CaM-Wolf: Causal-Aware Multimodal Agents for Social Deduction Games
Quick Answer
CaM-Wolf is the first social deduction game agent that integrates multimodal perception and causal reasoning, enhancing gameplay and human-AI interaction.
Quick Take
It processes video inputs and utilizes a causal-aware Reasoner trained via reinforcement learning, achieving superior performance in nuanced social dynamics. This advancement marks a significant step towards creating more human-like AI agents.
Key Points
- CaM-Wolf integrates multimodal perception for enhanced social deduction gameplay.
- It employs a causal-aware Reasoner trained through reinforcement learning.
- The agent processes video inputs to understand player behaviors and roles.
- User studies indicate improved human-AI interaction quality.
- This work advances the development of human-like AI agents.
DeepSignal Analysis
What happened
CaM-Wolf is introduced as a novel agent designed for social deduction games, integrating multimodal perception and causal reasoning. It processes video inputs and uses a causal-aware Reasoner trained through reinforcement learning, which enhances its gameplay capabilities. This development aims to improve human-AI interactions in complex social scenarios.
Key evidence
- CaM-Wolf is the first agent for social deduction games that combines multimodal perception and generation, addressing the limitations of text-based approaches.
- The agent processes video inputs from players and employs a causal-aware Reasoner, which establishes logical connections between observable behaviors and hidden roles.
- Experiments and user studies indicate that CaM-Wolf outperforms previous agents in gameplay and enhances the quality of interactions with human players.
Why it matters
The integration of multimodal perception and causal reasoning in CaM-Wolf represents a significant advancement in AI capabilities, particularly in social deduction games that require complex social skills. By moving beyond text-based interactions, this agent could facilitate more natural and engaging interactions between humans and AI. This progress may pave the way for future AI developments that can better understand and participate in nuanced social dynamics.
Paper Resources
Source Excerpt
Social deduction games (SDGs) such as Werewolf have become challenging testbeds for AI agents. These games require complex social skills such as reasoning, deception, and collaboration. While recent advances in (LLMs) have driven significant progress in SDG agents, current approaches are predominantly text-based, overlooking the multimodal nature that is fundamental to human social interaction. To bridge this gap, we introduce CaM-Wolf, the first SDG agent that integrates m
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.AI
See more →HOBA: Hierarchical On-Policy Bidding Agents for Adaptive Online Advertising
HOBA (Hierarchical On-policy Bidding Agents) is a novel hierarchical reinforcement learning framework that enhances online advertising bidding systems by improving adaptability and reducing hyperparameter tuning costs. It utilizes a for hyperparameter inference, a SARSA agent for expert model selection, and a dynamic expert pool for bid execution, achieving a +3.6% increase in target cost during large-scale deployment and outperforming state-of-the-art baselines on AuctionNet.