OPT 175B

Natural Language Processing6 results

• Question AnsweringonBoolQ

  Accuracy· 2022-12-22

  60.1

  best: 99.87 (Mistral-Nemo 12B (HPT))
  OPT-IML: Scaling Language Model Instruction Meta Learning through the Lens of GeneralizationarXiv:2212.12017
• Question AnsweringonTruthfulQA

  MC1· 2022-11-16

  0.21

  best: 0.59 (GPT-4 (RLHF))
  Galactica: A Large Language Model for SciencearXiv:2211.09085
• Word Sense DisambiguationonBIG-bench (Anachronisms)

  Accuracy· 2022-11-16

  49.1

  best: 69.1 (Chinchilla-70B (few-shot, k=5))
  Galactica: A Large Language Model for SciencearXiv:2211.09085
• Bias DetectiononStereoSet

  ICAT Score· 2022-11-16

  60

  best: 72.97 (GPT-2 (small))
  Galactica: A Large Language Model for SciencearXiv:2211.09085
• Bias DetectiononStereoSet

  LMS· 2022-11-16

  74.8

  best: 77.6 (GPT-3 (text-davinci-002))
  Galactica: A Large Language Model for SciencearXiv:2211.09085
• Bias DetectiononStereoSet

  SS· 2022-11-16

  59.9

  best: 60.8 (GPT-3 (text-davinci-002))
  Galactica: A Large Language Model for SciencearXiv:2211.09085