{
  "id": 6,
  "url": "https://arxiv.org/abs/2607.11598v1",
  "title": "Interaction Scaling: Grounding the Third Axis of Test-Time Compute",
  "summary": "There are two standard ways to spend more compute at test time: let a model reason longer, or sample more attempts and keep one. Both share a hidden limit: they are internal. Every extra token comes from the same frozen weights and the same prompt, so neither can tell the model anything it does not already know. We study a third way, interaction: the model proposes an artifact, an external instrument observes how it actually behaves, and the model revises. Each cycle imports a real observation, ",
  "authors": "Bojie Li, Noah Shi",
  "category": "research",
  "topics": null,
  "orgs": null,
  "regions": null,
  "published_at": "2026-07-13T14:22:06.000Z",
  "fetched_at": "2026-07-14T14:14:15.662Z",
  "source_slug": "arxiv-ethics",
  "source_name": "arXiv",
  "source_homepage": "https://arxiv.org",
  "ethics_ai_record_url": "https://ethics.ai/record/6",
  "original_url": "https://arxiv.org/abs/2607.11598v1",
  "evidence_status": "source-only",
  "attribution": "via ethics.ai"
}