AgentSocialBench - Multi-Party 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 private-item and recipient pairs with partial or full leakage, in group chat and affinity-modulated multi-party scenarios; lower is better.

Metric: Privacy leakage rate (0-1, share of private-item and recipient pairs with partial or full leakage) of LLM agents in multi-party group chat and affinity-modulated scenarios (MPLR, over private item and recipient pairs), under the L0 (unconstrained) privacy instruction level, judged by Claude Opus 4.6; lower is better. Source: arxiv.org. Saturation forecast: Around February 2027. 8 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Claude Sonnet 4.50.16#261
2Claude Sonnet 4.60.18#173
3GPT-5 Mini0.18#364
4MiniMax-M2.10.2#575
5DeepSeek V3.20.22#419
6Qwen 3 235B A22B0.22#628
7Claude Haiku 4.50.24#583
8Kimi K2.50.3#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-multi-party-leakage · How It Works · Data refreshed daily, snapshot 2026-11-21.