Zhang J. Do Language Models Prefer Better Demonstrations or Easier Output Tokens?. CAI. 2026;3(5):1-10. doi:10.70267/cai.v3n5.0110