@misc{indiciae5a35d206af4d,
  title = {Why Summaries Turn Neutral: Policy Attribution for Sentiment Drift in Reinforcement Learning from Human Feedback},
  author = {Mikhail Krasitskii and Alexander Gelbukh and Olga Kolesnikova and Grigori Sidorov},
  year = {2026},
  url = {https://arxiv.org/abs/2608.15530},
  note = {Source identifier: 2608.15530}
}