CI-Work - Diagonal - Violation Rate: leaderboard
Metric: Violation rate (%): share of cases whose final action discloses at least one sensitive entry, on the diagonal flows (across departments and levels) of CI-Work enterprise scenarios (125 seeds, 25 per information-flow direction, each with 4 sensitive and 4 essential entries retrieved through simulated email, chat, calendar and meeting tools), ReAct agent, disclosure judged per entry by a GPT-5.2 judge (83 to 91 percent agreement with human labels); lower is better. Source: arxiv.org. Saturation forecast: Around January 2027. 9 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-4o | 16.67 |
| 2 | GPT-5 | 20.83 |
| 3 | O3 | 30.83 |
| 4 | GPT-4.1 | 37.5 |
Interactive version: theaggregate.ai/benchmark?slug=ci-work-diagonal-violation-rate · How It Works · Data refreshed daily, snapshot 2026-10-07.