arXiv cs.CL· Marco Molinari, Gabriele Bettineschi·· 11 小时前AI 评分52
多智能体串谋与 J-space：用 J-PACT 检测智能体串谋
Multi-Agent Collusion and the J-Space
AI 导读
论文研究 LLM 多智能体系统中的串谋问题，在 DeepSeek-V4-Flash（284B）上用可解释性实验发现，串谋智能体在回复前的 J-space 中 86% 会表征虚假陈述、隐瞒或悔意；在隐写式记牌任务中，接收方的 J-space 承载隐藏计数，因果干预可改变该计数。
正文 · 原文
Abstract:LLM agents are increasingly deployed in Multi-Agent Systems (MAS), and they introduce the risk of coordination among agents that might escape human oversight. The J-space has been used to uncover deception in single LLMs, but its use for uncovering and steering collusion between agents remains unexplored. We show, with interpretability experiments on DeepSeek-V4-Flash (284B) on the NARCBench collusion benchmark, that colluding agents' J-space holds the collusion itself: before replying, 86% of colluders represent misrepresentation, concealment or remorse. In a steganographic card-counting task, the receiver's J-space holds a hidden count, and causal interventions alter its count. We also introduce the Jacobian Pair Contrast (J-PACT), an unsupervised method for detecting collusion between agents. We evaluate J-PACT on four open-weight models: GPT-OSS-20B, Llama-3.3-70B-IT, Qwen3-32B, and DeepSeek-V4-Flash (284B). J-PACT retains 85% of the above-chance AUROC of supervised NARCBench probes, matches them on steganographic scenarios, and names the colluding pair up to its complement in 99% of detected runs.
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2611.02153 [cs.CL] |
| (or arXiv:2611.02153v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2611.02153 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Marco Molinari [view email]
[v1]
Sun, 1 Nov 2026 19:36:13 UTC (1,101 KB)
来源：arXiv cs.CL · arxiv.org