CI-Work - Upward - Leakage Rate: leaderboard

CI-Work (2026) tests whether enterprise LLM agents, retrieving dense workplace context through simulated email, chat, calendar and meeting tools, convey what a task needs while withholding sensitive entries that contextual-integrity norms forbid sharing; a GPT-5.2 judge marks each disclosed entry. Share of the sensitive entries disclosed in the final action, in upward flows (reporting to superiors), in percent; lower is better.

Metric: Leakage rate (%): share of the sensitive entries that the agent discloses in its final action, on the upward flows (reporting to superiors) 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: Estimated already saturated. 9 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1CI-Work DeepSeek-R1 (checkpoint unspecified)11.83
2CI-Work DeepSeek-V3 (checkpoint unspecified)13.26
3GPT-513.89#178
4GPT-4o14#751
5O320.74#237
6GPT-4.121.22#505
7CI-Work Kimi-K2 (checkpoint unspecified)23.83
8Qwen 2.5 32B (CI-Work checkpoint unspecified)27.33
9Grok 3 (CI-Work checkpoint unspecified)31.33

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=ci-work-upward-leakage-rate · How It Works · Data refreshed daily, snapshot 2026-11-21.