pith. sign in

← back to paper

Review history

arxiv: 2606.22686 · 2 revisions

The Geometry of Refusal: Linear Instability in Safety-Aligned LLMs

  1. 2026-07-01 UNVERDICTED LOW v0.9.1-grok novelty 6.0
    32551 ms 5827 in 1347 out 2026-07-01T07:05:02.482569+00:00
  2. 2026-06-26 UNVERDICTED LOW v0.9.1-grok novelty 6.0
    19739 ms 5827 in 1087 out 2026-06-26T09:50:42.508767+00:00