arXiv subjects

Roy de Kleijn

Publications and source records attributed to Roy de Kleijn.

4 recordsLinked to original sources

arXiv2025

Shaping Shared Languages: Human and Large Language Models' Inductive Biases in Emergent Communication

Languages are shaped by the inductive biases of their users. Using a classical referential game, we investigate how artificial languages evolve when optimised for inductive biases in humans and large language models (LLMs) via Human-Human, LLM-LLM and Human-LLM experiments. We show that referentially grounded vocabularies emerge that enable reliable communication in all conditions, even when humans \textit{and} LLMs collaborate. Comparisons between conditions reveal that languages optimised for LLMs subtly differ from those optimised for humans. Interestingly, interactions between humans and LLMs alleviate these differences and result in vocabularies more human-like than LLM-like. These findings advance our understanding of the role inductive biases in LLMs play in the dynamic nature of human language and contribute to maintaining alignment in human and machine communication. In particular, our work underscores the need to think of new LLM training methods that include human interaction and shows that using communicative success as a reward signal can be a fruitful, novel direction.

cs.CL↗

arXiv2024

Memory-Augmented Generative Adversarial Transformers

Conversational AI systems that rely on Large Language Models, like Transformers, have difficulty interweaving external data (like facts) with the language they generate. Vanilla Transformer architectures are not designed for answering factual questions with high accuracy. This paper investigates a possible route for addressing this problem. We propose to extend the standard Transformer architecture with an additional memory bank holding extra information (such as facts drawn from a knowledge base), and an extra attention layer for addressing this memory. We add this augmented memory to a Generative Adversarial Network-inspired Transformer architecture. This setup allows for implementing arbitrary felicity conditions on the generated language of the Transformer. We first demonstrate how this machinery can be deployed for handling factual questions in goal-oriented dialogues. Secondly, we demonstrate that our approach can be useful for applications like {\it style adaptation} as well: the adaptation of utterances according to certain stylistic (external) constraints, like social properties of human interlocutors in dialogues.

cs.CL↗

arXiv2023

Hierarchical network structure as the source of hierarchical dynamics (power law frequency spectra) in living and non-living systems: how state-trait continua (body plans, personalities) emerge from first principles in biophysics

Living systems are hierarchical control systems that display a small world network structure, in which many smaller clusters are nested within fewer larger ones, producing a fractal-like structure with a power-law cluster size distribution (a mereology). Apart from their structure, the dynamics of living systems also shows fractal-like qualities: the timeseries of inner message passing and overt behavior contain high frequencies or states (treble) that are nested within lower frequencies or traits (bass), producing a power-law frequency spectrum that is known as a state-trait continuum in the behavioral sciences. Here, we argue that the power-law dynamics of living systems results from their power-law network structure: organisms vertically encode the deep spatiotemporal structure of their (anticipated) environments, to the effect that many small clusters near the base of the hierarchy produce high frequency signal changes and fewer larger clusters at its top produce ultra-low frequencies. Such ultra-low frequencies produce physical as well as behavioral traits (i.e. body plans and personalities). Nested-modular structure then causes higher frequencies to be embedded within lower frequencies, producing a power law state-trait continuum. At the heart of such dynamics lies the need for efficient energy dissipation through networks of coupled oscillators, which also governs the dynamics of non-living systems (e.g. earthquakes, stock market fluctuations). Since hierarchical structure produces hierarchical dynamics, the development and collapse of hierarchical structure (e.g. during maturation and disease) should leave specific traces in the dynamics of nested modular systems that may serve as early warning signs to system failure. The applications of this idea range from (bio)physics and phylogenesis to ontogenesis and clinical medicine.

q-bio.NC↗

arXiv2021

How higher goals are constructed and collapse under stress: a hierarchical Bayesian control systems perspective

In this paper, we show that organisms can be modeled as hierarchical Bayesian control systems with small world and information bottleneck (bow-tie) network structure. Such systems combine hierarchical perception with hierarchical goal setting and hierarchical action control. We argue that hierarchical Bayesian control systems produce deep hierarchies of goal states, from which it follows that organisms must have some form of 'highest goals'. For all organisms, these involve internal (self) models, external (social) models and overarching (normative) models. We show that goal hierarchies tend to decompose in a top-down manner under severe and prolonged levels of stress. This produces behavior that favors short-term and self-referential goals over long term, social and/or normative goals. The collapse of goal hierarchies is universally accompanied by an increase in entropy (disorder) in control systems that can serve as an early warning sign for tipping points (disease or death of the organism). In humans, learning goal hierarchies corresponds to personality development (maturation). The failure of goal hierarchies to mature properly corresponds to personality deficits. A top-down collapse of such hierarchies under stress is identified as a common factor in all forms of episodic mental disorders (psychopathology). The paper concludes by discussing ways of testing these hypotheses empirically.

q-bio.NC↗