{
  "id": 17925884,
  "title": "Sonar-MoE: In-Backward Expert Updates for Memory-Efficient Mixture-of-Experts Training",
  "url": "https://urgent.news/2026/11/02/sonar-moe-in-backward-expert-updates-for-memory-efficient-mixture-of",
  "topic": "ai",
  "section": "AI",
  "published": "2026-11-02T18:30:00.000Z",
  "source": {
    "name": "arXiv cs.AI",
    "slug": "arxiv-cs-ai",
    "url": "https://arxiv.org/abs/2611.03577v1"
  },
  "original_language": "en",
  "account": null,
  "summary": "Most parameters of a Mixture-of-Experts (MoE) model belong to its experts, and standard training keeps every expert gradient in GPU memory until the optimizer step. This memory grows with the number and size of the experts, so it rises as MoE models scale. We introduce Sonar-MoE, which updates each expert tensor during the backward pass and frees its gradient right away. It brings…",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}