Emergent properties have been widely adopted as a term to describe behavior not present in smaller models but observed in larger models (Wei et al., 2022a). Recent work suggests that the trade-off incurred by quantization is also an emergent property, with sharp drops in performance in models over 6B parameters. In this work, we ask Against a backdrop of increased research focus on why certain emergent properties surface at scale, this work provides a useful counter-example. We posit that it is possible to optimize for a quantization friendly training recipe that suppresses large activation magnitude outliers. Here, we find that outlier dimensions are not an inherent product of scale, but rather sensitive to the optimization conditions present during pre-training. This both opens up directions for more efficient quantization, and poses the question of whether other emergent properties are inherent or can be altered and conditioned by optimization and architecture design choices.

Foundation models trained at scale exhibit remarkable emergent behaviors, learning new capabilities beyond their initial training objectives. We find such emergent behaviors in biological vision models via large-scale contrastive vision-language training. To achieve this, we first curate TreeOfLife-200M, comprising 214 million images of living organisms, the largest and most diverse biological organism image dataset to date. We then train BioCLIP 2 on TreeOfLife-200M to distinguish different species. Despite the narrow training objective, BioCLIP 2 yields extraordinary accuracy when applied to various biological visual tasks such as habitat classification and trait prediction. We identify emergent properties in the learned embedding space of BioCLIP 2. At the inter-species level, the embedding distribution of different species aligns closely with functional and ecological meanings (e.g., beak sizes and habitats). At the intra-species level, instead of being diminished, the intra-species variations (e.g., life stages and sexes) are preserved and better separated in subspaces orthogonal to inter-species distinctions. We provide formal proof and analyses to explain why hierarchical supervision and contrastive objectives encourage these emergent properties. Crucially, our results reveal that these properties become increasingly significant with larger-scale training data, leading to a biologically meaningful embedding space.

The performance of Large Language Models (LLMs) often degrades when crucial information is in the middle of a long context, a "lost-in-the-middle" phenomenon that mirrors the primacy and recency effects in human memory. We propose that this behavior is not simply a flaw indicative of information loss but an adaptation to different information retrieval demands during pre-training: some tasks require uniform recall across the entire input (a long-term memory demand), while others prioritize the most recent information (a short-term memory demand). Consistent with this view, we show that this U-shaped performance curve emerges when LLMs (GPT-2 and Llama variants) are trained from scratch on two simple human memory paradigms simulating long-term and short-term memory demands. Our analysis reveals that while the recency effect directly aligns with short-term memory demand in the training data, the primacy effect is induced by the uniform long-term memory demand and is additionally influenced by the model's autoregressive properties and the formation of attention sinks. Our main findings from simple human memory paradigms also generalize to a sequence completion task, which more closely resembles the next-token prediction process in LLM pre-training. Together, our findings reveal how information retrieval demands, model architecture, and structural attention dynamics during model training can jointly produce positional bias observed in LLMs.

Research evaluation is a common and important task for academics and managers, and it is often supported by citation - based indicators (Hicks et al., 2015; Moed, 2005; Mukherjee, 2022). With the increasingly widespread use of Artificial Intelligence (AI) in research ( Mohammadi et al., 2025), it is important to check whether it can save expert time through support of the research evaluation task. ChatGPT research quality score estimates for journal articles are recent alternative s to citations as quantitative indicator s to support evaluations ( Kousha & Thelwall, 2025) . Their value lies in their positive correlation with expert judgement in all or nearly all fields, and at a slightly higher rate than for citation - based indicators ( Thelwall, 2025abc). Despite some systematic biases or disparities ( Thelwall & Kurt, 2025), t his property means that they are helpful when expert judgement fails, such as fo r areas outside of the assessor's expertise, as a cross - check for bias, and for evaluations where assessment expertise is unavailable or too expensive for the value of the task (Thelwall, 2025d) . Whilst a positive correlation with expert judgement has been established for three of the largest Large Language Models (LLMs) in 2025, ChatGPT 4o, ChatGPT 4o - mini, and Google Gemini Flash 1.5 ( Thelwall, 2025ac), these are all cloud - based services and may be too expensive or not private enough for some research evaluation purposes ( Nowak et al., 2025) . Moreover, cloud - based services can be withdrawn, updated, or made more costly, so research evaluation procedures may not be able to rely on them. Thus, there is a need to test whether any smaller "open weights" LLMs ( Sowe et al., 2024) that can be downloaded and used offline have a capability to estimate research quality.

The remarkable success of Large Language Models (LLMs) in generative tasks has raised fundamental questions about the nature of their acquired capabilities, which often appear to emerge unexpectedly without explicit training. This paper examines the emergent properties of Deep Neural Networks (DNNs) through both theoretical analysis and empirical observation, addressing the epistemological challenge of "creation without understanding" that characterises contemporary AI development. We explore how the neural approach's reliance on nonlinear, stochastic processes fundamentally differs from symbolic computational paradigms, creating systems whose macro-level behaviours cannot be analytically derived from micro-level neuron activities. Through analysis of scaling laws, grokking phenomena, and phase transitions in model capabilities, I demonstrate that emergent abilities arise from the complex dynamics of highly sensitive nonlinear systems rather than simply from parameter scaling alone. My investigation reveals that current debates over metrics, pre-training loss thresholds, and in-context learning miss the fundamental ontological nature of emergence in DNNs. I argue that these systems exhibit genuine emergent properties analogous to those found in other complex natural phenomena, where systemic capabilities emerge from cooperative interactions among simple components without being reducible to their individual behaviours. The paper concludes that understanding LLM capabilities requires recognising DNNs as a new domain of complex dynamical systems governed by universal principles of emergence, similar to those operating in physics, chemistry, and biology. This perspective shifts the focus from purely phenomenological definitions of emergence to understanding the internal dynamic transformations that enable these systems to acquire capabilities that transcend their individual components.

Emergence and swarms are widely discussed topics, yet no consensus exists on their formal definitions. This lack of agreement makes it difficult not only for new researchers to grasp these concepts, but also for experts who may use the same terms to mean different things. Many attempts have been made to objectively define 'swarm' or 'emergence,' with recent work highlighting the role of the external observer. Still, several researchers argue that once an observer's vantage point (e.g., scope, resolution, context) is established, the terms can be made objective or measured quantitatively. In this note, we propose a framework to discuss these ideas rigorously by separating externally observable states from latent, unobservable ones. This allows us to compare and contrast existing definitions of swarms and emergence on common ground. We argue that these concepts are ultimately subjective-shaped less by the system itself than by the perception and tacit knowledge of the observer. Specifically, we suggest that a 'swarm' is not defined by its group behavior alone, but by the process generating that behavior. Our broader goal is to support the design and deployment of robotic swarm systems, highlighting the critical distinction between multi-robot systems and true swarms.

In music information retrieval (MIR), contrastive self-supervised learning for general-purpose representation models is effective for global tasks such as automatic tagging. However, for local tasks such as chord estimation, it is widely assumed that contrastively trained general-purpose self-supervised models are inadequate and that more sophisticated SSL is necessary; e.g., masked modeling. Our paper challenges this assumption by revealing the potential of contrastive SSL paired with a transformer in local MIR tasks. We consider a lightweight vision transformer with one-dimensional patches in the time--frequency domain (ViT-1D) and train it with simple contrastive SSL through normalized temperature-scaled cross-entropy loss (NT-Xent). Although NT-Xent operates only over the class token, we observe that, potentially thanks to weight sharing, informative musical properties emerge in ViT-1D's sequence tokens. On global tasks, the temporal average of class and sequence tokens offers a performance increase compared to the class token alone, showing useful properties in the sequence tokens. On local tasks, sequence tokens perform unexpectedly well, despite not being specifically trained for. Furthermore, high-level musical features such as onsets emerge from layer-wise attention maps and self-similarity matrices show different layers capture different musical dimensions. Our paper does not focus on improving performance but advances the musical interpretation of transformers and sheds light on some overlooked abilities of contrastive SSL paired with transformers for sequence modeling in MIR.

Large Language Models (LLMs) are deep neural networks that, through training on huge amounts of text, learn to accurately predict the next word (or token) in a text. It has been surprising to many that next-token prediction has lead to impressive abilities, such as learning of syntax, code generation, writing in any style, and factual recall. It has been claimed in the LLM literature that, as the number of network parameters and amount of training data is scaled up, certain capabilities arise suddenly and unexpectedly, a phenomenon that these writers term "emergence". For example, Wei et al. [1] write, "we define emergent abilities of large language models as abilities that are not present in smaller-scale models but are present in large-scale models; thus they cannot be predicted by simply extrapolating the performance improvements on smaller-scale models." And in a recent review of emergent abilities in LLMs Berti et al. [2] survey around 100 papers the majority of which equate emergence with the discontinuous appearance of abilities with increasing data or model size.

Conceptual combination is a cognitive process that merges basic concepts, enabling the creation of complex expressions. During this process, the properties of combination (e.g., the whiteness of a peeled apple) can be inherited from basic concepts, newly emerge, or be canceled. However, previous studies have evaluated a limited set of properties and have not examined the generative process. To address this gap, we introduce the Conceptual Combination with Property Type dataset (CCPT), which consists of 12.3K annotated triplets of noun phrases, properties, and property types. Using CCPT, we establish three types of tasks to evaluate LLMs for conceptual combination thoroughly. Our key findings are threefold: (1) Our automatic metric grading property emergence and cancellation closely corresponds with human judgments. (2) LLMs, including OpenAI's o1, struggle to generate noun phrases which possess given emergent properties. (3) Our proposed method, inspired by cognitive psychology model that explains how relationships between concepts are formed, improves performances in all generative tasks. The dataset and experimental code are available at https://github.com/seokwon99/CCPT.git.

Despite the rapid pace at which deep networks are improving on standardized vision benchmarks, they are still outperformed by humans on real-world vision tasks. This paradoxical lack of generalization could be addressed by making deep networks more brain-like. Although several benchmarks have compared the ability of deep networks to predict brain responses to natural images, they do not capture subtle but important brain-like emergent properties. To resolve this issue, we report several well-known perceptual and neural emergent properties that can be tested on deep networks. To evaluate how various design factors impact brain-like properties, we systematically evaluated over 30 state-of-the-art networks with varying network architectures, training datasets and training regimes. Our main findings are as follows. First, network architecture had the strongest impact on brain-like properties compared to dataset and training regime variations. Second, networks varied widely in their alignment to the brain with no single network outperforming all others. Taken together, our results complement existing benchmarks by revealing brain-like properties that are either emergent or lacking in state-of-the-art deep networks.