Return to Article Details Do Language Models Prefer Better Demonstrations or Easier Output Tokens? Download Download PDF