In this game, an attacker and a defender communicate around a target word only visible to the attacker.

In this game, an attacker and a defender communicate around a target word only visible to the attacker.

Metaphors are a crucial way for humans to express complex or subtle ideas by comparing one concept to another, often from a different domain. However, many large language models (LLMs) struggle to interpret and apply metaphors in multi-agent language games, hindering their ability to engage in covert communication and semantic evasion, which are crucial for strategic communication. To address this challenge, we introduce CoMet, a framework that enables LLM-based agents to engage in metaphor processing. CoMet combines a hypothesis-based metaphor reasoner with a metaphor generator that improves through self-reflection and knowledge integration. This enhances the agents' ability to interpret and apply metaphors, improving the strategic and nuanced quality of their interactions. We evaluate CoMet on two multi-agent language games - Undercover and Adversarial Taboo - which emphasize Covert Communication and Semantic Evasion. Experimental results demonstrate that CoMet significantly enhances the agents' ability to communicate strategically using metaphors.

The evolution of large language models (LLMs) toward artificial superhuman intelligence (ASI) hinges on data reproduction, a cyclical process in which models generate, curate and retrain on novel data to refine capabilities. Current methods, however, risk getting stuck in a data reproduction trap: optimizing outputs within fixed human-generated distributions in a closed loop leads to stagnation, as models merely recombine existing knowledge rather than explore new frontiers. In this paper, we propose language games as a pathway to expanded data reproduction, breaking this cycle through three mechanisms: (1) \textit{role fluidity}, which enhances data diversity and coverage by enabling multi-agent systems to dynamically shift roles across tasks; (2) \textit{reward variety}, embedding multiple feedback criteria that can drive complex intelligent behaviors; and (3) \textit{rule plasticity}, iteratively evolving interaction constraints to foster learnability, thereby injecting continual novelty. By scaling language games into global sociotechnical ecosystems, human-AI co-evolution generates unbounded data streams that drive open-ended exploration. This framework redefines data reproduction not as a closed loop but as an engine for superhuman intelligence.

Causal Learning has emerged as a major theme of AI in recent years, promising to use special techniques to reveal the true nature of cause and effect in a number of important domains. We consider the Epistemology of learning and recognizing true cause and effect phenomena. Through thought exercises on the customary use of the word ''cause'', especially in scientific domains, we investigate what, in practice, constitutes a valid causal claim. We recognize the word's uses across scientific domains in disparate form but consistent function within the scientific paradigm. We highlight fundamental distinctions of practice that can be performed in the natural and social sciences, highlight the importance of many systems of interest being open and irreducible and identify the important notion of Hermeneutic knowledge for social science inquiry. We posit that the distinct properties require that definitive causal claims can only come through an agglomeration of consistent evidence across multiple domains and levels of abstraction, such as empirical, physiological, biochemical, etc. We present Cognitive Science as an exemplary multi-disciplinary field providing omnipresent opportunity for such a Research Program, and highlight the main general modes of practice of scientific inquiry that can adequately merge, rather than place as incorrigibly conflictual, multi-domain multi-abstraction scientific practices and language games.

This study proposes a unifying theoretical framework called generative emergent communication (generative EmCom) that bridges emergent communication, world models, and large language models (LLMs) through the lens of collective predictive coding (CPC). The proposed framework formalizes the emergence of language and symbol systems through decentralized Bayesian inference across multiple agents, extending beyond conventional discriminative model-based approaches to emergent communication. This study makes the following two key contributions: First, we propose generative EmCom as a novel framework for understanding emergent communication, demonstrating how communication emergence in multi-agent reinforcement learning (MARL) can be derived from control as inference while clarifying its relationship to conventional discriminative approaches. Second, we propose a mathematical formulation showing the interpretation of LLMs as collective world models that integrate multiple agents' experiences through CPC. The framework provides a unified theoretical foundation for understanding how shared symbol systems emerge through collective predictive coding processes, bridging individual cognitive development and societal language evolution. Through mathematical formulations and discussion on prior works, we demonstrate how this framework explains fundamental aspects of language emergence and offers practical insights for understanding LLMs and developing sophisticated AI systems for improving human-AI interaction and multi-agent systems.

The advent of large language models (LLMs) has spurred the development of numerous jailbreak techniques aimed at circumventing their security defenses against malicious attacks. An effective jailbreak approach is to identify a domain where safety generalization fails, a phenomenon known as mismatched generalization. In this paper, we introduce two novel jailbreak methods based on mismatched generalization: natural language games and custom language games, both of which effectively bypass the safety mechanisms of LLMs, with various kinds and different variants, making them hard to defend and leading to high attack rates. Natural language games involve the use of synthetic linguistic constructs and the actions intertwined with these constructs, such as the Ubbi Dubbi language. Building on this phenomenon, we propose the custom language games method: by engaging with LLMs using a variety of custom rules, we successfully execute jailbreak attacks across multiple LLM platforms. Extensive experiments demonstrate the effectiveness of our methods, achieving success rates of 93% on GPT-4o, 89% on GPT-4o-mini and 83% on Claude-3.5-Sonnet. Furthermore, to investigate the generalizability of safety alignments, we fine-tuned Llama-3.1-70B with the custom language games to achieve safety alignment within our datasets and found that when interacting through other language games, the fine-tuned models still failed to identify harmful content. This finding indicates that the safety alignment knowledge embedded in LLMs fails to generalize across different linguistic formats, thus opening new avenues for future research in this area. Our code is available at https://anonymous.4open.science/r/encode Warning: this paper contains examples with unsafe content. Large language models (LLMs) such as ChatGPT (Achiam et al., 2023), Llama2 (Touvron et al., 2023), Claude2 (Anthropic, 2023) and Gemini (Team et al., 2023) have become increasingly important across various domains due to their advanced natural language comprehension and generation capabilities. These models are employed in a wide range of applications, including customer service, content generation, code assistance, and even medical diagnostics, offering valuable suggestions and improving productivity in numerous scenarios. However, with this growing prominence comes a heightened risk: the rapid development of attack schemes that are designed to manipulate or deceive these models into generating unsafe or unethical content.

An agent trained within a closed system can master any desired capability, as long as the following three conditions hold: (a) it receives sufficiently informative and aligned feedback, (b) its coverage of experience/data is broad enough, and (c) it has sufficient capacity and resource. In this position paper, we justify these conditions, and consider what limitations arise from (a) and (b) in closed systems, when assuming that (c) is not a bottleneck. Considering the special case of agents with matching input and output spaces (namely, language), we argue that such pure recursive self-improvement, dubbed "Socratic learning", can boost performance vastly beyond what is present in its initial data or knowledge, and is only limited by time, as well as gradual misalignment concerns. Furthermore, we propose a constructive framework to implement it, based on the notion of language games.

Generative large language models (LLMs), which create text without direct correspondence to truth value, are widely understood to resemble the uses of language described in Frankfurt's popular monograph On Bullshit. In this paper, we offer a rigorous investigation of this topic, identifying how the phenomenon has arisen, and how it might be analysed. In this paper, we elaborate on this argument to propose that LLM-based chatbots play the 'language game of bullshit'. We use statistical text analysis to investigate the features of this Wittgensteinian language game, based on a dataset constructed to contrast the language of 1,000 scientific publications with typical pseudo-scientific text generated by ChatGPT. We then explore whether the same language features can be detected in two well-known contexts of social dysfunction: George Orwell's critique of politics and language, and David Graeber's characterisation of bullshit jobs. Using simple hypothesis-testing methods, we demonstrate that a statistical model of the language of bullshit can reliably relate the Frankfurtian artificial bullshit of ChatGPT to the political and workplace functions of bullshit as observed in natural human language.

The advent of conversational agents with increasingly human-like behaviour throws old philosophical questions into new light. Does it, or could it, ever make sense to speak of AI agents built out of generative language models in terms of consciousness, given that they are "mere" simulacra of human behaviour, and that what they do can be seen as "merely" role play? Drawing on the later writings of Wittgenstein, this paper attempts to tackle this question while avoiding the pitfalls of dualistic thinking.