@unpublished{22892, abstract = {Countless biological processes are fueled by energy-rich molecules like ATP and GTP that supply energy with extreme efficiency. However, designing similar energy-delivery schemes from the bottom up, essential for the development of powered nanostructures and other \emph{de novo} machinery, presents a significant challenge: how can an energy-rich structure be stable in solution yet still deliver this energy at precisely the right time? In this paper, we present a purely physical mechanism that solves this challenge, facilitating energy transfer akin to ATP hydrolysis, yet occurring between synthetic nanostructures without any biochemical interactions. This targeted energy delivery is achieved by exploiting a differentiable state-based model to balance the energy profiles that govern the structural transitions in the two nanostructures, creating a coupled relaxation pathway with minimal barriers that facilitates energy delivery. We verify the effectiveness and robustness of this mechanism through Langevin Dynamics simulations, demonstrating that a bath of the high-energy structures can systematically and repeatedly drive the target structure out of equilibrium, enabling it to perform tasks. As the mechanism operates only through explicit physical forces without any biochemistry or internal state variables, our results present generic and far-reaching design principles, setting the stage for the next generation of synthetic nanomachines.}, author = {Ehrmann, Andreas and Goodrich, Carl Peter}, booktitle = {arXiv}, title = {{Controlling energy delivery with bistable nanostructures}}, doi = {10.48550/arXiv.2506.14266}, year = {2026}, } @inproceedings{22894, abstract = {Large-scale deep learning models are known to memorize parts of the training set. In machine learning theory, memorization is often framed as interpolation or label fitting, and classical results show that this can be achieved when the number of parameters p in the model is larger than the number of training samples n. In this work, we consider memorization from the perspective of data reconstruction, demonstrating that this can be achieved when p is larger than dn, where d is the dimensionality of the data. More specifically, we show that, in the random features model, when p ≫ dn, the subspace spanned by the training samples in feature space gives sufficient information to identify the individual samples in input space. Our analysis suggests an optimization method to reconstruct the dataset from the model parameters, and we demonstrate that this method performs well on various architectures (random features, two-layer fully-connected and deep residual networks). Our results reveal a law of data reconstruction, according to which the entire training dataset can be recovered as p exceeds the threshold dn. }, author = {Iurada, Leonardo and Bombari, Simone and Tommasi, Tatiana and Mondelli, Marco}, booktitle = {14th International Conference on Learning Representations}, isbn = {9798331339678}, location = {Rio de Janeiro, Brazil}, pages = {145275--145314}, publisher = {OpenReview}, title = {{A law of data reconstruction for random features (and beyond)}}, volume = {2026}, year = {2026}, } @article{22950, abstract = {Little red dots (LRDs) are compact, red sources discovered by JWST at high redshift (z ≳ 4), marked by distinctive “V-shaped” spectral energy distributions (SEDs) and often interpreted as rapidly accreting active galactic nuclei (AGNs). Their true nature remains unclear though, and their evolutionary connection to their lower-redshift counterparts is still poorly constrained. Thus, we present WISEA J123635.56+621424.2 (here dubbed the Saguaro), a z = 2.0145 galaxy in GOODS-North, as a possible analog of high-redshift LRDs and a potential missing link in their evolutionary path toward lower-redshift systems. It features a compact LRD-like nucleus surrounded by a face-on spiral host. Its connections to LRDs include the following: (1) its nuclear spectrum shows a clear “V-shaped” SED, and (2) when redshifted to z = 7, surface brightness dimming makes the host undetectable, thus mimicking an LRD. This suggests that high-redshift LRDs may be embedded in extended hosts. To test this, we stack rest-frame UV images of 99 photometrically selected LRDs, revealing faint, diffuse emission. Stacking in redshift bins reveals mild radial growth, consistent with the expected galaxy size evolution. A simple analytic model confirms that surface brightness dimming alone can explain their compact appearance. Lastly, we show that the Saguaro is not unique by describing similar objects from the literature at z ≲ 3.5. Taken together, our results support a scenario in which LRDs may not be a distinct population, but could be the visible nuclei of galaxies undergoing a short-lived, (perhaps) AGN-dominated evolutionary phase, with their compact, red appearance driven largely by observational biases.}, author = {Rinaldi, Pierluigi and Rieke, George H. and Wu, Zihao and Gilbert, Carys J. E. and Pacucci, Fabio and Barchiesi, Luigi and Alberts, Stacey and Carniani, Stefano and Bunker, Andrew J. and Bhatawdekar, Rachana and D’Eugenio, Francesco and Ji, Zhiyuan and Johnson, Benjamin D. and Hainline, Kevin and Kokorev, Vasily and Kumari, Nimisha and Iani, Edoardo and Lyu, Jianwei and Maiolino, Roberto and Parlanti, Eleonora and Robertson, Brant E. and Sun, Yang and Vignali, Cristian and Williams, Christina C. and Willmer, Christopher N. A. and Zhu, Yongda}, issn = {1538-4357}, journal = {The Astrophysical Journal}, number = {2}, publisher = {IOP Publishing}, title = {{Beyond the dot: An LRD-like nucleus at the heart of an IR-bright galaxy and its implications for high-redshift LRDs}}, doi = {10.3847/1538-4357/ae80cd}, volume = {1006}, year = {2026}, } @inproceedings{22921, abstract = {Runtime monitoring of quantitative signals faces a fundamental trade-off between volatility and over-aggregation: instantaneous observations are noisy, while long-run averages obscure local structure. Localisation measures such as discounted averages offer a principled middle ground, yet remain poorly understood in runtime verification. This paper studies discounted sums from a monitoring perspective, in both deterministic and stochastic settings. We formalize the discounted monitoring problem and show that exact, sound monitoring of discounted sums cannot be achieved with finite memory. To overcome this impossibility, we introduce ε-approximately sound monitoring, deriving explicit bounds on memory and observation requirements. We then extend the framework to stochastic processes via expected discounted sums, defining pointwise and uniform (ε,δ)-soundness notions, establishing statistical optimality, and proving impossibility beyond a precision threshold. We also formalize the resource complexity of deterministic discounted monitoring via affine register machines and prove a tight worst-case lower bound. Finally, we present a specification language for arithmetic expressions over multiple discounted sums with synchronous and asynchronous semantics, and evaluate our approach on practical scenarios including algorithmic fairness.}, author = {Cano Cordoba, Filip and Henzinger, Thomas A and Kueffner, Konstantin and Sarac, Naci E}, booktitle = {37th International Conference on Concurrency Theory}, isbn = {9783959774475}, issn = {1868-8969}, keywords = {Runtime Verification, Probabilistic Systems, Quantitative Verification, Approximate Monitoring}, location = {Liverpool, United Kingdom}, publisher = {Schloss Dagstuhl - Leibniz-Zentrum für Informatik}, title = {{Monitoring discounted sum properties}}, doi = {10.4230/LIPIcs.CONCUR.2026.22}, volume = {391}, year = {2026}, } @inproceedings{22922, abstract = {In this note, we recall the history and our motivation behind the development of Strategy Logic and we discuss some of the work that ensued from its introduction.}, author = {Chatterjee, Krishnendu and Henzinger, Thomas A and Piterman, Nir}, booktitle = {37th International Conference on Concurrency Theory}, isbn = {9783959774475}, issn = {1868-8969}, keywords = {Strategy Logic, Games, Automata}, location = {Liverpool, United Kingdom}, publisher = {Schloss Dagstuhl - Leibniz-Zentrum für Informatik}, title = {{A look back at strategy logic}}, doi = {10.4230/LIPIcs.CONCUR.2026.6}, volume = {391}, year = {2026}, } @article{22951, abstract = {Anesthesia recovery is critical for resuming normal physiological and neuronal functions; however, the mechanisms involved remain elusive. Here, we identify a female-selective corticosterone-mediated microglia-neuron interaction during ketamine anesthesia recovery, absent in males. This microglia-neuron interaction induces plastic and functional neuronal changes, as evidenced by increased mEPSC frequency, which was occluded upon microglia depletion. We showed that this process is driven through up-regulation of the stress-responsive co-chaperone Fkbp5 mRNA and its protein, Fkbp51, in female microglia. Fkbp5/Fkbp51 is a key intermediary in a corticosteroid-induced stress response, and its involvement points toward a critical interface between endocrine signaling and microglia. To counteract the observed ketamine anesthesia-mediated increase in blood corticosterone during recovery, we removed the primary source of corticosterone by adrenalectomy. Close microglia-neuron interaction was reduced and increased again following corticosterone injection. Our findings identify a sex-specific microglia-mediated mechanism of neuronal plasticity during anesthesia recovery, driven by corticosterone, thereby enhancing our understanding of sex differences in brain function.}, author = {Venturino, Alessandro and Alam, Amin and Negrello, Thomas and Jin, Kelly and van Velthoven, Cindy T. J. and Cubero, Ryan J and Yeung, Jake and Koppensteiner, Peter and Tasic, Bosiljka and Siegert, Sandra}, issn = {2375-2548}, journal = {Science Advances}, number = {31}, publisher = {AAAS}, title = {{Corticosterone-linked microglial activity underpins sexually dimorphic neuroplasticity after ketamine anesthesia}}, doi = {10.1126/sciadv.adz6517}, volume = {12}, year = {2026}, } @article{22952, abstract = {Mechanosensitive ion transport plays a central role in enabling living systems to perceive and adapt to their environment through the deformation of soft, embedded ion channels. In this work, we demonstrate that ion transport within a two-dimensional graphite nanoslit can be rationally engineered to achieve a bipolar, pressure-sensitive response without any structural deformation. The mechanosensitivity arises from the selective charging of one channel inlet, which acts as a reversible source of mobile charge carriers. These excess ions can then be advected in or out of the channel by the pressure-driven water flow, thereby modulating the ionic conductance. This mechanism is captured through a comprehensive electrohydrodynamic model that analytically accounts for coupled diffusion, convection, surface transport, diffusio-osmosis, and interfacial slippage, both inside and outside the nanoslit. The theoretical framework quantitatively reproduces the experimental data, showing that a simple surface charge pattern can give rise to complex, pressure-dependent conductance. These findings reveal how rich nonlinear couplings at the nanoscale can be harnessed to design adaptive, bioinspired nanofluidic systems, exemplified here by ionic pressure sensors.}, author = {Lizée, Mathieu and Zhang, Zhijia and Coquinot, Baptiste and Yang, Qian and Bocquet, Lydéric}, issn = {2375-2548}, journal = {Science Advances}, number = {31}, publisher = {AAAS}, title = {{Architecting mechanosensitive nanofluidic transport in graphite nanoslits}}, doi = {10.1126/sciadv.aed4451}, volume = {12}, year = {2026}, } @article{22954, abstract = {The aim of the present paper is to propose and study a dissipative variant of symplectic billiards within planar strictly convex domains. The associated billiard map is dissipative, thus it admits a compact invariant set, the so-called Birkhoff attractor. Its complexity depends on the rate of dissipation as well as on the geometry of the billiard table. We prove that (a) for strong dissipation, the Birkhoff attractor is a normally contracted graph over the zero section; (b) for mild dissipation, the Birkhoff attractor within a centrally symmetric domain is an indecomposable continuum whose restricted dynamics has positive topological entropy. We compare these results with those for dissipative Birkhoff billiards, studied in [15].}, author = {Baracco, Luca and Bernardi, Olga and Florio, Anna and Nardi, Alessandra}, issn = {1572-9222}, journal = {Journal of Dynamics and Differential Equations}, keywords = {Dissipative symplectic billiard, Birkhoff attractor, Conformally symplectic twist map}, publisher = {Springer}, title = {{Birkhoff attractors for dissipative symplectic billiards}}, doi = {10.1007/s10884-026-10528-9}, year = {2026}, } @article{22953, abstract = {We prove that any finitely smooth, axially symmetric, strictly convex domain with everywhere positive curvature and sufficiently close to an ellipse is area spectrally rigid. This means that any area-isospectral family of domains in this class is necessarily equi-affine. We use techniques, adapted to symplectic billiards, inspired by the paper of J. De Simoi, V. Kaloshin and Q. Wei [5].}, author = {Baracco, Luca and Bernardi, Olga and Nardi, Alessandra}, issn = {1930-532X}, journal = {Journal of Modern Dynamics}, pages = {469--488}, publisher = {American Institute of Mathematical Sciences}, title = {{Area spectral rigidity for axially symmetric symplectic billiard tables}}, doi = {10.3934/jmd.2026012}, volume = {22}, year = {2026}, } @article{22957, abstract = {Linguistic inequality remains a silent barrier in contemporary science, particularly in rapidly evolving and interdisciplinary fields such as bioinformatics. This article examines how the Spanish-speaking community has responded to this gap through the creation of the Simposio de Estudiantes Hispanohablantes de Bioinformática y Biología Computacional (SEH2Bioinfo), a collective initiative that transforms language into a tool for inclusion and collaboration. Drawing on a historical review of the use of Spanish in scientific communication and an analysis of the two editions of SEH2Bioinfo (2022 and 2025), complemented by surveys of more than 1500 participants, we explore the effects of language on confidence, participation, and academic visibility. The findings show that Spanish can serve as a space for training and mentorship without compromising rigor, while simultaneously strengthening international networks and promoting open science. The growth of SEH2Bioinfo, its collaborative approach, and its intercontinental reach establish the symposium as a model for linguistic equity in scientific practice. Its experience demonstrates that expanding linguistic diversity in bioinformatics is not only possible but essential to building a global scientific community that is fairer, more connected, and representative.}, author = {Padilla Franzotti, Carla L and Olguín-Orellana, Gabriel J and Palopoli, Nicolas and Urquiza-Zurich, Sebastián and Garcia-Moreno, Adrian and Monzón, Sara and Cabrera-Pasadas, Mónica and Soler Sáez, Irene and Perpiñá-Clérigues, Carla and Puche Quiñonez, Rafael J and Vélez Segura, Jennifer and Lovaco-Flores, Jose Abel and Osorio Mogollón, Cleidy M and Esquinas-Román, Eva M and Miserendino, Maria C and Eranti, Pradeep and Castillo-Orozco, Ana and Cuesta-Astroz, Yesid and Parra, R Gonzalo}, issn = {2635-0041}, journal = {Bioinformatics Advances}, number = {1}, publisher = {Oxford University Press}, title = {{Bioinformatics in Español: Expanding open, diverse, and collaborative science through SEH2Bioinfo}}, doi = {10.1093/bioadv/vbag182}, volume = {6}, year = {2026}, } @article{22958, abstract = {We propose a sensing scheme for detecting weak forces that achieves sensitivity scaling proportional to the number of individual sensors without relying on entanglement or other non-classical resources. Our scheme utilizes coherent averaging across a chain of optomechanical cavities, unidirectionally coupled via a laser beam. As the beam passes through the cavities, it accumulates phase shifts induced by a common external force acting on the mechanical elements. Remarkably, this fully classical approach achieves the sensitivity scaling typically associated with quantum-enhanced protocols, providing a robust and experimentally feasible route to precision sensing. Potential applications range from high-sensitivity gravitational field measurements at the Large Hadron Collider to probing dark matter interactions and detecting gravitational waves. This work opens a new pathway for leveraging coherent light-matter interactions for force sensing.}, author = {Marchese, Marta M and Braun, Daniel and Nimmrichter, Stefan and Rätzel, Dennis}, issn = {1367-2630}, journal = {New Journal of Physics}, number = {8}, publisher = {IOP Publishing}, title = {{Cascaded optomechanical sensing for small signals}}, doi = {10.1088/1367-2630/ae8b16}, volume = {28}, year = {2026}, } @article{22959, abstract = {With this status report, we aim to provide a timely snapshot of the protein engineering field as a broad and rapidly advancing discipline that integrates computational, molecular biology, structure-guided, evolutionary, and synthetic approaches to create new and improved proteins with tailored structures and useful functions. The report is organized into eight thematic areas spanning core methodologies and major application domains, including enzymes, therapeutics, detection, synthetic biology, and materials. Contributions from experts across these areas highlight both the historical foundations and recent advances in their respective fields, with particular emphasis on the growing influence of machine learning and artificial intelligence-based methods. Emerging from this broad overview is a central message: protein engineering appears to be entering a golden age, defined by a rapidly accelerating pace of progress, even as significant challenges in design, screening, and real-world application remain. Looking ahead, the continued integration of computational and experimental strategies is poised to further accelerate the impact of protein engineering across an expanding range of economically and societally important sectors, from therapeutics and molecular imaging to diagnostics, plastic recycling, and industrial chemistry.}, author = {Ai, Hui-wang and Arnold, Frances H and Barrick, Doug and Bayley, Hagan and Brownless, Alfie-Louise R and Campbell, Robert E and Castells-Graells, Roger and Chaiyen, Pimchai and Chica, Roberto A and Chisholm, Lauren O and Christ, Daniel and Chu, Guo-Chao and Chungyoun, Michael and Eijsink, Vincent G H and Elowitz, Michael B and Fansher, Douglas J and Fasan, Rudi and Foley, Gabriel and Frei, Michelle S and Gagsteiger, Andreas and Gao, Ruiyan and Gillam, Elizabeth M J and Gray, Jeffrey J and Guerra, Luis F and Höcker, Birte and Jin, Zhiyang and Kamerlin, Shina Caroline Lynn and Koga, Nobuyasu and Kontermann, Roland E and Kuhlman, Brian and Langlois, Olivia A and Li, Dongyang and Lins, Roberto D and Liu, Chang C and Liu, Lei and López Barreiro, Diego and Lu, Peilong and Mehta, Sohum and Miyazaki, Ikumi and Müller, Manuel M and Muyldermans, Serge and Nemoto, Naoto and Nivala, Jeff and Nuñez, James K and Ornelas, Izaiah J and Pelletier, Joelle N and Plückthun, Andreas and Porth, Owen T and Prakinee, Kridsadakorn and Qing, Yujia and Reisenbauer, Julia and Schmidt, Sandy and Seifert, Oliver and Shapiro, Mikhail G and Sharma, Rachna and Singh, Nilmani and Smith, Benjamin J and Spangler, Jamie B and Tatsumi-Koga, Rie and Terai, Takuya and Terholsen, Henrik and Tolhurst, Mattias and Tripp, Katherine W and Trossmann, Vanessa Tanja and Tucker, Chandra L and Turak, Onur and Van Deventer, James A and Várnai, Anikó and Viana, Isabelle F T and Weeks, Amy M and Wittrup, K Dane and Wong, Kingsley L-J and Xi, Chenggang and Yeates, Todd O and Yeh, Andy Hsien-Wei and Zhang, Jin and Zhang, Tiger S and Zhao, Huimin and Zheng, Fang and Zhu, Jingyi and Zinn, Sacha}, issn = {1741-0134}, journal = {Protein Engineering, Design and Selection}, publisher = {Oxford University Press}, title = {{Protein engineering: Status report}}, doi = {10.1093/protein/gzag020}, volume = {39}, year = {2026}, } @phdthesis{21198, abstract = {In recent years there has been a massive increase in the amount of data generated in a decentralized manner. Ever more powerful edge devices, such as smartphones, have become ubiquitous in most societies on earth. Through text typed, photos taken and apps used, these devices, which we refer to as clients, generate enormous amounts of high quality and complex data. Moreover, the nature of these devices means the data they generate is often sensitive and privacy concerns prevent it being gathered and stored in a central location. This presents a challenge to the modern machine learning paradigm that requires central access to large amounts of data. Federated learning (FL) has emerged as one of the answers to this problem. Rather than bringing the data to the model, FL sends the model to the data. Model training takes place on device, with periodically synchronized updates, allowing data to remain locally stored. While this approach offers significant privacy advantages it comes with its own set of unique challenges. These include: data heterogeneity, the notion that different devices generate data in distinct ways which can negatively impact training dynamics; systems heterogeneity, meaning that different devices may have differing hardware specifications; high communication costs, which are induced by the repeated transferring of models over the network and low device computational power, which limits the use of larger models on device. In this thesis we present a range of methods for federated learning. We focus primarily on the challenge of data heterogeneity, though the methods presented are designed to be well adapted to the other challenges of a federated setting, such as the constraints of limited compute and communication overhead. We first present a method for explicitly modeling client data heterogeneity. The approach formulates clients as samples from a certain probability distribution and infers the parameters of this distribution from the available training clients. This learned distribution then represents the heterogeneity present among the clients and can be sampled from in order to create new simulated clients that are similar to the real clients we have observed so far. Following this we present two methods for directly dealing with data heterogeneity through personalization. Highly heterogeneous client data distributions can mean that learning a single global model becomes suboptimal, and some form of personalization of models to each individual client is required. Our approaches are based around hypernetworks, which we use to generate personalized model parameters without the need for additional training or finetuning. In the first approach we focus on generating full parameterizations of client models using learned embeddings of client data and labels, with a hypernetwork located on the central server. In the second approach we address the more challenging scenario where we want to generate a personalized model for a client without any label information. The hypernetwork is trained to generate a low dimensional representation of a client’s personalized model parameters, allowing it to be transferred to and run on the client devices. In our final presented method, we change our focus and rather than aim to directly address the challenge of data heterogeneity, we instead ensure we are unaffected by it. This is done in the context of k-means clustering and we present a method for federated clustering with a focus on added privacy guarantees.}, author = {Scott, Jonathan A}, issn = {2663-337X}, pages = {158}, publisher = {Institute of Science and Technology Austria}, title = {{Data heterogeneity and personalization in federated learning}}, doi = {10.15479/AT-ISTA-21198}, year = {2026}, } @phdthesis{21854, abstract = {As neural-network-based models grow both in size and popularity, interest has grown in making the models smaller and more efficient to train. To that end, many methods have been proposed to prune models by reducing their number of nonzero parameters. Additionally, parameter-efficient fine-tuning, in which a much smaller number of parameters than the total contained in the model is updated during training, has become very popular, especially in the space of Large Language Models. At the same time, the increasingly routine deployment of machine learning in real-world applications has spurred a drive to make them more trustworthy - in the sense of, among other things, being unbiased, interpretable, and editable. In this thesis, we examine the interplay between efficiency and trustworthiness. First, we analyze the effects of model pruning on bias in computer vision models, demonstrating that increased sparsity leads to greater bias, largely as a function of increased model uncertainty in marginal cases. Based on this observation, we propose several bias mitigation techniques. Then, we demonstrate that example-specific model pruning can improve model interpretation methods while improving pruning efficiency to make example-specific model pruning feasible in real time. Then, we investigate the effectiveness of parameter-efficient and data-efficient model personalization via fine-tuning, demonstrating that it is highly feasible with very small computational and data resources. Finally, we consider efficiency in editing model knowledge using a custom synthetic data framework, demonstrating that parameter-efficient, low-rank fine-tuning frequently outperforms full-rank fine-tuning, and, additionally, that restricting which model blocks are fine-tuned frequently improves results. Together, the results in this thesis provide new insights and techniques for combining trustworthiness and efficiency during neural network inference and training. }, author = {Iofinova, Eugenia B}, issn = {2663-337X}, pages = {237}, publisher = {Institute of Science and Technology Austria}, title = {{On the utility and effects of efficiency in artificial neural networks}}, doi = {10.15479/AT-ISTA-21854}, year = {2026}, } @article{22960, abstract = {We construct and characterize a sample of 146 little red dots (LRDs) across 2.0