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