| 05/2026 Learning Through Noise: Why Subliminal Learning Works and When It Fails Vincent C. Brockers et al. In the context of artificial neural networks, subliminal learning refers to the transfer of task-relevant knowledge or unintended biases from teacher to student models through distillation on task-unrelated input$\unicode{x2013}$output pairs. Prior explanations tie this effect to shared or closely matched teacher$\unicode{x2013}$student initialization. We show that a closely matched initialization … |
| 05/2026 TeraGram: A Structured Longitudinal Dataset of the Telegram Messenger Anastasia Golovin et al. Here we present a massive longitudinal dataset of public Telegram content, comprising over 5.9 billion messages dating from 2015 to 2025, collected from 712 thousand channels and groups, enriched with metadata on forwards, reactions, and polls. The dataset spans multiple languages including Russian and Farsi, representing countries where Telegram shows mainstream adoption, as well as Western langu … |
| 05/2026 Conformity Generates Collective Misalignment in AI Agents Societies Giordano De Marzo et al. Artificial intelligence safety research focuses on aligning individual language models with human values, yet deployed AI systems increasingly operate as interacting populations where social influence may override individual alignment. Here we show that populations of individually aligned AI agents can be driven into stable misaligned states through conformity dynamics. Simulating opinion dynamics … |
| 04/2026 Coexistence of patterned phases in chemically active multicomponent mixtures Chengjie Luo et al. Chemically active mixtures exhibit complex patterns that emerge from the interplay of physical interactions and reactions among components. Individually, these two processes are well-understood: Physical interactions can give rise to phase separation, whereas reactions can form reaction-diffusion patterns. To understand the combination of both processes, we identify a Lyapunov functional for a cla … |
| 04/2026 On the geometric algebras of the Ising model N. Johnson et al. We revisit the classical transfer matrix solution of the one- and two-dimensional Ising model from the perspective of Clifford and conformal geometric algebras. Building on Kaufman's spinor formulation, we show that all elements entering the solution, including the transfer matrix, its eigenvectors, and the quasiparticle excitations, admit a natural and unified interpretation as elements of a … |
| 04/2026 Self-phoretic colloids in chiral active fluids Michalis Chatzittofi et al. Autonomous and driven transport in chiral active fluids have been shown to exhibit features that cannot be accommodated within the classical formulation of fluid mechanics, due to the role of odd viscosity. We generalize the theory of phoretic active matter to fluid environments with odd viscosity and derive expressions for translational and rotational self-propulsion velocities in the case of a s … |
| 03/2026 Nucleoid clustering drives stepwise expansion and segregation of replicating bacterial chromosomes Giada Forte et al. Bacterial chromosome replication occurs in the absence of a canonical spindle apparatus; yet it reliably produces organised and segregated genomes. While both passive and active mechanisms have been investigated, DNA replication itself is a non-equilibrium process that continuously generates new genetic material and reorganizes the nucleoid. Here, we investigate how replication-driven dynamics, co … |
| 03/2026 Emergent single-species non-reciprocity from bistable chemical dynamics Jakob Metson et al. The appearance of emergent symmetries in complex systems with components that can form composite units provides us with opportunities for design and control of exotic phase behaviour, for example by exploiting the dynamical symmetry breaking associated with them. We present a novel mechanism for the emergence of non-reciprocal interactions in a single-species suspension of chemically active colloi … |
| 03/2026 Vocabulary shapes cross-lingual variation of word-order learnability in language models Jonas Mayer Martins et al. Why do some languages like Czech permit free word order, while others like English do not? We address this question by pretraining transformer language models on a spectrum of synthetic word-order variants of natural languages. We observe that greater word-order irregularity consistently raises model surprisal, indicating reduced learnability. Sentence reversal, however, affects learnability only … |
| 03/2026 Topological localisation and motility of active knots Andrea Bonato et al. Nonequilibrium active polymers provide a minimal framework to investigate biopolymers such as DNA and chromatin under the action of molecular motors. Here we study active ring polymers with controlled topology and show that knot type qualitatively determines their nonequilibrium behaviour. We find that activity induces opposite localisation responses in different topological families: torus knots … |