{
  "id": 4212,
  "url": "https://arxiv.org/abs/2605.17064v2",
  "title": "Towards Human-Level Book-Writing Capability",
  "summary": "Large language models are optimized for instruction following and agentic tasks remain poorly aligned with the requirements of high-quality creative writing. We show that a purpose-built creative writing model can outperform both GPT-5.5 and Claude Opus 4.8 on writing quality evaluation. Fiction frequently depends on behaviors that assistant-tuned models are explicitly trained to avoid, particularly deception, moral ambiguity, and unreliable narration. As a result, generated stories often appear",
  "authors": "Jan Zierstek, Matteo Batelic, Maya Medjad, Tim Schönenberger",
  "category": "research",
  "topics": "agents-autonomy",
  "orgs": "openai,anthropic",
  "regions": null,
  "published_at": "2026-05-16T16:10:41.000Z",
  "fetched_at": "2026-07-14T16:30:50.572Z",
  "source_slug": "arxiv-ethics",
  "source_name": "arXiv",
  "source_homepage": "https://arxiv.org",
  "ethics_ai_record_url": "https://ethics.ai/record/4212",
  "original_url": "https://arxiv.org/abs/2605.17064v2",
  "evidence_status": "source-only",
  "attribution": "via ethics.ai"
}