{
  "id": 227,
  "url": "https://arxiv.org/abs/2607.04432v1",
  "title": "Covert Trait Propagation Is Representation Alignment: Mechanistic Evidence from Hidden-Channel Distillation",
  "summary": "A student model trained on pure uniform noise can still inherit its teacher's digit-classification ability, provided the two share initialization. Previous work proves this transfer is guaranteed when the teacher's learning rate is small enough, but does not explain where in the network the channel lives or what sets its capacity. Working in an MLP distillation setting on MNIST, we show these channels are not purely informational: geometric alignment gates access to the information the channel c",
  "authors": "Kargi Chauhan, Aditya Shah",
  "category": "research",
  "topics": "safety-alignment,children-education",
  "orgs": null,
  "regions": null,
  "published_at": "2026-07-05T17:52:09.000Z",
  "fetched_at": "2026-07-14T14:14:24.246Z",
  "source_slug": "arxiv-ethics",
  "source_name": "arXiv",
  "source_homepage": "https://arxiv.org",
  "ethics_ai_record_url": "https://ethics.ai/record/227",
  "original_url": "https://arxiv.org/abs/2607.04432v1",
  "evidence_status": "source-only",
  "attribution": "via ethics.ai"
}