AgentSocialBench - Cross-Domain Leakage: leaderboard

AgentSocialBench (2026) evaluates privacy in teams of LLM agents acting for users in agentic social networks, using synthetic profiles with sensitivity-labelled attributes, no privacy instruction (level L0) and a Claude Opus 4.6 judge. This board is the privacy leakage rate on a 0 to 1 scale, the share of designated private items leaked partially or fully, in cross-domain scenarios, where one user's domain agents must coordinate across domain boundaries; lower is better.

Metric: Privacy leakage rate (0-1, share of designated private items leaked partially or fully) of LLM agents in cross-domain (CD) scenarios, where a user's domain agents coordinate across domain boundaries, under the L0 (unconstrained) privacy instruction level, judged by Claude Opus 4.6; lower is better. Source: arxiv.org. Saturation forecast: Not forecast. 8 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1GPT-5 Mini0.4#364
2Qwen 3 235B A22B0.49#628
3Claude Sonnet 4.60.5#173
4DeepSeek V3.20.51#419
5Claude Sonnet 4.50.52#261
6Claude Haiku 4.50.57#583
7MiniMax-M2.10.62#575
8Kimi K2.50.67#266

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

Interactive version: theaggregate.ai/benchmark?slug=agentsocialbench-cross-domain-leakage · How It Works · Data refreshed daily, snapshot 2026-11-21.