AINewsnow

Study examines how inference-time expert deactivation affects routing and weights in sparse models

This story is from 2026-11-06. It is preserved in the archive; the latest stories are on the live feed.

A new arXiv paper analyzes expert deactivation at inference time, showing it changes both token routing and expert weights, and that performance effects depend on how these changes are implemented.

Read the full story at arXiv cs.CL ↗

Timeline · 1 report

  1. 2026-11-06 05:00 · arXiv cs.CL
    Matching the Training-Time Weight Rule in Inference-Time Expert Deactivation

More stories

  1. OpenAI to keep expanding AI data centers despite election results favoring restrictions — DigiTimes Asia
  2. Claude Code ecosystem expands with memory, browser, and workspace tools — DEV Community — AI
  3. US AI Safety Center Staff Now Report to White House, Raising Independence Concerns — AP News
  4. Hydra-PSIE expands 'Coeziune OM–IA' DOI registry to 38 published works — DEV Community — AI
  5. SemiAnalysis details rapid buildout of AI data center region in Inner Mongolia — Techmeme
  6. AI Coding Agents Shift From Prompt Files to Deterministic Scripts and Sandboxed Execution — DEV Community — AI
  7. Tencent AI chief warns AI will soon master biology, raising safety risks — SCMP China Tech
  8. Dev News Digest: OpenSSH advisory, Claude Code email leak, GitHub Actions lockfiles GA — DEV Community — AI

Get the daily brief of stories like this at 6:30 every morning →