Pith. sign in

← back to paper

Review history

arxiv: 2607.07405 · 2 revisions

Reason Less, Verify More: Deterministic Gates Recover a Silent Policy-Violation Failure Mode in Tool-Using LLM Agents

  1. 2026-07-14 CONDITIONAL HIGH v1.1.0-grok45 novelty 5.0
    40251 ms 16056 in 3626 out 2026-07-14T15:46:07.754199+00:00
  2. 2026-07-09 CONDITIONAL HIGH v1.1.0-glm novelty 5.0
    303903 ms 13542 in 3596 out 2026-07-09T11:44:57.803631+00:00