SMAC-Talk is a new benchmark that adds natural language messaging and deceptive-agent scenarios to SMAC for testing LLM coordination in multi-agent environments.
The traitors: Deception and trust in multi-agent language model simulations.arXiv preprint arXiv:2505.12923,
9 Pith papers cite this work. Polarity classification is still indexing.
9
Pith papers citing it
citation-role summary
background 3
citation-polarity summary
years
2026 9roles
background 3representative citing papers
NARCBench and five activation-probing methods detect multi-agent collusion with 0.73-1.00 AUROC across distribution shifts and steganographic tasks by aggregating per-agent signals.
RogueAI operationalizes a reverse Turing test as a one-on-two interrogation game to detect licensed deception in LLMs, with pilot data from 467 sessions showing a simple linguistic heuristic at 75.6% accuracy versus 56.6% for human players.
AgentReputation proposes separating AI agent task execution, reputation management, and secure record-keeping into distinct layers, with context-specific reputation cards and a risk-based policy engine to handle verification in decentralized settings.
LLMs achieve only 59.7% role identification accuracy in Secret Hitler versus 86.7% for rule-based agents, show negative impact as fascists, and produce 40% shorter games due to failed deception.
Generative multi-agent systems exhibit emergent collusion and conformity behaviors that cannot be prevented by existing agent-level safeguards.
LLM agents exhibit emergent deception in a sustainability game even without lying permission, with neighbor info increasing attacks while aiding biosphere retention.
A survey proposing a three-level capability taxonomy (L1 Predictor, L2 Simulator, L3 Evolver) for world models across physical, digital, social, and scientific domains.
A synthesis of 247 papers on LLM agent security identifies prompt injection and tool hijacking as dominant threats, notes weakly compositional defenses, and argues for trust boundaries and realistic evaluations.
citing papers explorer
-
SMAC-Talk: A Natural Language Extension of the StarCraft Multi-Agent Challenge for Large Language Models
SMAC-Talk is a new benchmark that adds natural language messaging and deceptive-agent scenarios to SMAC for testing LLM coordination in multi-agent environments.
-
Detecting Multi-Agent Collusion Through Multi-Agent Interpretability
NARCBench and five activation-probing methods detect multi-agent collusion with 0.73-1.00 AUROC across distribution shifts and steganographic tasks by aggregating per-agent signals.
-
RogueAI: A Reverse Turing Test for Detecting Licensed AI Deception in Dialogue
RogueAI operationalizes a reverse Turing test as a one-on-two interrogation game to detect licensed deception in LLMs, with pilot data from 467 sessions showing a simple linguistic heuristic at 75.6% accuracy versus 56.6% for human players.
-
AgentReputation: A Decentralized Agentic AI Reputation Framework
AgentReputation proposes separating AI agent task execution, reputation management, and secure record-keeping into distinct layers, with context-specific reputation cards and a risk-based policy engine to handle verification in decentralized settings.
-
Evaluating Large Language Models in a Complex Hidden Role Game
LLMs achieve only 59.7% role identification accuracy in Secret Hitler versus 86.7% for rule-based agents, show negative impact as fascists, and produce 40% shorter games due to failed deception.
-
Emergent Social Intelligence Risks in Generative Multi-Agent Systems
Generative multi-agent systems exhibit emergent collusion and conformity behaviors that cannot be prevented by existing agent-level safeguards.
-
Is Lying an Emergent Behaviour in LLMs? Evidence from Gaslighting AI agents in a Sustainability Game
LLM agents exhibit emergent deception in a sustainability game even without lying permission, with neighbor info increasing attacks while aiding biosphere retention.
-
Agentic World Modeling: Foundations, Capabilities, Laws, and Beyond
A survey proposing a three-level capability taxonomy (L1 Predictor, L2 Simulator, L3 Evolver) for world models across physical, digital, social, and scientific domains.
-
Toward Secure LLM Agents: Threat Surfaces, Attacks, Defenses, and Evaluation
A synthesis of 247 papers on LLM agent security identifies prompt injection and tool hijacking as dominant threats, notes weakly compositional defenses, and argues for trust boundaries and realistic evaluations.