eval_results/flores_devtest.json

{
  "test_set": "FLORES-200 devtest",
  "source": "https://github.com/facebookresearch/flores",
  "n_pairs": 1012,
  "checkpoint": "final_v2.3.pt",
  "model": "ControlMT v2.3 (139M)",
  "decoding": {
    "method": "beam_search",
    "num_beams": 6,
    "length_penalty": 1.2,
    "no_repeat_ngram_size": 3,
    "anti_lm_alpha": 0.5,
    "max_length": 256
  },
  "scoring_models": {
    "comet_kiwi": "Unbabel/wmt22-cometkiwi-da",
    "comet_da": "Unbabel/wmt22-comet-da",
    "surface": "sacrebleu (default tokenization)"
  },
  "scores": {
    "kn2en": {
      "kiwi": 0.8437,
      "comet": 0.8459,
      "bleu": 27.20,
      "chrf": 55.84
    },
    "en2kn": {
      "kiwi": 0.8663,
      "comet": 0.8443,
      "bleu": 18.50,
      "chrf": 56.12
    }
  },
  "ship_floor_verdict": {
    "comet_kn2en": "PASS (>= 0.82); 0.0059 above floor",
    "comet_en2kn": "PASS (>= 0.82); 0.0043 above floor",
    "kiwi_kn2en": "PASS (>= 0.80 aspirational)",
    "kiwi_en2kn": "ASPIRATIONAL (>= 0.85 mark — above)"
  },
  "contamination_disclosure": "FLORES-200 was created by Meta in 2022 from Wikipedia by human translators. Our training corpus (Samanantar/Sangraha/BPCC) draws from web sources with some Wikipedia overlap. The model has not seen FLORES devtest sentences specifically, but may share subject matter / entity coverage. Same risk applies to every MT model published on this benchmark."
}