{
  "id": 4388,
  "url": "https://arxiv.org/abs/2605.13537v1",
  "title": "Temper and Tilt Lead to SLOP: Reward Hacking Mitigation with Inference-Time Alignment",
  "summary": "Inference-time alignment techniques offer a lightweight alternative or complement to costly reinforcement learning, while enabling continual adaptation as alignment objectives and reward targets evolve. Existing theoretical analyses justify these methods as approximations to sampling from distributions optimally tilted toward a given reward model. We extend these techniques by introducing reference-model temperature adjustment, which leads to further generalization of inference-time alignment to",
  "authors": "Ye Wang, Jing Liu, Toshiaki Koike-Akino",
  "category": "research",
  "topics": "safety-alignment",
  "orgs": null,
  "regions": null,
  "published_at": "2026-05-13T13:47:06.000Z",
  "fetched_at": "2026-07-14T16:30:59.235Z",
  "source_slug": "arxiv-ethics",
  "source_name": "arXiv",
  "source_homepage": "https://arxiv.org",
  "ethics_ai_record_url": "https://ethics.ai/record/4388",
  "original_url": "https://arxiv.org/abs/2605.13537v1",
  "evidence_status": "source-only",
  "attribution": "via ethics.ai"
}