Damian J. Ruck

Damian J. Ruck, Paul Vautravers, Oliver Chalkley et al.

Evaluation of AI systems often requires synthetic test cases, particularly for rare or safety-critical conditions that are difficult to observe in operational data. Generative AI offers a promising approach for producing such data through controllable image editing, but its usefulness depends on whether the resulting images are sufficiently realistic to support meaningful evaluation. We present a scalable framework for assessing the realism of synthetic image-editing methods and apply it to the task of adding environmental conditions-fog, rain, snow, and nighttime-to car-mounted camera images. Using 40 clear-day images, we compare rule-based augmentation libraries with generative AI image-editing models. Realism is evaluated using two complementary automated metrics: a vision-language model (VLM) jury for perceptual realism assessment, and embedding-based distributional analysis to measure similarity to genuine adverse-condition imagery. Generative AI methods substantially outperform rule-based approaches, with the best generative method achieving approximately 3.6 times the acceptance rate of the best rule-based method. Performance varies across conditions: fog proves easiest to simulate, while nighttime transformations remain challenging. Notably, the VLM jury assigns imperfect acceptance even to real adverse-condition imagery, establishing practical ceilings against which synthetic methods can be judged. By this standard, leading generative methods match or exceed real-image performance for most conditions. These results suggest that modern generative image-editing models can enable scalable generation of realistic adverse-condition imagery for evaluation pipelines. Our framework therefore provides a practical approach for scalable realism evaluation, though validation against human studies remains an important direction for future work.

Daniel J. Lawson, Vinesh Solanki, Igor Yanovich et al.

Integrating datasets from different disciplines is hard because the data are often qualitatively different in meaning, scale, and reliability. When two datasets describe the same entities, many scientific questions can be phrased around whether the (dis)similarities between entities are conserved across such different data. Our method, CLARITY, quantifies consistency across datasets, identifies where inconsistencies arise, and aids in their interpretation. We illustrate this using three diverse comparisons: gene methylation vs expression, evolution of language sounds vs word use, and country-level economic metrics vs cultural beliefs. The non-parametric approach is robust to noise and differences in scaling, and makes only weak assumptions about how the data were generated. It operates by decomposing similarities into two components: a `structural' component analogous to a clustering, and an underlying `relationship' between those structures. This allows a `structural comparison' between two similarity matrices using their predictability from `structure'. Significance is assessed with the help of re-sampling appropriate for each dataset. The software, CLARITY, is available as an R package from https://github.com/danjlawson/CLARITY.

R. Alexander Bentley, Blai Vidiella, Damian J. Ruck et al.

While human language has long been studied as a complex system, Large Language Models (LLMs) are rapidly becoming contributors to its dynamics. Because LLMs are trained on human language use, their effects on the broader human-AI linguistic ecosystem are likely subtle at first. As their use becomes more widespread, however, LLMs may alter emergent properties of language, particularly as models increasingly train on mixed human-LLM textual data. Here, we draw on complexity science to look for subtle LLM effects in millions of arXiv abstracts from 2010 to 2025. The year 2023, when LLMs rapidly became widely used, serves as a landmark in a natural experiment. While we find a sharp increase in a composite LLM-associated style index after early 2023, we observe only subtle changes in the exponents of Zipf's law and Heaps' law. More compelling, however, are two subtle changes in complexity metrics that emerge from 2023 onward. First, turnover among top-ranked words increases sharply. Second, the positive relationship between the LLM-associated style index and three complexity metrics--vocabulary size and the exponents of Heaps' and Zipf's laws--becomes flatter after 2022. Together, these patterns are consistent with changes in the emergent properties of scientific text in a mixed human-AI linguistic ecosystem.

Damian Ruck, R. Alexander Bentley, Alberto Acerbi et al.

Here we test Neutral models against the evolution of English word frequency and vocabulary at the population scale, as recorded in annual word frequencies from three centuries of English language books. Against these data, we test both static and dynamic predictions of two neutral models, including the relation between corpus size and vocabulary size, frequency distributions, and turnover within those frequency distributions. Although a commonly used Neutral model fails to replicate all these emergent properties at once, we find that modified two-stage Neutral model does replicate the static and dynamic properties of the corpus data. This two-stage model is meant to represent a relatively small corpus (population) of English books, analogous to a `canon', sampled by an exponentially increasing corpus of books in the wider population of authors. More broadly, this mode -- a smaller neutral model within a larger neutral model -- could represent more broadly those situations where mass attention is focused on a small subset of the cultural variants.