← back to paper

Review history

arxiv: 2605.03242 · 2 revisions

Enhancing Agent Safety Judgment: Controlled Benchmark Rewriting and Analogical Reasoning for Deceptive Out-of-Distribution Scenarios

1. 2026-05-07 UNVERDICTED LOW v0.9.0 novelty 7.0
   42798 ms 5517 in 1261 out 2026-05-07T16:53:39.504625+00:00
2. 2026-05-07 UNVERDICTED LOW v0.9.0 novelty 6.0
   21292 ms 5495 in 1010 out 2026-05-07T01:59:18.890360+00:00