rule": {
   "checkpoint": "/Users/jamartin/OE_lemmatiser_retrain/experiments/C12_prefix_rule/best_model.pt",
   "universe": "revised3",
   "training_summary": {
    "name": "C12_prefix_rule",
    "aug": "attested",
    "aug_prob": 0.2,
    "extra_train": {},
    "universe": "revised3",
    "corpus_rows": 84050,
    "total_train_rows": 84050,
    "epochs": 80,
    "batch_size": 72,
    "lr": 0.00015,
    "label_smoothing": 0.15,
    "pos_weight": 0.2,
    "dropout": 0.15,
    "cycles": 4,
    "patience": 20,
    "seed": 42,
    "pretrained_encoder": true,
    "steps_per_epoch": 1168,
    "total_steps": 93440,
    "parameters": 26437278,
    "device": "mps",
    "started": "2026-08-28T00:59:07",
    "best_dev_lemma": 0.7784123358081096,
    "best_epoch": 79,
    "epochs_run": 80,
    "wall_seconds": 8711.6,
    "finished": "2026-08-28T03:24:18",
    "augmenter_stats": {
     "attested_substitutions": 856904,
     "synthetic_fallbacks": 489351
    }
   },
   "dev": {
    "n": 10506,
    "greedy": {
     "lemma": 77.84,
     "lemma_notation_insensitive": 77.84,
     "pos": 82.38
    },
    "beam10": {
     "lemma": 77.89,
     "lemma_notation_insensitive": 77.89,
     "pos": 82.38
    },
    "oracle_in_beam10": 95.32,
    "breakdown_greedy": {
     "by_pos": {
      "NOUN": {
       "n": 3314,
       "acc": 72.69
      },
      "VERB": {
       "n": 3128,
       "acc": 82.32
      },
      "ADJ": {
       "n": 1561,
       "acc": 77.19
      },
      "PROPN": {
       "n": 1369,
       "acc": 82.54
      },
      "X": {
       "n": 418,
       "acc": 89.71
      },
      "ADV": {
       "n": 331,
       "acc": 64.05
      },
      "DET": {
       "n": 122,
       "acc": 69.67
      },
      "NUM": {
       "n": 97,
       "acc": 75.26
      },
      "ADP": {
       "n": 78,
       "acc": 67.95
      },
      "PRON": {
       "n": 64,
       "acc": 73.44
      }
     },
     "seen_form": {
      "n": 2149,
      "acc": 65.05
     },
     "unseen_form": {
      "n": 8357,
      "acc": 81.13
     },
     "gold_consistent": {
      "n": 1085,
      "acc": 98.53
     },
     "gold_ambiguous": {
      "n": 218,
      "acc": 66.97
     },
     "gold_conflicting": {
      "n": 846,
      "acc": 21.63
     }
    }
   },
   "test": {
    "n": 10507,
    "greedy": {
     "lemma": 77.4,
     "lemma_notation_insensitive": 77.4,
     "pos": 82.85
    },
    "beam10": {
     "lemma": 77.58,
     "lemma_notation_insensitive": 77.58,
     "pos": 82.85
    },
    "oracle_in_beam10": 95.17,
    "breakdown_greedy": {
     "by_pos": {
      "NOUN": {
       "n": 3230,
       "acc": 73.07
      },
      "VERB": {
       "n": 3126,
       "acc": 80.93
      },
      "ADJ": {
       "n": 1559,
       "acc": 75.43
      },
      "PROPN": {
       "n": 1395,
       "acc": 82.72
      },
      "X": {
       "n": 463,
       "acc": 87.26
      },
      "ADV": {
       "n": 356,
       "acc": 64.61
      },
      "DET": {
       "n": 136,
       "acc": 72.06
      },
      "PRON": {
       "n": 76,
       "acc": 68.42
      },
      "NUM": {
       "n": 73,
       "acc": 79.45
      },
      "ADP": {
       "n": 71,
       "acc": 77.46
      }
     },
     "seen_form": {
      "n": 2140,
      "acc": 64.77
     },
     "unseen_form": {
      "n": 8367,
      "acc": 80.63
     },
     "gold_consistent": {
      "n": 1081,
      "acc": 97.96
     },
     "gold_ambiguous": {
      "n": 223,
      "acc": 64.57
     },
     "gold_conflicting": {
      "n": 836,
      "acc": 21.89
     }
    }
   }
  },
  "C13_adjudicated": {
   "checkpoint": "/Users/jamartin/OE_lemmatiser_retrain/experiments/C13_adjudicated/best_model.pt",
   "universe": "revised4",
   "training_summary": {
    "name": "C13_adjudicated",
    "aug": "attested",
    "aug_prob": 0.2,
    "extra_train": {},
    "universe": "revised4",
    "corpus_rows": 84050,
    "total_train_rows": 84050,
    "epochs": 80,
    "batch_size": 72,
    "lr": 0.00015,
    "label_smoothing": 0.15,
    "pos_weight": 0.2,
    "dropout": 0.15,
    "cycles": 4,
    "patience": 20,
    "seed": 42,
    "pretrained_encoder": true,
    "steps_per_epoch": 1168,
    "total_steps": 93440,
    "parameters": 26437278,
    "device": "mps",
    "started": "2026-08-28T10:28:34",
    "best_dev_lemma": 0.7823148676946506,
    "best_epoch": 79,
    "epochs_run": 80,
    "wall_seconds": 8896.3,
    "finished": "2026-08-28T12:56:50",
    "augmenter_stats": {
     "attested_substitutions": 855765,
     "synthetic_fallbacks": 490533
    }
   },
   "dev": {
    "n": 10506,
    "greedy": {
     "lemma": 78.23,
     "lemma_notation_insensitive": 78.23,
     "pos": 82.42
    },
    "beam10": {
     "lemma": 78.41,
     "lemma_notation_insensitive": 78.41,
     "pos": 82.42
    },
    "oracle_in_beam10": 95.48,
    "breakdown_greedy": {
     "by_pos": {
      "NOUN": {
       "n": 3314,
       "acc": 73.3
      },
      "VERB": {
       "n": 3128,
       "acc": 83.12
      },
      "ADJ": {
       "n": 1561,
       "acc": 76.81
      },
      "PROPN": {
       "n": 1369,
       "acc": 84.08
      },
      "X": {
       "n": 418,
       "acc": 88.04
      },
      "ADV": {
       "n": 331,
       "acc": 62.84
      },
      "DET": {
       "n": 122,
       "acc": 66.39
      },
      "NUM": {
       "n": 97,
       "acc": 69.07
      },
      "ADP": {
       "n": 78,
       "acc": 69.23
      },
      "PRON": {
       "n": 64,
       "acc": 73.44
      }
     },
     "seen_form": {
      "n": 2149,
      "acc": 67.75
     },
     "unseen_form": {
      "n": 8357,
      "acc": 80.93
     },
     "gold_consistent": {
      "n": 1249,
      "acc": 95.6
     },
     "gold_ambiguous": {
      "n": 212,
      "acc": 63.21
     },
     "gold_conflicting": {
      "n": 688,
      "acc": 18.6
     }
    }
   },
   "test": {
    "n": 10507,
    "greedy": {
     "lemma": 78.21,
     "lemma_notation_insensitive": 78.21,
     "pos": 82.59
    },
    "beam10": {
     "lemma": 78.45,
     "lemma_notation_insensitive": 78.45,
     "pos": 82.59
    },
    "oracle_in_beam10": 95.45,
    "breakdown_greedy": {
     "by_pos": {
      "NOUN": {
       "n": 3230,
       "acc": 74.27
      },
      "VERB": {
       "n": 3126,
       "acc": 82.44
      },
      "ADJ": {
       "n": 1559,
       "acc": 76.07
      },
      "PROPN": {
       "n": 1395,
       "acc": 82.51
      },
      "X": {
       "n": 463,
       "acc": 85.96
      },
      "ADV": {
       "n": 356,
       "acc": 64.33
      },
      "DET": {
       "n": 136,
       "acc": 73.53
      },
      "PRON": {
       "n": 76,
       "acc": 68.42
      },
      "NUM": {
       "n": 73,
       "acc": 78.08
      },
      "ADP": {
       "n": 71,
       "acc": 77.46
      }
     },
     "seen_form": {
      "n": 2140,
      "acc": 69.16
     },
     "unseen_form": {
      "n": 8367,
      "acc": 80.53
     },
     "gold_consistent": {
      "n": 1272,
      "acc": 95.75
     },
     "gold_ambiguous": {
      "n": 215,
      "acc": 66.05
     },
     "gold_conflicting": {
      "n": 653,
      "acc": 18.38
     }
    }
   }
  },
  "C14_expert": {
   "checkpoint": "/Users/jamartin/OE_lemmatiser_retrain/experiments/C14_expert/best_model.pt",
   "universe": "revised6",
   "training_summary": {
    "name": "C14_expert",
    "aug": "attested",
    "aug_prob": 0.2,
    "extra_train": {},
    "universe": "revised6",
    "corpus_rows": 84050,
    "total_train_rows": 84050,
    "epochs": 80,
    "batch_size": 72,
    "lr": 0.00015,
    "label_smoothing": 0.15,
    "pos_weight": 0.2,
    "dropout": 0.15,
    "cycles": 4,
    "patience": 20,
    "seed": 42,
    "pretrained_encoder": true,
    "steps_per_epoch": 1168,
    "total_steps": 93440,
    "parameters": 26437278,
    "device": "mps",
    "started": "2026-08-28T13:05:46",
    "best_dev_lemma": 0.7850751951265943,
    "best_epoch": 78,
    "epochs_run": 80,
    "wall_seconds": 9036.9,
    "finished": "2026-08-28T15:36:23",
    "augmenter_stats": {
     "attested_substitutions": 855903,
     "synthetic_fallbacks": 489435
    }
   },
   "dev": {
    "n": 10506,
    "greedy": {
     "lemma": 78.51,
     "lemma_notation_insensitive": 78.51,
     "pos": 82.45
    },
    "beam10": {
     "lemma": 78.76,
     "lemma_notation_insensitive": 78.76,
     "pos": 82.45
    },
    "oracle_in_beam10": 95.63,
    "breakdown_greedy": {
     "by_pos": {
      "NOUN": {
       "n": 3314,
       "acc": 74.14
      },
      "VERB": {
       "n": 3128,
       "acc": 83.12
      },
      "ADJ": {
       "n": 1561,
       "acc": 76.87
      },
      "PROPN": {
       "n": 1369,
       "acc": 83.42
      },
      "X": {
       "n": 418,
       "acc": 87.32
      },
      "ADV": {
       "n": 331,
       "acc": 65.26
      },
      "DET": {
       "n": 122,
       "acc": 69.67
      },
      "NUM": {
       "n": 97,
       "acc": 71.13
      },
      "ADP": {
       "n": 78,
       "acc": 66.67
      },
      "PRON": {
       "n": 64,
       "acc": 75.0
      }
     },
     "seen_form": {
      "n": 2149,
      "acc": 69.01
     },
     "unseen_form": {
      "n": 8357,
      "acc": 80.95
     },
     "gold_consistent": {
      "n": 1272,
      "acc": 95.2
     },
     "gold_ambiguous": {
      "n": 211,
      "acc": 63.98
     },
     "gold_conflicting": {
      "n": 666,
      "acc": 20.57
     }
    }
   },
   "test": {
    "n": 10507,
    "greedy": {
     "lemma": 78.21,
     "lemma_notation_insensitive": 78.21,
     "pos": 82.78
    },
    "beam10": {
     "lemma": 78.3,
     "lemma_notation_insensitive": 78.3,
     "pos": 82.78
    },
    "oracle_in_beam10": 95.23,
    "breakdown_greedy": {
     "by_pos": {
      "NOUN": {
       "n": 3230,
       "acc": 74.4
      },
      "VERB": {
       "n": 3126,
       "acc": 82.34
      },
      "ADJ": {
       "n": 1559,
       "acc": 76.14
      },
      "PROPN": {
       "n": 1395,
       "acc": 81.94
      },
      "X": {
       "n": 463,
       "acc": 86.83
      },
      "ADV": {
       "n": 356,
       "acc": 62.64
      },
      "DET": {
       "n": 136,
       "acc": 75.0
      },
      "PRON": {
       "n": 76,
       "acc": 68.42
      },
      "NUM": {
       "n": 73,
       "acc": 84.93
      },
      "ADP": {
       "n": 71,
       "acc": 74.65
      }
     },
     "seen_form": {
      "n": 2140,
      "acc": 69.11
     },
     "unseen_form": {
      "n": 8367,
      "acc": 80.53
     },
     "gold_consistent": {
      "n": 1280,
      "acc": 94.77
     },
     "gold_ambiguous": {
      "n": 212,
      "acc": 67.45
     },
     "gold_conflicting": {
      "n": 648,
      "acc": 18.98
     }
    }
   }
  },
  "C15_canonical": {
   "checkpoint": "/Users/jamartin/OE_lemmatiser_retrain/experiments/C15_canonical/best_model.pt",
   "universe": "revised7",
   "training_summary": {
    "name": "C15_canonical",
    "aug": "attested",
    "aug_prob": 0.2,
    "extra_train": {},
    "universe": "revised7",
    "corpus_rows": 84050,
    "total_train_rows": 84050,
    "epochs": 80,
    "batch_size": 72,
    "lr": 0.00015,
    "label_smoothing": 0.15,
    "pos_weight": 0.2,
    "dropout": 0.15,
    "cycles": 4,
    "patience": 20,
    "seed": 42,
    "pretrained_encoder": true,
    "steps_per_epoch": 1168,
    "total_steps": 93440,
    "parameters": 26437278,
    "device": "mps",
    "started": "2026-08-28T16:02:58",
    "best_dev_lemma": 0.7861222158766419,
    "best_epoch": 80,
    "epochs_run": 80,
    "wall_seconds": 9139.5,
    "finished": "2026-08-28T18:35:18",
    "augmenter_stats": {
     "attested_substitutions": 855678,
     "synthetic_fallbacks": 490556
    }
   },
   "dev": {
    "n": 10506,
    "greedy": {
     "lemma": 78.61,
     "lemma_notation_insensitive": 78.61,
     "pos": 82.39
    },
    "beam10": {
     "lemma": 78.81,
     "lemma_notation_insensitive": 78.81,
     "pos": 82.39
    },
    "oracle_in_beam10": 95.49,
    "breakdown_greedy": {
     "by_pos": {
      "NOUN": {
       "n": 3314,
       "acc": 74.56
      },
      "VERB": {
       "n": 3128,
       "acc": 82.7
      },
      "ADJ": {
       "n": 1561,
       "acc": 76.94
      },
      "PROPN": {
       "n": 1369,
       "acc": 84.15
      },
      "X": {
       "n": 418,
       "acc": 86.84
      },
      "ADV": {
       "n": 331,
       "acc": 64.35
      },
      "DET": {
       "n": 122,
       "acc": 68.85
      },
      "NUM": {
       "n": 97,
       "acc": 71.13
      },
      "ADP": {
       "n": 78,
       "acc": 71.79
      },
      "PRON": {
       "n": 64,
       "acc": 78.12
      }
     },
     "seen_form": {
      "n": 2149,
      "acc": 69.24
     },
     "unseen_form": {
      "n": 8357,
      "acc": 81.02
     },
     "gold_consistent": {
      "n": 1308,
      "acc": 95.18
     },
     "gold_ambiguous": {
      "n": 211,
      "acc": 63.51
     },
     "gold_conflicting": {
      "n": 630,
      "acc": 17.3
     }
    }
   },
   "test": {
    "n": 10507,
    "greedy": {
     "lemma": 78.32,
     "lemma_notation_insensitive": 78.32,
     "pos": 82.79
    },
    "beam10": {
     "lemma": 78.51,
     "lemma_notation_insensitive": 78.51,
     "pos": 82.79
    },
    "oracle_in_beam10": 95.37,
    "breakdown_greedy": {
     "by_pos": {
      "NOUN": {
       "n": 3230,
       "acc": 74.64
      },
      "VERB": {
       "n": 3126,
       "acc": 82.34
      },
      "ADJ": {
       "n": 1559,
       "acc": 75.88
      },
      "PROPN": {
       "n": 1395,
       "acc": 82.65
      },
      "X": {
       "n": 463,
       "acc": 85.53
      },
      "ADV": {
       "n": 356,
       "acc": 64.33
      },
      "DET": {
       "n": 136,
       "acc": 76.47
      },
      "PRON": {
       "n": 76,
       "acc": 68.42
      },
      "NUM": {
       "n": 73,
       "acc": 79.45
      },
      "ADP": {
       "n": 71,
       "acc": 76.06
      }
     },
     "seen_form": {
      "n": 2140,
      "acc": 69.63
     },
     "unseen_form": {
      "n": 8367,
      "acc": 80.54
     },
     "gold_consistent": {
      "n": 1312,
      "acc": 95.05
     },
     "gold_ambiguous": {
      "n": 215,
      "acc": 65.58
     },
     "gold_conflicting": {
      "n": 613,
      "acc": 16.64
     }
    }
   }
  }
 },
 "postprocessing": {
  "C0_replication": {
   "checkpoint": "/Users/jamartin/OE_lemmatiser_retrain/experiments/C0_replication/best_model.pt",
   "label": "C0_replication",
   "beam_size": 10,
   "universe": "original",
   "splits": {
    "dev": {
     "n": 10506,
     "oracle_in_beam": 93.19,
     "strategies": {
      "beam_top1": {
       "lemma_accuracy": 69.12,
       "lemma_accuracy_nomacron": 73.58,
       "pos_accuracy": 82.38
      },
      "filter_train": {
       "lemma_accuracy": 64.14,
       "lemma_accuracy_nomacron": 68.94,
       "pos_accuracy": 82.38
      },
      "filter_dict": {
       "lemma_accuracy": 67.19,
       "lemma_accuracy_nomacron": 73.94,
       "pos_accuracy": 82.38
      },
      "soft_dict_0.5": {
       "lemma_accuracy": 69.67,
       "lemma_accuracy_nomacron": 74.15,
       "pos_accuracy": 82.38
      },
      "soft_train_0.5": {
       "lemma_accuracy": 69.32,
       "lemma_accuracy_nomacron": 73.83,
       "pos_accuracy": 82.38
      },
      "soft_dict_1.0": {
       "lemma_accuracy": 69.93,
       "lemma_accuracy_nomacron": 74.48,
       "pos_accuracy": 82.38
      },
      "soft_train_1.0": {
       "lemma_accuracy": 69.4,
       "lemma_accuracy_nomacron": 73.97,
       "pos_accuracy": 82.38
      },
      "soft_dict_2.0": {
       "lemma_accuracy": 70.21,
       "lemma_accuracy_nomacron": 75.06,
       "pos_accuracy": 82.38
      },
      "soft_train_2.0": {
       "lemma_accuracy": 69.3,
       "lemma_accuracy_nomacron": 73.99,
       "pos_accuracy": 82.38
      },
      "soft_dict_4.0": {
       "lemma_accuracy": 69.3,
       "lemma_accuracy_nomacron": 75.35,
       "pos_accuracy": 82.38
      },
      "soft_train_4.0": {
       "lemma_accuracy": 68.6,
       "lemma_accuracy_nomacron": 73.42,
       "pos_accuracy": 82.38
      },
      "macron_dict": {
       "lemma_accuracy": 54.98,
       "lemma_accuracy_nomacron": 73.58,
       "pos_accuracy": 82.38
      },
      "macron_corpus": {
       "lemma_accuracy": 66.18,
       "lemma_accuracy_nomacron": 73.58,
       "pos_accuracy": 82.38
      },
      "soft_dict_1.0+macron_corpus": {
       "lemma_accuracy": 67.13,
       "lemma_accuracy_nomacron": 74.48,
       "pos_accuracy": 82.38
      }
     }
    },
    "test": {
     "n": 10507,
     "oracle_in_beam": 93.39,
     "strategies": {
      "beam_top1": {
       "lemma_accuracy": 68.08,
       "lemma_accuracy_nomacron": 72.88,
       "pos_accuracy": 82.86
      },
      "filter_train": {
       "lemma_accuracy": 63.46,
       "lemma_accuracy_nomacron": 68.51,
       "pos_accuracy": 82.86
      },
      "filter_dict": {
       "lemma_accuracy": 66.55,
       "lemma_accuracy_nomacron": 73.53,
       "pos_accuracy": 82.86
      },
      "soft_dict_0.5": {
       "lemma_accuracy": 68.73,
       "lemma_accuracy_nomacron": 73.53,
       "pos_accuracy": 82.86
      },
      "soft_train_0.5": {
       "lemma_accuracy": 68.44,
       "lemma_accuracy_nomacron": 73.25,
       "pos_accuracy": 82.86
      },
      "soft_dict_1.0": {
       "lemma_accuracy": 69.0,
       "lemma_accuracy_nomacron": 73.87,
       "pos_accuracy": 82.86
      },
      "soft_train_1.0": {
       "lemma_accuracy": 68.51,
       "lemma_accuracy_nomacron": 73.38,
       "pos_accuracy": 82.86
      },
      "soft_dict_2.0": {
       "lemma_accuracy": 69.32,
       "lemma_accuracy_nomacron": 74.51,
       "pos_accuracy": 82.86
      },
      "soft_train_2.0": {
       "lemma_accuracy": 68.53,
       "lemma_accuracy_nomacron": 73.55,
       "pos_accuracy": 82.86
      },
      "soft_dict_4.0": {
       "lemma_accuracy": 68.54,
       "lemma_accuracy_nomacron": 74.67,
       "pos_accuracy": 82.86
      },
      "soft_train_4.0": {
       "lemma_accuracy": 67.77,
       "lemma_accuracy_nomacron": 72.9,
       "pos_accuracy": 82.86
      },
      "macron_dict": {
       "lemma_accuracy": 55.07,
       "lemma_accuracy_nomacron": 72.88,
       "pos_accuracy": 82.86
      },
      "macron_corpus": {
       "lemma_accuracy": 65.05,
       "lemma_accuracy_nomacron": 72.88,
       "pos_accuracy": 82.86
      },
      "soft_dict_1.0+macron_corpus": {
       "lemma_accuracy": 66.14,
       "lemma_accuracy_nomacron": 73.87,
       "pos_accuracy": 82.86
      }
     }
    }
   }
  },
  "C10_revised": {
   "checkpoint": "../experiments/C10_revised/best_model.pt",
   "label": "C10_revised",
   "beam_size": 10,
   "universe": "revised",
   "splits": {
    "dev": {
     "n": 10506,
     "oracle_in_beam": 95.44,
     "strategies": {
      "beam_top1": {
       "lemma_accuracy": 77.43,
       "lemma_accuracy_nomacron": 77.43,
       "pos_accuracy": 82.34
      },
      "filter_train": {
       "lemma_accuracy": 73.38,
       "lemma_accuracy_nomacron": 73.38,
       "pos_accuracy": 82.34
      },
      "filter_dict": {
       "lemma_accuracy": 80.35,
       "lemma_accuracy_nomacron": 80.35,
       "pos_accuracy": 82.34
      },
      "soft_dict_0.5": {
       "lemma_accuracy": 78.03,
       "lemma_accuracy_nomacron": 78.03,
       "pos_accuracy": 82.34
      },
      "soft_train_0.5": {
       "lemma_accuracy": 77.79,
       "lemma_accuracy_nomacron": 77.79,
       "pos_accuracy": 82.34
      },
      "soft_dict_1.0": {
       "lemma_accuracy": 78.57,
       "lemma_accuracy_nomacron": 78.57,
       "pos_accuracy": 82.34
      },
      "soft_train_1.0": {
       "lemma_accuracy": 78.0,
       "lemma_accuracy_nomacron": 78.0,
       "pos_accuracy": 82.34
      },
      "soft_dict_2.0": {
       "lemma_accuracy": 79.37,
       "lemma_accuracy_nomacron": 79.37,
       "pos_accuracy": 82.34
      },
      "soft_train_2.0": {
       "lemma_accuracy": 78.25,
       "lemma_accuracy_nomacron": 78.25,
       "pos_accuracy": 82.34
      },
      "soft_dict_4.0": {
       "lemma_accuracy": 80.81,
       "lemma_accuracy_nomacron": 80.81,
       "pos_accuracy": 82.34
      },
      "soft_train_4.0": {
       "lemma_accuracy": 78.01,
       "lemma_accuracy_nomacron": 78.01,
       "pos_accuracy": 82.34
      },
      "soft_dict_8.0": {
       "lemma_accuracy": 80.35,
       "lemma_accuracy_nomacron": 80.35,
       "pos_accuracy": 82.34
      },
      "soft_train_8.0": {
       "lemma_accuracy": 73.38,
       "lemma_accuracy_nomacron": 73.38,
       "pos_accuracy": 82.34
      },
      "soft_dict_16.0": {
       "lemma_accuracy": 80.35,
       "lemma_accuracy_nomacron": 80.35,
       "pos_accuracy": 82.34
      },
      "soft_train_16.0": {
       "lemma_accuracy": 73.38,
       "lemma_accuracy_nomacron": 73.38,
       "pos_accuracy": 82.34
      }
     }
    },
    "test": {
     "n": 10507,
     "oracle_in_beam": 95.17,
     "strategies": {
      "beam_top1": {
       "lemma_accuracy": 77.17,
       "lemma_accuracy_nomacron": 77.17,
       "pos_accuracy": 82.96
      },
      "filter_train": {
       "lemma_accuracy": 72.95,
       "lemma_accuracy_nomacron": 72.95,
       "pos_accuracy": 82.96
      },
      "filter_dict": {
       "lemma_accuracy": 80.11,
       "lemma_accuracy_nomacron": 80.11,
       "pos_accuracy": 82.96
      },
      "soft_dict_0.5": {
       "lemma_accuracy": 77.84,
       "lemma_accuracy_nomacron": 77.84,
       "pos_accuracy": 82.96
      },
      "soft_train_0.5": {
       "lemma_accuracy": 77.42,
       "lemma_accuracy_nomacron": 77.42,
       "pos_accuracy": 82.96
      },
      "soft_dict_1.0": {
       "lemma_accuracy": 78.31,
       "lemma_accuracy_nomacron": 78.31,
       "pos_accuracy": 82.96
      },
      "soft_train_1.0": {
       "lemma_accuracy": 77.71,
       "lemma_accuracy_nomacron": 77.71,
       "pos_accuracy": 82.96
      },
      "soft_dict_2.0": {
       "lemma_accuracy": 79.25,
       "lemma_accuracy_nomacron": 79.25,
       "pos_accuracy": 82.96
      },
      "soft_train_2.0": {
       "lemma_accuracy": 78.09,
       "lemma_accuracy_nomacron": 78.09,
       "pos_accuracy": 82.96
      },
      "soft_dict_4.0": {
       "lemma_accuracy": 80.31,
       "lemma_accuracy_nomacron": 80.31,
       "pos_accuracy": 82.96
      },
      "soft_train_4.0": {
       "lemma_accuracy": 77.64,
       "lemma_accuracy_nomacron": 77.64,
       "pos_accuracy": 82.96
      },
      "soft_dict_8.0": {
       "lemma_accuracy": 80.11,
       "lemma_accuracy_nomacron": 80.11,
       "pos_accuracy": 82.96
      },
      "soft_train_8.0": {
       "lemma_accuracy": 72.95,
       "lemma_accuracy_nomacron": 72.95,
       "pos_accuracy": 82.96
      },
      "soft_dict_16.0": {
       "lemma_accuracy": 80.11,
       "lemma_accuracy_nomacron": 80.11,
       "pos_accuracy": 82.96
      },
      "soft_train_16.0": {
       "lemma_accuracy": 72.95,
       "lemma_accuracy_nomacron": 72.95,
       "pos_accuracy": 82.96
      }
     }
    }
   }
  },
  "C11_revised_apparatus": {
   "checkpoint": "../experiments/C11_revised_apparatus/best_model.pt",
   "label": "C11_revised_apparatus",
   "beam_size": 10,
   "universe": "revised2",
   "splits": {
    "dev": {
     "n": 10506,
     "oracle_in_beam": 95.36,
     "strategies": {
      "beam_top1": {
       "lemma_accuracy": 77.45,
       "lemma_accuracy_nomacron": 77.45,
       "pos_accuracy": 82.31
      },
      "filter_train": {
       "lemma_accuracy": 72.92,
       "lemma_accuracy_nomacron": 72.92,
       "pos_accuracy": 82.31
      },
      "filter_dict": {
       "lemma_accuracy": 80.02,
       "lemma_accuracy_nomacron": 80.02,
       "pos_accuracy": 82.31
      },
      "soft_dict_0.5": {
       "lemma_accuracy": 77.99,
       "lemma_accuracy_nomacron": 77.99,
       "pos_accuracy": 82.31
      },
      "soft_train_0.5": {
       "lemma_accuracy": 77.6,
       "lemma_accuracy_nomacron": 77.6,
       "pos_accuracy": 82.31
      },
      "soft_dict_1.0": {
       "lemma_accuracy": 78.42,
       "lemma_accuracy_nomacron": 78.42,
       "pos_accuracy": 82.31
      },
      "soft_train_1.0": {
       "lemma_accuracy": 77.74,
       "lemma_accuracy_nomacron": 77.74,
       "pos_accuracy": 82.31
      },
      "soft_dict_2.0": {
       "lemma_accuracy": 79.35,
       "lemma_accuracy_nomacron": 79.35,
       "pos_accuracy": 82.31
      },
      "soft_train_2.0": {
       "lemma_accuracy": 77.86,
       "lemma_accuracy_nomacron": 77.86,
       "pos_accuracy": 82.31
      },
      "soft_dict_4.0": {
       "lemma_accuracy": 80.51,
       "lemma_accuracy_nomacron": 80.51,
       "pos_accuracy": 82.31
      },
      "soft_train_4.0": {
       "lemma_accuracy": 77.48,
       "lemma_accuracy_nomacron": 77.48,
       "pos_accuracy": 82.31
      },
      "soft_dict_8.0": {
       "lemma_accuracy": 80.02,
       "lemma_accuracy_nomacron": 80.02,
       "pos_accuracy": 82.31
      },
      "soft_train_8.0": {
       "lemma_accuracy": 72.92,
       "lemma_accuracy_nomacron": 72.92,
       "pos_accuracy": 82.31
      },
      "soft_dict_16.0": {
       "lemma_accuracy": 80.02,
       "lemma_accuracy_nomacron": 80.02,
       "pos_accuracy": 82.31
      },
      "soft_train_16.0": {
       "lemma_accuracy": 72.92,
       "lemma_accuracy_nomacron": 72.92,
       "pos_accuracy": 82.31
      }
     }
    },
    "test": {
     "n": 10507,
     "oracle_in_beam": 95.2,
     "strategies": {
      "beam_top1": {
       "lemma_accuracy": 76.82,
       "lemma_accuracy_nomacron": 76.82,
       "pos_accuracy": 82.65
      },
      "filter_train": {
       "lemma_accuracy": 72.45,
       "lemma_accuracy_nomacron": 72.45,
       "pos_accuracy": 82.65
      },
      "filter_dict": {
       "lemma_accuracy": 79.58,
       "lemma_accuracy_nomacron": 79.58,
       "pos_accuracy": 82.65
      },
      "soft_dict_0.5": {
       "lemma_accuracy": 77.52,
       "lemma_accuracy_nomacron": 77.52,
       "pos_accuracy": 82.65
      },
      "soft_train_0.5": {
       "lemma_accuracy": 77.15,
       "lemma_accuracy_nomacron": 77.15,
       "pos_accuracy": 82.65
      },
      "soft_dict_1.0": {
       "lemma_accuracy": 77.9,
       "lemma_accuracy_nomacron": 77.9,
       "pos_accuracy": 82.65
      },
      "soft_train_1.0": {
       "lemma_accuracy": 77.2,
       "lemma_accuracy_nomacron": 77.2,
       "pos_accuracy": 82.65
      },
      "soft_dict_2.0": {
       "lemma_accuracy": 78.75,
       "lemma_accuracy_nomacron": 78.75,
       "pos_accuracy": 82.65
      },
      "soft_train_2.0": {
       "lemma_accuracy": 77.41,
       "lemma_accuracy_nomacron": 77.41,
       "pos_accuracy": 82.65
      },
      "soft_dict_4.0": {
       "lemma_accuracy": 80.05,
       "lemma_accuracy_nomacron": 80.05,
       "pos_accuracy": 82.65
      },
      "soft_train_4.0": {
       "lemma_accuracy": 77.25,
       "lemma_accuracy_nomacron": 77.25,
       "pos_accuracy": 82.65
      },
      "soft_dict_8.0": {
       "lemma_accuracy": 79.58,
       "lemma_accuracy_nomacron": 79.58,
       "pos_accuracy": 82.65
      },
      "soft_train_8.0": {
       "lemma_accuracy": 72.45,
       "lemma_accuracy_nomacron": 72.45,
       "pos_accuracy": 82.65
      },
      "soft_dict_16.0": {
       "lemma_accuracy": 79.58,
       "lemma_accuracy_nomacron": 79.58,
       "pos_accuracy": 82.65
      },
      "soft_train_16.0": {
       "lemma_accuracy": 72.45,
       "lemma_accuracy_nomacron": 72.45,
       "pos_accuracy": 82.65
      }
     }
    }
   }
  },
  "C12_prefix_rule": {
   "checkpoint": "../experiments/C12_prefix_rule/best_model.pt",
   "label": "C12_prefix_rule",
   "beam_size": 10,
   "universe": "revised3",
   "splits": {
    "dev": {
     "n": 10506,
     "oracle_in_beam": 95.32,
     "strategies": {
      "beam_top1": {
       "lemma_accuracy": 77.89,
       "lemma_accuracy_nomacron": 77.89,
       "pos_accuracy": 82.38
      },
      "filter_train": {
       "lemma_accuracy": 73.52,
       "lemma_accuracy_nomacron": 73.52,
       "pos_accuracy": 82.38
      },
      "filter_dict": {
       "lemma_accuracy": 80.62,
       "lemma_accuracy_nomacron": 80.62,
       "pos_accuracy": 82.38
      },
      "soft_dict_0.5": {
       "lemma_accuracy": 78.45,
       "lemma_accuracy_nomacron": 78.45,
       "pos_accuracy": 82.38
      },
      "soft_train_0.5": {
       "lemma_accuracy": 78.17,
       "lemma_accuracy_nomacron": 78.17,
       "pos_accuracy": 82.38
      },
      "soft_dict_1.0": {
       "lemma_accuracy": 79.1,
       "lemma_accuracy_nomacron": 79.1,
       "pos_accuracy": 82.38
      },
      "soft_train_1.0": {
       "lemma_accuracy": 78.36,
       "lemma_accuracy_nomacron": 78.36,
       "pos_accuracy": 82.38
      },
      "soft_dict_2.0": {
       "lemma_accuracy": 80.12,
       "lemma_accuracy_nomacron": 80.12,
       "pos_accuracy": 82.38
      },
      "soft_train_2.0": {
       "lemma_accuracy": 78.54,
       "lemma_accuracy_nomacron": 78.54,
       "pos_accuracy": 82.38
      },
      "soft_dict_4.0": {
       "lemma_accuracy": 81.09,
       "lemma_accuracy_nomacron": 81.09,
       "pos_accuracy": 82.38
      },
      "soft_train_4.0": {
       "lemma_accuracy": 78.04,
       "lemma_accuracy_nomacron": 78.04,
       "pos_accuracy": 82.38
      },
      "soft_dict_8.0": {
       "lemma_accuracy": 80.62,
       "lemma_accuracy_nomacron": 80.62,
       "pos_accuracy": 82.38
      },
      "soft_train_8.0": {
       "lemma_accuracy": 73.52,
       "lemma_accuracy_nomacron": 73.52,
       "pos_accuracy": 82.38
      },
      "soft_dict_16.0": {
       "lemma_accuracy": 80.62,
       "lemma_accuracy_nomacron": 80.62,
       "pos_accuracy": 82.38
      },
      "soft_train_16.0": {
       "lemma_accuracy": 73.52,
       "lemma_accuracy_nomacron": 73.52,
       "pos_accuracy": 82.38
      }
     }
    },
    "test": {
     "n": 10507,
     "oracle_in_beam": 95.17,
     "strategies": {
      "beam_top1": {
       "lemma_accuracy": 77.58,
       "lemma_accuracy_nomacron": 77.58,
       "pos_accuracy": 82.85
      },
      "filter_train": {
       "lemma_accuracy": 73.17,
       "lemma_accuracy_nomacron": 73.17,
       "pos_accuracy": 82.85
      },
      "filter_dict": {
       "lemma_accuracy": 80.48,
       "lemma_accuracy_nomacron": 80.48,
       "pos_accuracy": 82.85
      },
      "soft_dict_0.5": {
       "lemma_accuracy": 78.13,
       "lemma_accuracy_nomacron": 78.13,
       "pos_accuracy": 82.85
      },
      "soft_train_0.5": {
       "lemma_accuracy": 77.76,
       "lemma_accuracy_nomacron": 77.76,
       "pos_accuracy": 82.85
      },
      "soft_dict_1.0": {
       "lemma_accuracy": 78.56,
       "lemma_accuracy_nomacron": 78.56,
       "pos_accuracy": 82.85
      },
      "soft_train_1.0": {
       "lemma_accuracy": 77.96,
       "lemma_accuracy_nomacron": 77.96,
       "pos_accuracy": 82.85
      },
      "soft_dict_2.0": {
       "lemma_accuracy": 79.39,
       "lemma_accuracy_nomacron": 79.39,
       "pos_accuracy": 82.85
      },
      "soft_train_2.0": {
       "lemma_accuracy": 78.08,
       "lemma_accuracy_nomacron": 78.08,
       "pos_accuracy": 82.85
      },
      "soft_dict_4.0": {
       "lemma_accuracy": 80.49,
       "lemma_accuracy_nomacron": 80.49,
       "pos_accuracy": 82.85
      },
      "soft_train_4.0": {
       "lemma_accuracy": 77.81,
       "lemma_accuracy_nomacron": 77.81,
       "pos_accuracy": 82.85
      },
      "soft_dict_8.0": {
       "lemma_accuracy": 80.48,
       "lemma_accuracy_nomacron": 80.48,
       "pos_accuracy": 82.85
      },
      "soft_train_8.0": {
       "lemma_accuracy": 73.17,
       "lemma_accuracy_nomacron": 73.17,
       "pos_accuracy": 82.85
      },
      "soft_dict_16.0": {
       "lemma_accuracy": 80.48,
       "lemma_accuracy_nomacron": 80.48,
       "pos_accuracy": 82.85
      },
      "soft_train_16.0": {
       "lemma_accuracy": 73.17,
       "lemma_accuracy_nomacron": 73.17,
       "pos_accuracy": 82.85
      }
     }
    }
   }
  },
  "C13_adjudicated": {
   "checkpoint": "../experiments/C13_adjudicated/best_model.pt",
   "label": "C13_adjudicated",
   "beam_size": 10,
   "universe": "revised4",
   "splits": {
    "dev": {
     "n": 10506,
     "oracle_in_beam": 95.48,
     "strategies": {
      "beam_top1": {
       "lemma_accuracy": 78.41,
       "lemma_accuracy_nomacron": 78.41,
       "pos_accuracy": 82.42
      },
      "filter_train": {
       "lemma_accuracy": 74.01,
       "lemma_accuracy_nomacron": 74.01,
       "pos_accuracy": 82.42
      },
      "filter_dict": {
       "lemma_accuracy": 81.23,
       "lemma_accuracy_nomacron": 81.23,
       "pos_accuracy": 82.42
      },
      "soft_dict_0.5": {
       "lemma_accuracy": 78.98,
       "lemma_accuracy_nomacron": 78.98,
       "pos_accuracy": 82.42
      },
      "soft_train_0.5": {
       "lemma_accuracy": 78.69,
       "lemma_accuracy_nomacron": 78.69,
       "pos_accuracy": 82.42
      },
      "soft_dict_1.0": {
       "lemma_accuracy": 79.43,
       "lemma_accuracy_nomacron": 79.43,
       "pos_accuracy": 82.42
      },
      "soft_train_1.0": {
       "lemma_accuracy": 78.86,
       "lemma_accuracy_nomacron": 78.86,
       "pos_accuracy": 82.42
      },
      "soft_dict_2.0": {
       "lemma_accuracy": 80.34,
       "lemma_accuracy_nomacron": 80.34,
       "pos_accuracy": 82.42
      },
      "soft_train_2.0": {
       "lemma_accuracy": 79.14,
       "lemma_accuracy_nomacron": 79.14,
       "pos_accuracy": 82.42
      },
      "soft_dict_4.0": {
       "lemma_accuracy": 81.54,
       "lemma_accuracy_nomacron": 81.54,
       "pos_accuracy": 82.42
      },
      "soft_train_4.0": {
       "lemma_accuracy": 78.67,
       "lemma_accuracy_nomacron": 78.67,
       "pos_accuracy": 82.42
      },
      "soft_dict_8.0": {
       "lemma_accuracy": 81.23,
       "lemma_accuracy_nomacron": 81.23,
       "pos_accuracy": 82.42
      },
      "soft_train_8.0": {
       "lemma_accuracy": 74.01,
       "lemma_accuracy_nomacron": 74.01,
       "pos_accuracy": 82.42
      },
      "soft_dict_16.0": {
       "lemma_accuracy": 81.23,
       "lemma_accuracy_nomacron": 81.23,
       "pos_accuracy": 82.42
      },
      "soft_train_16.0": {
       "lemma_accuracy": 74.01,
       "lemma_accuracy_nomacron": 74.01,
       "pos_accuracy": 82.42
      }
     }
    },
    "test": {
     "n": 10507,
     "oracle_in_beam": 95.45,
     "strategies": {
      "beam_top1": {
       "lemma_accuracy": 78.45,
       "lemma_accuracy_nomacron": 78.45,
       "pos_accuracy": 82.59
      },
      "filter_train": {
       "lemma_accuracy": 73.81,
       "lemma_accuracy_nomacron": 73.81,
       "pos_accuracy": 82.59
      },
      "filter_dict": {
       "lemma_accuracy": 81.26,
       "lemma_accuracy_nomacron": 81.26,
       "pos_accuracy": 82.59
      },
      "soft_dict_0.5": {
       "lemma_accuracy": 78.88,
       "lemma_accuracy_nomacron": 78.88,
       "pos_accuracy": 82.59
      },
      "soft_train_0.5": {
       "lemma_accuracy": 78.73,
       "lemma_accuracy_nomacron": 78.73,
       "pos_accuracy": 82.59
      },
      "soft_dict_1.0": {
       "lemma_accuracy": 79.43,
       "lemma_accuracy_nomacron": 79.43,
       "pos_accuracy": 82.59
      },
      "soft_train_1.0": {
       "lemma_accuracy": 78.95,
       "lemma_accuracy_nomacron": 78.95,
       "pos_accuracy": 82.59
      },
      "soft_dict_2.0": {
       "lemma_accuracy": 80.48,
       "lemma_accuracy_nomacron": 80.48,
       "pos_accuracy": 82.59
      },
      "soft_train_2.0": {
       "lemma_accuracy": 79.19,
       "lemma_accuracy_nomacron": 79.19,
       "pos_accuracy": 82.59
      },
      "soft_dict_4.0": {
       "lemma_accuracy": 81.45,
       "lemma_accuracy_nomacron": 81.45,
       "pos_accuracy": 82.59
      },
      "soft_train_4.0": {
       "lemma_accuracy": 78.7,
       "lemma_accuracy_nomacron": 78.7,
       "pos_accuracy": 82.59
      },
      "soft_dict_8.0": {
       "lemma_accuracy": 81.26,
       "lemma_accuracy_nomacron": 81.26,
       "pos_accuracy": 82.59
      },
      "soft_train_8.0": {
       "lemma_accuracy": 73.81,
       "lemma_accuracy_nomacron": 73.81,
       "pos_accuracy": 82.59
      },
      "soft_dict_16.0": {
       "lemma_accuracy": 81.26,
       "lemma_accuracy_nomacron": 81.26,
       "pos_accuracy": 82.59
      },
      "soft_train_16.0": {
       "lemma_accuracy": 73.81,
       "lemma_accuracy_nomacron": 73.81,
       "pos_accuracy": 82.59
      }
     }
    }
   }
  },
  "C14_expert": {
   "checkpoint": "../experiments/C14_expert/best_model.pt",
   "label": "C14_expert",
   "beam_size": 10,
   "universe": "revised6",
   "splits": {
    "dev": {
     "n": 10506,
     "oracle_in_beam": 95.63,
     "strategies": {
      "beam_top1": {
       "lemma_accuracy": 78.76,
       "lemma_accuracy_nomacron": 78.76,
       "pos_accuracy": 82.45
      },
      "filter_train": {
       "lemma_accuracy": 73.8,
       "lemma_accuracy_nomacron": 73.8,
       "pos_accuracy": 82.45
      },
      "filter_dict": {
       "lemma_accuracy": 81.89,
       "lemma_accuracy_nomacron": 81.89,
       "pos_accuracy": 82.45
      },
      "soft_dict_0.5": {
       "lemma_accuracy": 79.41,
       "lemma_accuracy_nomacron": 79.41,
       "pos_accuracy": 82.45
      },
      "soft_train_0.5": {
       "lemma_accuracy": 79.02,
       "lemma_accuracy_nomacron": 79.02,
       "pos_accuracy": 82.45
      },
      "soft_dict_1.0": {
       "lemma_accuracy": 79.95,
       "lemma_accuracy_nomacron": 79.95,
       "pos_accuracy": 82.45
      },
      "soft_train_1.0": {
       "lemma_accuracy": 79.23,
       "lemma_accuracy_nomacron": 79.23,
       "pos_accuracy": 82.45
      },
      "soft_dict_2.0": {
       "lemma_accuracy": 80.82,
       "lemma_accuracy_nomacron": 80.82,
       "pos_accuracy": 82.45
      },
      "soft_train_2.0": {
       "lemma_accuracy": 79.44,
       "lemma_accuracy_nomacron": 79.44,
       "pos_accuracy": 82.45
      },
      "soft_dict_4.0": {
       "lemma_accuracy": 82.22,
       "lemma_accuracy_nomacron": 82.22,
       "pos_accuracy": 82.45
      },
      "soft_train_4.0": {
       "lemma_accuracy": 79.15,
       "lemma_accuracy_nomacron": 79.15,
       "pos_accuracy": 82.45
      },
      "soft_dict_8.0": {
       "lemma_accuracy": 81.89,
       "lemma_accuracy_nomacron": 81.89,
       "pos_accuracy": 82.45
      },
      "soft_train_8.0": {
       "lemma_accuracy": 73.8,
       "lemma_accuracy_nomacron": 73.8,
       "pos_accuracy": 82.45
      },
      "soft_dict_16.0": {
       "lemma_accuracy": 81.89,
       "lemma_accuracy_nomacron": 81.89,
       "pos_accuracy": 82.45
      },
      "soft_train_16.0": {
       "lemma_accuracy": 73.8,
       "lemma_accuracy_nomacron": 73.8,
       "pos_accuracy": 82.45
      }
     }
    },
    "test": {
     "n": 10507,
     "oracle_in_beam": 95.23,
     "strategies": {
      "beam_top1": {
       "lemma_accuracy": 78.3,
       "lemma_accuracy_nomacron": 78.3,
       "pos_accuracy": 82.78
      },
      "filter_train": {
       "lemma_accuracy": 73.1,
       "lemma_accuracy_nomacron": 73.1,
       "pos_accuracy": 82.78
      },
      "filter_dict": {
       "lemma_accuracy": 81.3,
       "lemma_accuracy_nomacron": 81.3,
       "pos_accuracy": 82.78
      },
      "soft_dict_0.5": {
       "lemma_accuracy": 78.87,
       "lemma_accuracy_nomacron": 78.87,
       "pos_accuracy": 82.78
      },
      "soft_train_0.5": {
       "lemma_accuracy": 78.55,
       "lemma_accuracy_nomacron": 78.55,
       "pos_accuracy": 82.78
      },
      "soft_dict_1.0": {
       "lemma_accuracy": 79.38,
       "lemma_accuracy_nomacron": 79.38,
       "pos_accuracy": 82.78
      },
      "soft_train_1.0": {
       "lemma_accuracy": 78.73,
       "lemma_accuracy_nomacron": 78.73,
       "pos_accuracy": 82.78
      },
      "soft_dict_2.0": {
       "lemma_accuracy": 80.27,
       "lemma_accuracy_nomacron": 80.27,
       "pos_accuracy": 82.78
      },
      "soft_train_2.0": {
       "lemma_accuracy": 79.09,
       "lemma_accuracy_nomacron": 79.09,
       "pos_accuracy": 82.78
      },
      "soft_dict_4.0": {
       "lemma_accuracy": 81.45,
       "lemma_accuracy_nomacron": 81.45,
       "pos_accuracy": 82.78
      },
      "soft_train_4.0": {
       "lemma_accuracy": 78.84,
       "lemma_accuracy_nomacron": 78.84,
       "pos_accuracy": 82.78
      },
      "soft_dict_8.0": {
       "lemma_accuracy": 81.3,
       "lemma_accuracy_nomacron": 81.3,
       "pos_accuracy": 82.78
      },
      "soft_train_8.0": {
       "lemma_accuracy": 73.1,
       "lemma_accuracy_nomacron": 73.1,
       "pos_accuracy": 82.78
      },
      "soft_dict_16.0": {
       "lemma_accuracy": 81.3,
       "lemma_accuracy_nomacron": 81.3,
       "pos_accuracy": 82.78
      },
      "soft_train_16.0": {
       "lemma_accuracy": 73.1,
       "lemma_accuracy_nomacron": 73.1,
       "pos_accuracy": 82.78
      }
     }
    }
   }
  },
  "C15_canonical": {
   "checkpoint": "../experiments/C15_canonical/best_model.pt",
   "label": "C15_canonical",
   "beam_size": 10,
   "universe": "revised7",
   "splits": {
    "dev": {
     "n": 10506,
     "oracle_in_beam": 95.49,
     "strategies": {
      "beam_top1": {
       "lemma_accuracy": 78.81,
       "lemma_accuracy_nomacron": 78.81,
       "pos_accuracy": 82.39
      },
      "filter_train": {
       "lemma_accuracy": 73.23,
       "lemma_accuracy_nomacron": 73.23,
       "pos_accuracy": 82.39
      },
      "filter_dict": {
       "lemma_accuracy": 81.56,
       "lemma_accuracy_nomacron": 81.56,
       "pos_accuracy": 82.39
      },
      "soft_dict_0.5": {
       "lemma_accuracy": 79.39,
       "lemma_accuracy_nomacron": 79.39,
       "pos_accuracy": 82.39
      },
      "soft_train_0.5": {
       "lemma_accuracy": 79.09,
       "lemma_accuracy_nomacron": 79.09,
       "pos_accuracy": 82.39
      },
      "soft_dict_1.0": {
       "lemma_accuracy": 79.88,
       "lemma_accuracy_nomacron": 79.88,
       "pos_accuracy": 82.39
      },
      "soft_train_1.0": {
       "lemma_accuracy": 79.26,
       "lemma_accuracy_nomacron": 79.26,
       "pos_accuracy": 82.39
      },
      "soft_dict_2.0": {
       "lemma_accuracy": 80.7,
       "lemma_accuracy_nomacron": 80.7,
       "pos_accuracy": 82.39
      },
      "soft_train_2.0": {
       "lemma_accuracy": 79.48,
       "lemma_accuracy_nomacron": 79.48,
       "pos_accuracy": 82.39
      },
      "soft_dict_4.0": {
       "lemma_accuracy": 81.8,
       "lemma_accuracy_nomacron": 81.8,
       "pos_accuracy": 82.39
      },
      "soft_train_4.0": {
       "lemma_accuracy": 79.18,
       "lemma_accuracy_nomacron": 79.18,
       "pos_accuracy": 82.39
      },
      "soft_dict_8.0": {
       "lemma_accuracy": 81.56,
       "lemma_accuracy_nomacron": 81.56,
       "pos_accuracy": 82.39
      },
      "soft_train_8.0": {
       "lemma_accuracy": 73.23,
       "lemma_accuracy_nomacron": 73.23,
       "pos_accuracy": 82.39
      },
      "soft_dict_16.0": {
       "lemma_accuracy": 81.56,
       "lemma_accuracy_nomacron": 81.56,
       "pos_accuracy": 82.39
      },
      "soft_train_16.0": {
       "lemma_accuracy": 73.23,
       "lemma_accuracy_nomacron": 73.23,
       "pos_accuracy": 82.39
      }
     }
    },
    "test": {
     "n": 10507,
     "oracle_in_beam": 95.37,
     "strategies": {
      "beam_top1": {
       "lemma_accuracy": 78.51,
       "lemma_accuracy_nomacron": 78.51,
       "pos_accuracy": 82.79
      },
      "filter_train": {
       "lemma_accuracy": 73.0,
       "lemma_accuracy_nomacron": 73.0,
       "pos_accuracy": 82.79
      },
      "filter_dict": {
       "lemma_accuracy": 81.44,
       "lemma_accuracy_nomacron": 81.44,
       "pos_accuracy": 82.79
      },
      "soft_dict_0.5": {
       "lemma_accuracy": 78.95,
       "lemma_accuracy_nomacron": 78.95,
       "pos_accuracy": 82.79
      },
      "soft_train_0.5": {
       "lemma_accuracy": 78.74,
       "lemma_accuracy_nomacron": 78.74,
       "pos_accuracy": 82.79
      },
      "soft_dict_1.0": {
       "lemma_accuracy": 79.4,
       "lemma_accuracy_nomacron": 79.4,
       "pos_accuracy": 82.79
      },
      "soft_train_1.0": {
       "lemma_accuracy": 78.9,
       "lemma_accuracy_nomacron": 78.9,
       "pos_accuracy": 82.79
      },
      "soft_dict_2.0": {
       "lemma_accuracy": 80.41,
       "lemma_accuracy_nomacron": 80.41,
       "pos_accuracy": 82.79
      },
      "soft_train_2.0": {
       "lemma_accuracy": 79.35,
       "lemma_accuracy_nomacron": 79.35,
       "pos_accuracy": 82.79
      },
      "soft_dict_4.0": {
       "lemma_accuracy": 81.57,
       "lemma_accuracy_nomacron": 81.57,
       "pos_accuracy": 82.79
      },
      "soft_train_4.0": {
       "lemma_accuracy": 79.08,
       "lemma_accuracy_nomacron": 79.08,
       "pos_accuracy": 82.79
      },
      "soft_dict_8.0": {
       "lemma_accuracy": 81.44,
       "lemma_accuracy_nomacron": 81.44,
       "pos_accuracy": 82.79
      },
      "soft_train_8.0": {
       "lemma_accuracy": 73.0,
       "lemma_accuracy_nomacron": 73.0,
       "pos_accuracy": 82.79
      },
      "soft_dict_16.0": {
       "lemma_accuracy": 81.44,
       "lemma_accuracy_nomacron": 81.44,
       "pos_accuracy": 82.79
      },
      "soft_train_16.0": {
       "lemma_accuracy": 73.0,
       "lemma_accuracy_nomacron": 73.0,
       "pos_accuracy": 82.79
      }
     }
    }
   }
  },
  "C1_variant": {
   "checkpoint": "/Users/jamartin/OE_lemmatiser_retrain/experiments/C1_variant/best_model.pt",
   "label": "C1_variant",
   "beam_size": 10,
   "universe": "original",
   "splits": {
    "dev": {
     "n": 10506,
     "oracle_in_beam": 93.27,
     "strategies": {
      "beam_top1": {
       "lemma_accuracy": 69.08,
       "lemma_accuracy_nomacron": 73.58,
       "pos_accuracy": 82.53
      },
      "filter_train": {
       "lemma_accuracy": 64.57,
       "lemma_accuracy_nomacron": 69.33,
       "pos_accuracy": 82.53
      },
      "filter_dict": {
       "lemma_accuracy": 67.49,
       "lemma_accuracy_nomacron": 74.38,
       "pos_accuracy": 82.53
      },
      "soft_dict_0.5": {
       "lemma_accuracy": 69.6,
       "lemma_accuracy_nomacron": 74.07,
       "pos_accuracy": 82.53
      },
      "soft_train_0.5": {
       "lemma_accuracy": 69.36,
       "lemma_accuracy_nomacron": 73.86,
       "pos_accuracy": 82.53
      },
      "soft_dict_1.0": {
       "lemma_accuracy": 69.84,
       "lemma_accuracy_nomacron": 74.44,
       "pos_accuracy": 82.53
      },
      "soft_train_1.0": {
       "lemma_accuracy": 69.4,
       "lemma_accuracy_nomacron": 73.96,
       "pos_accuracy": 82.53
      },
      "soft_dict_2.0": {
       "lemma_accuracy": 70.16,
       "lemma_accuracy_nomacron": 75.05,
       "pos_accuracy": 82.53
      },
      "soft_train_2.0": {
       "lemma_accuracy": 69.41,
       "lemma_accuracy_nomacron": 74.14,
       "pos_accuracy": 82.53
      },
      "soft_dict_4.0": {
       "lemma_accuracy": 69.58,
       "lemma_accuracy_nomacron": 75.59,
       "pos_accuracy": 82.53
      },
      "soft_train_4.0": {
       "lemma_accuracy": 68.88,
       "lemma_accuracy_nomacron": 73.67,
       "pos_accuracy": 82.53
      },
      "soft_dict_8.0": {
       "lemma_accuracy": 67.49,
       "lemma_accuracy_nomacron": 74.38,
       "pos_accuracy": 82.53
      },
      "soft_train_8.0": {
       "lemma_accuracy": 64.57,
       "lemma_accuracy_nomacron": 69.33,
       "pos_accuracy": 82.53
      },
      "soft_dict_16.0": {
       "lemma_accuracy": 67.49,
       "lemma_accuracy_nomacron": 74.38,
       "pos_accuracy": 82.53
      },
      "soft_train_16.0": {
       "lemma_accuracy": 64.57,
       "lemma_accuracy_nomacron": 69.33,
       "pos_accuracy": 82.53
      },
      "macron_dict": {
       "lemma_accuracy": 55.08,
       "lemma_accuracy_nomacron": 73.58,
       "pos_accuracy": 82.53
      },
      "macron_corpus": {
       "lemma_accuracy": 66.13,
       "lemma_accuracy_nomacron": 73.58,
       "pos_accuracy": 82.53
      },
      "soft_dict_1.0+macron_corpus": {
       "lemma_accuracy": 67.01,
       "lemma_accuracy_nomacron": 74.44,
       "pos_accuracy": 82.53
      }
     }
    },
    "test": {
     "n": 10507,
     "oracle_in_beam": 93.26,
     "strategies": {
      "beam_top1": {
       "lemma_accuracy": 68.18,
       "lemma_accuracy_nomacron": 72.97,
       "pos_accuracy": 82.68
      },
      "filter_train": {
       "lemma_accuracy": 63.91,
       "lemma_accuracy_nomacron": 69.01,
       "pos_accuracy": 82.68
      },
      "filter_dict": {
       "lemma_accuracy": 66.72,
       "lemma_accuracy_nomacron": 73.88,
       "pos_accuracy": 82.68
      },
      "soft_dict_0.5": {
       "lemma_accuracy": 68.54,
       "lemma_accuracy_nomacron": 73.45,
       "pos_accuracy": 82.68
      },
      "soft_train_0.5": {
       "lemma_accuracy": 68.47,
       "lemma_accuracy_nomacron": 73.29,
       "pos_accuracy": 82.68
      },
      "soft_dict_1.0": {
       "lemma_accuracy": 68.96,
       "lemma_accuracy_nomacron": 73.86,
       "pos_accuracy": 82.68
      },
      "soft_train_1.0": {
       "lemma_accuracy": 68.57,
       "lemma_accuracy_nomacron": 73.43,
       "pos_accuracy": 82.68
      },
      "soft_dict_2.0": {
       "lemma_accuracy": 69.39,
       "lemma_accuracy_nomacron": 74.62,
       "pos_accuracy": 82.68
      },
      "soft_train_2.0": {
       "lemma_accuracy": 68.75,
       "lemma_accuracy_nomacron": 73.81,
       "pos_accuracy": 82.68
      },
      "soft_dict_4.0": {
       "lemma_accuracy": 68.58,
       "lemma_accuracy_nomacron": 75.01,
       "pos_accuracy": 82.68
      },
      "soft_train_4.0": {
       "lemma_accuracy": 68.09,
       "lemma_accuracy_nomacron": 73.29,
       "pos_accuracy": 82.68
      },
      "soft_dict_8.0": {
       "lemma_accuracy": 66.72,
       "lemma_accuracy_nomacron": 73.88,
       "pos_accuracy": 82.68
      },
      "soft_train_8.0": {
       "lemma_accuracy": 63.91,
       "lemma_accuracy_nomacron": 69.01,
       "pos_accuracy": 82.68
      },
      "soft_dict_16.0": {
       "lemma_accuracy": 66.72,
       "lemma_accuracy_nomacron": 73.88,
       "pos_accuracy": 82.68
      },
      "soft_train_16.0": {
       "lemma_accuracy": 63.91,
       "lemma_accuracy_nomacron": 69.01,
       "pos_accuracy": 82.68
      },
      "macron_dict": {
       "lemma_accuracy": 55.18,
       "lemma_accuracy_nomacron": 72.97,
       "pos_accuracy": 82.68
      },
      "macron_corpus": {
       "lemma_accuracy": 65.17,
       "lemma_accuracy_nomacron": 72.97,
       "pos_accuracy": 82.68
      },
      "soft_dict_1.0+macron_corpus": {
       "lemma_accuracy": 66.11,
       "lemma_accuracy_nomacron": 73.86,
       "pos_accuracy": 82.68
      }
     }
    }
   }
  },
  "C5_normalised": {
   "checkpoint": "/Users/jamartin/OE_lemmatiser_retrain/experiments/C5_normalised/best_model.pt",
   "label": "C5_normalised",
   "beam_size": 10,
   "universe": "normalised",
   "splits": {
    "dev": {
     "n": 10506,
     "oracle_in_beam": 94.48,
     "strategies": {
      "beam_top1": {
       "lemma_accuracy": 74.96,
       "lemma_accuracy_nomacron": 74.96,
       "pos_accuracy": 82.72
      },
      "filter_train": {
       "lemma_accuracy": 69.92,
       "lemma_accuracy_nomacron": 69.92,
       "pos_accuracy": 82.72
      },
      "filter_dict": {
       "lemma_accuracy": 77.16,
       "lemma_accuracy_nomacron": 77.16,
       "pos_accuracy": 82.72
      },
      "soft_dict_0.5": {
       "lemma_accuracy": 75.71,
       "lemma_accuracy_nomacron": 75.71,
       "pos_accuracy": 82.72
      },
      "soft_train_0.5": {
       "lemma_accuracy": 75.25,
       "lemma_accuracy_nomacron": 75.25,
       "pos_accuracy": 82.72
      },
      "soft_dict_1.0": {
       "lemma_accuracy": 76.26,
       "lemma_accuracy_nomacron": 76.26,
       "pos_accuracy": 82.72
      },
      "soft_train_1.0": {
       "lemma_accuracy": 75.36,
       "lemma_accuracy_nomacron": 75.36,
       "pos_accuracy": 82.72
      },
      "soft_dict_2.0": {
       "lemma_accuracy": 76.97,
       "lemma_accuracy_nomacron": 76.97,
       "pos_accuracy": 82.72
      },
      "soft_train_2.0": {
       "lemma_accuracy": 75.42,
       "lemma_accuracy_nomacron": 75.42,
       "pos_accuracy": 82.72
      },
      "soft_dict_4.0": {
       "lemma_accuracy": 77.93,
       "lemma_accuracy_nomacron": 77.93,
       "pos_accuracy": 82.72
      },
      "soft_train_4.0": {
       "lemma_accuracy": 74.69,
       "lemma_accuracy_nomacron": 74.69,
       "pos_accuracy": 82.72
      },
      "soft_dict_8.0": {
       "lemma_accuracy": 77.16,
       "lemma_accuracy_nomacron": 77.16,
       "pos_accuracy": 82.72
      },
      "soft_train_8.0": {
       "lemma_accuracy": 69.92,
       "lemma_accuracy_nomacron": 69.92,
       "pos_accuracy": 82.72
      },
      "soft_dict_16.0": {
       "lemma_accuracy": 77.16,
       "lemma_accuracy_nomacron": 77.16,
       "pos_accuracy": 82.72
      },
      "soft_train_16.0": {
       "lemma_accuracy": 69.92,
       "lemma_accuracy_nomacron": 69.92,
       "pos_accuracy": 82.72
      }
     }
    },
    "test": {
     "n": 10507,
     "oracle_in_beam": 94.43,
     "strategies": {
      "beam_top1": {
       "lemma_accuracy": 74.65,
       "lemma_accuracy_nomacron": 74.65,
       "pos_accuracy": 82.79
      },
      "filter_train": {
       "lemma_accuracy": 69.71,
       "lemma_accuracy_nomacron": 69.71,
       "pos_accuracy": 82.79
      },
      "filter_dict": {
       "lemma_accuracy": 76.72,
       "lemma_accuracy_nomacron": 76.72,
       "pos_accuracy": 82.79
      },
      "soft_dict_0.5": {
       "lemma_accuracy": 75.22,
       "lemma_accuracy_nomacron": 75.22,
       "pos_accuracy": 82.79
      },
      "soft_train_0.5": {
       "lemma_accuracy": 74.81,
       "lemma_accuracy_nomacron": 74.81,
       "pos_accuracy": 82.79
      },
      "soft_dict_1.0": {
       "lemma_accuracy": 75.78,
       "lemma_accuracy_nomacron": 75.78,
       "pos_accuracy": 82.79
      },
      "soft_train_1.0": {
       "lemma_accuracy": 75.02,
       "lemma_accuracy_nomacron": 75.02,
       "pos_accuracy": 82.79
      },
      "soft_dict_2.0": {
       "lemma_accuracy": 76.66,
       "lemma_accuracy_nomacron": 76.66,
       "pos_accuracy": 82.79
      },
      "soft_train_2.0": {
       "lemma_accuracy": 75.2,
       "lemma_accuracy_nomacron": 75.2,
       "pos_accuracy": 82.79
      },
      "soft_dict_4.0": {
       "lemma_accuracy": 77.56,
       "lemma_accuracy_nomacron": 77.56,
       "pos_accuracy": 82.79
      },
      "soft_train_4.0": {
       "lemma_accuracy": 74.65,
       "lemma_accuracy_nomacron": 74.65,
       "pos_accuracy": 82.79
      },
      "soft_dict_8.0": {
       "lemma_accuracy": 76.72,
       "lemma_accuracy_nomacron": 76.72,
       "pos_accuracy": 82.79
      },
      "soft_train_8.0": {
       "lemma_accuracy": 69.72,
       "lemma_accuracy_nomacron": 69.72,
       "pos_accuracy": 82.79
      },
      "soft_dict_16.0": {
       "lemma_accuracy": 76.72,
       "lemma_accuracy_nomacron": 76.72,
       "pos_accuracy": 82.79
      },
      "soft_train_16.0": {
       "lemma_accuracy": 69.71,
       "lemma_accuracy_nomacron": 69.71,
       "pos_accuracy": 82.79
      }
     }
    }
   }
  },
  "C6_combined_normalised": {
   "checkpoint": "/Users/jamartin/OE_lemmatiser_retrain/experiments/C6_combined_normalised/best_model.pt",
   "label": "C6_combined_normalised",
   "beam_size": 10,
   "universe": "normalised",
   "splits": {
    "dev": {
     "n": 10506,
     "oracle_in_beam": 94.84,
     "strategies": {
      "beam_top1": {
       "lemma_accuracy": 70.61,
       "lemma_accuracy_nomacron": 70.61,
       "pos_accuracy": 81.48
      },
      "filter_train": {
       "lemma_accuracy": 65.86,
       "lemma_accuracy_nomacron": 65.86,
       "pos_accuracy": 81.48
      },
      "filter_dict": {
       "lemma_accuracy": 71.51,
       "lemma_accuracy_nomacron": 71.51,
       "pos_accuracy": 81.48
      },
      "soft_dict_0.5": {
       "lemma_accuracy": 70.98,
       "lemma_accuracy_nomacron": 70.98,
       "pos_accuracy": 81.48
      },
      "soft_train_0.5": {
       "lemma_accuracy": 71.27,
       "lemma_accuracy_nomacron": 71.27,
       "pos_accuracy": 81.48
      },
      "soft_dict_1.0": {
       "lemma_accuracy": 71.26,
       "lemma_accuracy_nomacron": 71.26,
       "pos_accuracy": 81.48
      },
      "soft_train_1.0": {
       "lemma_accuracy": 71.52,
       "lemma_accuracy_nomacron": 71.52,
       "pos_accuracy": 81.48
      },
      "soft_dict_2.0": {
       "lemma_accuracy": 71.78,
       "lemma_accuracy_nomacron": 71.78,
       "pos_accuracy": 81.48
      },
      "soft_train_2.0": {
       "lemma_accuracy": 71.33,
       "lemma_accuracy_nomacron": 71.33,
       "pos_accuracy": 81.48
      },
      "soft_dict_4.0": {
       "lemma_accuracy": 72.19,
       "lemma_accuracy_nomacron": 72.19,
       "pos_accuracy": 81.48
      },
      "soft_train_4.0": {
       "lemma_accuracy": 70.03,
       "lemma_accuracy_nomacron": 70.03,
       "pos_accuracy": 81.48
      },
      "soft_dict_8.0": {
       "lemma_accuracy": 71.51,
       "lemma_accuracy_nomacron": 71.51,
       "pos_accuracy": 81.48
      },
      "soft_train_8.0": {
       "lemma_accuracy": 65.86,
       "lemma_accuracy_nomacron": 65.86,
       "pos_accuracy": 81.48
      },
      "soft_dict_16.0": {
       "lemma_accuracy": 71.51,
       "lemma_accuracy_nomacron": 71.51,
       "pos_accuracy": 81.48
      },
      "soft_train_16.0": {
       "lemma_accuracy": 65.86,
       "lemma_accuracy_nomacron": 65.86,
       "pos_accuracy": 81.48
      }
     }
    },
    "test": {
     "n": 10507,
     "oracle_in_beam": 94.83,
     "strategies": {
      "beam_top1": {
       "lemma_accuracy": 69.87,
       "lemma_accuracy_nomacron": 69.87,
       "pos_accuracy": 81.67
      },
      "filter_train": {
       "lemma_accuracy": 65.89,
       "lemma_accuracy_nomacron": 65.89,
       "pos_accuracy": 81.67
      },
      "filter_dict": {
       "lemma_accuracy": 71.32,
       "lemma_accuracy_nomacron": 71.32,
       "pos_accuracy": 81.67
      },
      "soft_dict_0.5": {
       "lemma_accuracy": 70.52,
       "lemma_accuracy_nomacron": 70.52,
       "pos_accuracy": 81.67
      },
      "soft_train_0.5": {
       "lemma_accuracy": 70.92,
       "lemma_accuracy_nomacron": 70.92,
       "pos_accuracy": 81.67
      },
      "soft_dict_1.0": {
       "lemma_accuracy": 70.91,
       "lemma_accuracy_nomacron": 70.91,
       "pos_accuracy": 81.67
      },
      "soft_train_1.0": {
       "lemma_accuracy": 71.31,
       "lemma_accuracy_nomacron": 71.31,
       "pos_accuracy": 81.67
      },
      "soft_dict_2.0": {
       "lemma_accuracy": 71.32,
       "lemma_accuracy_nomacron": 71.32,
       "pos_accuracy": 81.67
      },
      "soft_train_2.0": {
       "lemma_accuracy": 71.36,
       "lemma_accuracy_nomacron": 71.36,
       "pos_accuracy": 81.67
      },
      "soft_dict_4.0": {
       "lemma_accuracy": 71.73,
       "lemma_accuracy_nomacron": 71.73,
       "pos_accuracy": 81.67
      },
      "soft_train_4.0": {
       "lemma_accuracy": 70.18,
       "lemma_accuracy_nomacron": 70.18,
       "pos_accuracy": 81.67
      },
      "soft_dict_8.0": {
       "lemma_accuracy": 71.32,
       "lemma_accuracy_nomacron": 71.32,
       "pos_accuracy": 81.67
      },
      "soft_train_8.0": {
       "lemma_accuracy": 65.89,
       "lemma_accuracy_nomacron": 65.89,
       "pos_accuracy": 81.67
      },
      "soft_dict_16.0": {
       "lemma_accuracy": 71.32,
       "lemma_accuracy_nomacron": 71.32,
       "pos_accuracy": 81.67
      },
      "soft_train_16.0": {
       "lemma_accuracy": 65.89,
       "lemma_accuracy_nomacron": 65.89,
       "pos_accuracy": 81.67
      }
     }
    }
   }
  },
  "C7_variant_normalised": {
   "checkpoint": "/Users/jamartin/OE_lemmatiser_retrain/experiments/C7_variant_normalised/best_model.pt",
   "label": "C7_variant_normalised",
   "beam_size": 10,
   "universe": "normalised",
   "splits": {
    "dev": {
     "n": 10506,
     "oracle_in_beam": 94.58,
     "strategies": {
      "beam_top1": {
       "lemma_accuracy": 75.02,
       "lemma_accuracy_nomacron": 75.02,
       "pos_accuracy": 82.54
      },
      "filter_train": {
       "lemma_accuracy": 70.41,
       "lemma_accuracy_nomacron": 70.41,
       "pos_accuracy": 82.54
      },
      "filter_dict": {
       "lemma_accuracy": 77.56,
       "lemma_accuracy_nomacron": 77.56,
       "pos_accuracy": 82.54
      },
      "soft_dict_0.5": {
       "lemma_accuracy": 75.68,
       "lemma_accuracy_nomacron": 75.68,
       "pos_accuracy": 82.54
      },
      "soft_train_0.5": {
       "lemma_accuracy": 75.26,
       "lemma_accuracy_nomacron": 75.26,
       "pos_accuracy": 82.54
      },
      "soft_dict_1.0": {
       "lemma_accuracy": 76.15,
       "lemma_accuracy_nomacron": 76.15,
       "pos_accuracy": 82.54
      },
      "soft_train_1.0": {
       "lemma_accuracy": 75.35,
       "lemma_accuracy_nomacron": 75.35,
       "pos_accuracy": 82.54
      },
      "soft_dict_2.0": {
       "lemma_accuracy": 76.99,
       "lemma_accuracy_nomacron": 76.99,
       "pos_accuracy": 82.54
      },
      "soft_train_2.0": {
       "lemma_accuracy": 75.52,
       "lemma_accuracy_nomacron": 75.52,
       "pos_accuracy": 82.54
      },
      "soft_dict_4.0": {
       "lemma_accuracy": 78.21,
       "lemma_accuracy_nomacron": 78.21,
       "pos_accuracy": 82.54
      },
      "soft_train_4.0": {
       "lemma_accuracy": 75.22,
       "lemma_accuracy_nomacron": 75.22,
       "pos_accuracy": 82.54
      },
      "soft_dict_8.0": {
       "lemma_accuracy": 77.56,
       "lemma_accuracy_nomacron": 77.56,
       "pos_accuracy": 82.54
      },
      "soft_train_8.0": {
       "lemma_accuracy": 70.41,
       "lemma_accuracy_nomacron": 70.41,
       "pos_accuracy": 82.54
      },
      "soft_dict_16.0": {
       "lemma_accuracy": 77.56,
       "lemma_accuracy_nomacron": 77.56,
       "pos_accuracy": 82.54
      },
      "soft_train_16.0": {
       "lemma_accuracy": 70.41,
       "lemma_accuracy_nomacron": 70.41,
       "pos_accuracy": 82.54
      }
     }
    },
    "test": {
     "n": 10507,
     "oracle_in_beam": 94.42,
     "strategies": {
      "beam_top1": {
       "lemma_accuracy": 74.73,
       "lemma_accuracy_nomacron": 74.73,
       "pos_accuracy": 82.77
      },
      "filter_train": {
       "lemma_accuracy": 69.75,
       "lemma_accuracy_nomacron": 69.75,
       "pos_accuracy": 82.77
      },
      "filter_dict": {
       "lemma_accuracy": 76.86,
       "lemma_accuracy_nomacron": 76.86,
       "pos_accuracy": 82.77
      },
      "soft_dict_0.5": {
       "lemma_accuracy": 75.21,
       "lemma_accuracy_nomacron": 75.21,
       "pos_accuracy": 82.77
      },
      "soft_train_0.5": {
       "lemma_accuracy": 74.92,
       "lemma_accuracy_nomacron": 74.92,
       "pos_accuracy": 82.77
      },
      "soft_dict_1.0": {
       "lemma_accuracy": 75.82,
       "lemma_accuracy_nomacron": 75.82,
       "pos_accuracy": 82.77
      },
      "soft_train_1.0": {
       "lemma_accuracy": 75.24,
       "lemma_accuracy_nomacron": 75.24,
       "pos_accuracy": 82.77
      },
      "soft_dict_2.0": {
       "lemma_accuracy": 76.77,
       "lemma_accuracy_nomacron": 76.77,
       "pos_accuracy": 82.77
      },
      "soft_train_2.0": {
       "lemma_accuracy": 75.48,
       "lemma_accuracy_nomacron": 75.48,
       "pos_accuracy": 82.77
      },
      "soft_dict_4.0": {
       "lemma_accuracy": 77.7,
       "lemma_accuracy_nomacron": 77.7,
       "pos_accuracy": 82.77
      },
      "soft_train_4.0": {
       "lemma_accuracy": 74.96,
       "lemma_accuracy_nomacron": 74.96,
       "pos_accuracy": 82.77
      },
      "soft_dict_8.0": {
       "lemma_accuracy": 76.86,
       "lemma_accuracy_nomacron": 76.86,
       "pos_accuracy": 82.77
      },
      "soft_train_8.0": {
       "lemma_accuracy": 69.75,
       "lemma_accuracy_nomacron": 69.75,
       "pos_accuracy": 82.77
      },
      "soft_dict_16.0": {
       "lemma_accuracy": 76.86,
       "lemma_accuracy_nomacron": 76.86,
       "pos_accuracy": 82.77
      },
      "soft_train_16.0": {
       "lemma_accuracy": 69.75,
       "lemma_accuracy_nomacron": 69.75,
       "pos_accuracy": 82.77
      }
     }
    }
   }
  },
  "C8_conflictfree_normalised": {
   "checkpoint": "/Users/jamartin/OE_lemmatiser_retrain/experiments/C8_conflictfree_normalised/best_model.pt",
   "label": "C8_conflictfree_normalised",
   "beam_size": 10,
   "universe": "normalised",
   "splits": {
    "dev": {
     "n": 10506,
     "oracle_in_beam": 94.99,
     "strategies": {
      "beam_top1": {
       "lemma_accuracy": 75.02,
       "lemma_accuracy_nomacron": 75.02,
       "pos_accuracy": 82.65
      },
      "filter_train": {
       "lemma_accuracy": 69.32,
       "lemma_accuracy_nomacron": 69.32,
       "pos_accuracy": 82.65
      },
      "filter_dict": {
       "lemma_accuracy": 75.66,
       "lemma_accuracy_nomacron": 75.66,
       "pos_accuracy": 82.65
      },
      "soft_dict_0.5": {
       "lemma_accuracy": 75.43,
       "lemma_accuracy_nomacron": 75.43,
       "pos_accuracy": 82.65
      },
      "soft_train_0.5": {
       "lemma_accuracy": 75.5,
       "lemma_accuracy_nomacron": 75.5,
       "pos_accuracy": 82.65
      },
      "soft_dict_1.0": {
       "lemma_accuracy": 75.72,
       "lemma_accuracy_nomacron": 75.72,
       "pos_accuracy": 82.65
      },
      "soft_train_1.0": {
       "lemma_accuracy": 75.54,
       "lemma_accuracy_nomacron": 75.54,
       "pos_accuracy": 82.65
      },
      "soft_dict_2.0": {
       "lemma_accuracy": 76.17,
       "lemma_accuracy_nomacron": 76.17,
       "pos_accuracy": 82.65
      },
      "soft_train_2.0": {
       "lemma_accuracy": 75.42,
       "lemma_accuracy_nomacron": 75.42,
       "pos_accuracy": 82.65
      },
      "soft_dict_4.0": {
       "lemma_accuracy": 76.53,
       "lemma_accuracy_nomacron": 76.53,
       "pos_accuracy": 82.65
      },
      "soft_train_4.0": {
       "lemma_accuracy": 74.26,
       "lemma_accuracy_nomacron": 74.26,
       "pos_accuracy": 82.65
      },
      "soft_dict_8.0": {
       "lemma_accuracy": 75.66,
       "lemma_accuracy_nomacron": 75.66,
       "pos_accuracy": 82.65
      },
      "soft_train_8.0": {
       "lemma_accuracy": 69.32,
       "lemma_accuracy_nomacron": 69.32,
       "pos_accuracy": 82.65
      },
      "soft_dict_16.0": {
       "lemma_accuracy": 75.66,
       "lemma_accuracy_nomacron": 75.66,
       "pos_accuracy": 82.65
      },
      "soft_train_16.0": {
       "lemma_accuracy": 69.32,
       "lemma_accuracy_nomacron": 69.32,
       "pos_accuracy": 82.65
      }
     }
    },
    "test": {
     "n": 10507,
     "oracle_in_beam": 94.73,
     "strategies": {
      "beam_top1": {
       "lemma_accuracy": 73.96,
       "lemma_accuracy_nomacron": 73.96,
       "pos_accuracy": 82.82
      },
      "filter_train": {
       "lemma_accuracy": 68.82,
       "lemma_accuracy_nomacron": 68.82,
       "pos_accuracy": 82.82
      },
      "filter_dict": {
       "lemma_accuracy": 74.98,
       "lemma_accuracy_nomacron": 74.98,
       "pos_accuracy": 82.82
      },
      "soft_dict_0.5": {
       "lemma_accuracy": 74.44,
       "lemma_accuracy_nomacron": 74.44,
       "pos_accuracy": 82.82
      },
      "soft_train_0.5": {
       "lemma_accuracy": 74.51,
       "lemma_accuracy_nomacron": 74.51,
       "pos_accuracy": 82.82
      },
      "soft_dict_1.0": {
       "lemma_accuracy": 74.79,
       "lemma_accuracy_nomacron": 74.79,
       "pos_accuracy": 82.82
      },
      "soft_train_1.0": {
       "lemma_accuracy": 74.63,
       "lemma_accuracy_nomacron": 74.63,
       "pos_accuracy": 82.82
      },
      "soft_dict_2.0": {
       "lemma_accuracy": 75.31,
       "lemma_accuracy_nomacron": 75.31,
       "pos_accuracy": 82.82
      },
      "soft_train_2.0": {
       "lemma_accuracy": 74.73,
       "lemma_accuracy_nomacron": 74.73,
       "pos_accuracy": 82.82
      },
      "soft_dict_4.0": {
       "lemma_accuracy": 75.72,
       "lemma_accuracy_nomacron": 75.72,
       "pos_accuracy": 82.82
      },
      "soft_train_4.0": {
       "lemma_accuracy": 73.86,
       "lemma_accuracy_nomacron": 73.86,
       "pos_accuracy": 82.82
      },
      "soft_dict_8.0": {
       "lemma_accuracy": 74.98,
       "lemma_accuracy_nomacron": 74.98,
       "pos_accuracy": 82.82
      },
      "soft_train_8.0": {
       "lemma_accuracy": 68.82,
       "lemma_accuracy_nomacron": 68.82,
       "pos_accuracy": 82.82
      },
      "soft_dict_16.0": {
       "lemma_accuracy": 74.98,
       "lemma_accuracy_nomacron": 74.98,
       "pos_accuracy": 82.82
      },
      "soft_train_16.0": {
       "lemma_accuracy": 68.82,
       "lemma_accuracy_nomacron": 68.82,
       "pos_accuracy": 82.82
      }
     }
    }
   }
  },
  "C9_unseenonly_normalised": {
   "checkpoint": "/Users/jamartin/OE_lemmatiser_retrain/experiments/C9_unseenonly_normalised/best_model.pt",
   "label": "C9_unseenonly_normalised",
   "beam_size": 10,
   "universe": "normalised",
   "splits": {
    "dev": {
     "n": 10506,
     "oracle_in_beam": 95.12,
     "strategies": {
      "beam_top1": {
       "lemma_accuracy": 74.45,
       "lemma_accuracy_nomacron": 74.45,
       "pos_accuracy": 83.36
      },
      "filter_train": {
       "lemma_accuracy": 68.75,
       "lemma_accuracy_nomacron": 68.75,
       "pos_accuracy": 83.36
      },
      "filter_dict": {
       "lemma_accuracy": 75.0,
       "lemma_accuracy_nomacron": 75.0,
       "pos_accuracy": 83.36
      },
      "soft_dict_0.5": {
       "lemma_accuracy": 75.0,
       "lemma_accuracy_nomacron": 75.0,
       "pos_accuracy": 83.36
      },
      "soft_train_0.5": {
       "lemma_accuracy": 74.87,
       "lemma_accuracy_nomacron": 74.87,
       "pos_accuracy": 83.36
      },
      "soft_dict_1.0": {
       "lemma_accuracy": 75.37,
       "lemma_accuracy_nomacron": 75.37,
       "pos_accuracy": 83.36
      },
      "soft_train_1.0": {
       "lemma_accuracy": 75.06,
       "lemma_accuracy_nomacron": 75.06,
       "pos_accuracy": 83.36
      },
      "soft_dict_2.0": {
       "lemma_accuracy": 75.67,
       "lemma_accuracy_nomacron": 75.67,
       "pos_accuracy": 83.36
      },
      "soft_train_2.0": {
       "lemma_accuracy": 74.78,
       "lemma_accuracy_nomacron": 74.78,
       "pos_accuracy": 83.36
      },
      "soft_dict_4.0": {
       "lemma_accuracy": 76.01,
       "lemma_accuracy_nomacron": 76.01,
       "pos_accuracy": 83.36
      },
      "soft_train_4.0": {
       "lemma_accuracy": 73.73,
       "lemma_accuracy_nomacron": 73.73,
       "pos_accuracy": 83.36
      },
      "soft_dict_8.0": {
       "lemma_accuracy": 75.0,
       "lemma_accuracy_nomacron": 75.0,
       "pos_accuracy": 83.36
      },
      "soft_train_8.0": {
       "lemma_accuracy": 68.75,
       "lemma_accuracy_nomacron": 68.75,
       "pos_accuracy": 83.36
      },
      "soft_dict_16.0": {
       "lemma_accuracy": 75.0,
       "lemma_accuracy_nomacron": 75.0,
       "pos_accuracy": 83.36
      },
      "soft_train_16.0": {
       "lemma_accuracy": 68.75,
       "lemma_accuracy_nomacron": 68.75,
       "pos_accuracy": 83.36
      }
     }
    },
    "test": {
     "n": 10507,
     "oracle_in_beam": 95.1,
     "strategies": {
      "beam_top1": {
       "lemma_accuracy": 73.81,
       "lemma_accuracy_nomacron": 73.81,
       "pos_accuracy": 83.14
      },
      "filter_train": {
       "lemma_accuracy": 68.78,
       "lemma_accuracy_nomacron": 68.78,
       "pos_accuracy": 83.14
      },
      "filter_dict": {
       "lemma_accuracy": 74.58,
       "lemma_accuracy_nomacron": 74.58,
       "pos_accuracy": 83.14
      },
      "soft_dict_0.5": {
       "lemma_accuracy": 74.2,
       "lemma_accuracy_nomacron": 74.2,
       "pos_accuracy": 83.14
      },
      "soft_train_0.5": {
       "lemma_accuracy": 74.33,
       "lemma_accuracy_nomacron": 74.33,
       "pos_accuracy": 83.14
      },
      "soft_dict_1.0": {
       "lemma_accuracy": 74.58,
       "lemma_accuracy_nomacron": 74.58,
       "pos_accuracy": 83.14
      },
      "soft_train_1.0": {
       "lemma_accuracy": 74.55,
       "lemma_accuracy_nomacron": 74.55,
       "pos_accuracy": 83.14
      },
      "soft_dict_2.0": {
       "lemma_accuracy": 75.18,
       "lemma_accuracy_nomacron": 75.18,
       "pos_accuracy": 83.14
      },
      "soft_train_2.0": {
       "lemma_accuracy": 74.74,
       "lemma_accuracy_nomacron": 74.74,
       "pos_accuracy": 83.14
      },
      "soft_dict_4.0": {
       "lemma_accuracy": 75.43,
       "lemma_accuracy_nomacron": 75.43,
       "pos_accuracy": 83.14
      },
      "soft_train_4.0": {
       "lemma_accuracy": 73.78,
       "lemma_accuracy_nomacron": 73.78,
       "pos_accuracy": 83.14
      },
      "soft_dict_8.0": {
       "lemma_accuracy": 74.58,
       "lemma_accuracy_nomacron": 74.58,
       "pos_accuracy": 83.14
      },
      "soft_train_8.0": {
       "lemma_accuracy": 68.78,
       "lemma_accuracy_nomacron": 68.78,
       "pos_accuracy": 83.14
      },
      "soft_dict_16.0": {
       "lemma_accuracy": 74.58,
       "lemma_accuracy_nomacron": 74.58,
       "pos_accuracy": 83.14
      },
      "soft_train_16.0": {
       "lemma_accuracy": 68.78,
       "lemma_accuracy_nomacron": 68.78,
       "pos_accuracy": 83.14
      }
     }
    }
   }
  },
  "baseline_published": {
   "checkpoint": "/Users/jamartin/OE_lemmatiser_retrain/baseline_project/experiments/checkpoints/improved_v2/best_model.pt",
   "label": "baseline_published",
   "beam_size": 10,
   "universe": "original",
   "splits": {
    "dev": {
     "n": 10506,
     "oracle_in_beam": 93.3,
     "strategies": {
      "beam_top1": {
       "lemma_accuracy": 69.06,
       "lemma_accuracy_nomacron": 73.46,
       "pos_accuracy": 82.12
      },
      "filter_train": {
       "lemma_accuracy": 64.29,
       "lemma_accuracy_nomacron": 69.07,
       "pos_accuracy": 82.12
      },
      "filter_dict": {
       "lemma_accuracy": 67.31,
       "lemma_accuracy_nomacron": 74.11,
       "pos_accuracy": 82.12
      },
      "soft_dict_0.5": {
       "lemma_accuracy": 69.57,
       "lemma_accuracy_nomacron": 74.03,
       "pos_accuracy": 82.12
      },
      "soft_train_0.5": {
       "lemma_accuracy": 69.27,
       "lemma_accuracy_nomacron": 73.72,
       "pos_accuracy": 82.12
      },
      "soft_dict_1.0": {
       "lemma_accuracy": 69.94,
       "lemma_accuracy_nomacron": 74.46,
       "pos_accuracy": 82.12
      },
      "soft_train_1.0": {
       "lemma_accuracy": 69.4,
       "lemma_accuracy_nomacron": 73.91,
       "pos_accuracy": 82.12
      },
      "soft_dict_2.0": {
       "lemma_accuracy": 70.1,
       "lemma_accuracy_nomacron": 74.97,
       "pos_accuracy": 82.12
      },
      "soft_train_2.0": {
       "lemma_accuracy": 69.15,
       "lemma_accuracy_nomacron": 73.8,
       "pos_accuracy": 82.12
      },
      "soft_dict_4.0": {
       "lemma_accuracy": 69.25,
       "lemma_accuracy_nomacron": 75.23,
       "pos_accuracy": 82.12
      },
      "soft_train_4.0": {
       "lemma_accuracy": 68.3,
       "lemma_accuracy_nomacron": 73.07,
       "pos_accuracy": 82.12
      },
      "soft_dict_8.0": {
       "lemma_accuracy": 67.31,
       "lemma_accuracy_nomacron": 74.11,
       "pos_accuracy": 82.12
      },
      "soft_train_8.0": {
       "lemma_accuracy": 64.29,
       "lemma_accuracy_nomacron": 69.07,
       "pos_accuracy": 82.12
      },
      "soft_dict_16.0": {
       "lemma_accuracy": 67.31,
       "lemma_accuracy_nomacron": 74.11,
       "pos_accuracy": 82.12
      },
      "soft_train_16.0": {
       "lemma_accuracy": 64.29,
       "lemma_accuracy_nomacron": 69.07,
       "pos_accuracy": 82.12
      },
      "macron_dict": {
       "lemma_accuracy": 55.06,
       "lemma_accuracy_nomacron": 73.46,
       "pos_accuracy": 82.12
      },
      "macron_corpus": {
       "lemma_accuracy": 66.19,
       "lemma_accuracy_nomacron": 73.46,
       "pos_accuracy": 82.12
      },
      "soft_dict_1.0+macron_corpus": {
       "lemma_accuracy": 67.2,
       "lemma_accuracy_nomacron": 74.46,
       "pos_accuracy": 82.12
      }
     }
    },
    "test": {
     "n": 10507,
     "oracle_in_beam": 92.98,
     "strategies": {
      "beam_top1": {
       "lemma_accuracy": 68.21,
       "lemma_accuracy_nomacron": 72.76,
       "pos_accuracy": 82.75
      },
      "filter_train": {
       "lemma_accuracy": 63.84,
       "lemma_accuracy_nomacron": 68.78,
       "pos_accuracy": 82.75
      },
      "filter_dict": {
       "lemma_accuracy": 66.44,
       "lemma_accuracy_nomacron": 73.45,
       "pos_accuracy": 82.75
      },
      "soft_dict_0.5": {
       "lemma_accuracy": 68.87,
       "lemma_accuracy_nomacron": 73.41,
       "pos_accuracy": 82.75
      },
      "soft_train_0.5": {
       "lemma_accuracy": 68.53,
       "lemma_accuracy_nomacron": 73.16,
       "pos_accuracy": 82.75
      },
      "soft_dict_1.0": {
       "lemma_accuracy": 69.12,
       "lemma_accuracy_nomacron": 73.85,
       "pos_accuracy": 82.75
      },
      "soft_train_1.0": {
       "lemma_accuracy": 68.67,
       "lemma_accuracy_nomacron": 73.41,
       "pos_accuracy": 82.75
      },
      "soft_dict_2.0": {
       "lemma_accuracy": 69.45,
       "lemma_accuracy_nomacron": 74.38,
       "pos_accuracy": 82.75
      },
      "soft_train_2.0": {
       "lemma_accuracy": 68.6,
       "lemma_accuracy_nomacron": 73.46,
       "pos_accuracy": 82.75
      },
      "soft_dict_4.0": {
       "lemma_accuracy": 68.38,
       "lemma_accuracy_nomacron": 74.62,
       "pos_accuracy": 82.75
      },
      "soft_train_4.0": {
       "lemma_accuracy": 67.78,
       "lemma_accuracy_nomacron": 72.8,
       "pos_accuracy": 82.75
      },
      "soft_dict_8.0": {
       "lemma_accuracy": 66.44,
       "lemma_accuracy_nomacron": 73.45,
       "pos_accuracy": 82.75
      },
      "soft_train_8.0": {
       "lemma_accuracy": 63.84,
       "lemma_accuracy_nomacron": 68.78,
       "pos_accuracy": 82.75
      },
      "soft_dict_16.0": {
       "lemma_accuracy": 66.44,
       "lemma_accuracy_nomacron": 73.45,
       "pos_accuracy": 82.75
      },
      "soft_train_16.0": {
       "lemma_accuracy": 63.84,
       "lemma_accuracy_nomacron": 68.78,
       "pos_accuracy": 82.75
      },
      "macron_dict": {
       "lemma_accuracy": 55.11,
       "lemma_accuracy_nomacron": 72.76,
       "pos_accuracy": 82.75
      },
      "macron_corpus": {
       "lemma_accuracy": 65.03,
       "lemma_accuracy_nomacron": 72.76,
       "pos_accuracy": 82.75
      },
      "soft_dict_1.0+macron_corpus": {
       "lemma_accuracy": 66.11,
       "lemma_accuracy_nomacron": 73.85,
       "pos_accuracy": 82.75
      }
     }
    }
   }
  }
 }
}