@article{21437,
  abstract     = {Altermagnets are a class of collinear magnets that exhibit non-relativistic spin splitting (NRSS) of electronic bands in the absence of net magnetization. Their potential to generate large spin polarization without spin-orbit coupling has created strong interest in probes that access the underlying order parameter directly. In this Perspective, we show that linear magneto-birefringence (LMB) provides a natural and broadly applicable route to detecting altermagnetic order. Building on the correspondence between the momentum-space structure of NRSS and the ferroic ordering of magnetic multipoles in real space, we demonstrate how $d$-wave and $g$-wave NRSS textures yield distinct LMB responses. We present a symmetry-based framework that identifies the optical geometries and field configurations required to isolate specific multipole components, enabling domain imaging and providing benchmarks for theoretical models of LMB.},
  author       = {Sunko, Veronika and Orenstein, J.},
  issn         = {2397-4648},
  journal      = {npj Quantum Materials},
  publisher    = {Springer Nature},
  title        = {{Linear magneto-birefringence as a probe of altermagnetism}},
  doi          = {10.1038/s41535-026-00901-8},
  year         = {2026},
}

@article{22101,
  abstract     = {Evolutionary biology examines how the genetic and phenotypic composition
of populations changes over time. An important goal is to determine the
fixation probability of a single advantageous mutant that arises in a homogeneous
population of N residents. Many real populations experience environmental
gradients that cause mutations to be beneficial in some spatial
regions but harmful in others. Here, we study the fixation probability of a
mutant placed on a simple one-dimensional spatial structure that experiences
such a gradient. The mutant’s fitness varies linearly from1 − s to 1 + s, whereas
the resident fitness is constant and equal to 1. The existing literature suggests
that such heterogeneity in the mutant’s fitness should lead to a decrease in its
fixation probability. However, in this work, we find that small, non-negligible
gradients (s < 1=√N) substantially increase the fixation probability,while larger
gradients (s > (log N)/√N) substantially decrease it.Moreover, we quantify the
strength of this phenomenon analytically and we precisely delimit the range of
the gradients for which it occurs. Our computer simulations closely match
those findings. Altogether, our results indicate that subjecting a simple
population structure to natural environmental conditions can produce strong
counterintuitive effects.},
  author       = {Svoboda, Jakub and Nemati, Hossein and Tkadlec, Josef and Kaveh, Kamran and Chatterjee, Krishnendu},
  issn         = {2041-1723},
  journal      = {Nature Communications},
  publisher    = {Springer Nature},
  title        = {{The effect of the fitness gradient on fixation probability}},
  doi          = {10.1038/s41467-026-71777-2},
  volume       = {17},
  year         = {2026},
}

@inproceedings{22103,
  abstract     = {Modern AI systems increasingly rely on opaque, highly complex models whose inner workings remain inaccessible even to experts. This opacity creates challenges for trust, accountability, and compliance with
emerging regulatory expectations such as the “right to an explanation”. While traditional explainability methods—feature attributions, counterfactuals, surrogate models—and interpretable model classes provide valuable insights for engineers, they often fall short of delivering the contextual, conversational explanations that
real users expect. Large Language Models (LLMs) offer a promising new avenue for explanation due to their
ability to engage interactively, adapt to user needs, and translate technical outputs into more accessible reasoning. However, their tendencies toward hallucination, conflict avoidance, and oversimplification introduce
serious risks when used as explanatory agents. This paper analyzes these opportunities and limitations, examines verification strategies for ensuring explanation fidelity, and situates LLM-generated explanations within
broader concerns about public trust. The paper concludes by outlining best practices and future research directions for building robust, verifiable, and human-aligned explanation systems.},
  author       = {Cano Cordoba, Filip},
  booktitle    = {Proceedings of the 18th International Conference on Agents and Artificial Intelligence},
  isbn         = {9789897587962},
  issn         = {2184-433X},
  keywords     = {Explainable AI, Large Language Models, Trust in AI},
  location     = {Marbella, Spain},
  pages        = {4689--4696},
  publisher    = {Science and Technology Publications},
  title        = {{Explaining decisions one conversation at a time: Opportunities and risks of LLMs as explainability assistants}},
  doi          = {10.5220/0014483200004052},
  volume       = {5},
  year         = {2026},
}

@article{22105,
  abstract     = {Protein conformational energy landscapes are shaped not only by intramolecular interactions but also by their environment. In protein crystals and protein–protein complexes, intermolecular contacts alter this energy landscape, but the exact nature of this alteration is difficult to decipher. Understanding how the crystal lattice affects protein dynamics is crucial for crystallography-based studies of motion, yet its influence on collective motions remains unclear. Aromatic ring flips in the hydrophobic core represent sensitive probes of such dynamics. Here, we compare the kinetics of aromatic ring flips in the protein GB1 in crystals, in complex with its binding partner IgG, and in solution, combining advanced isotope labelling with quantitative NMR methods. We show that rings in the core flip nearly a thousand times less frequently in crystals than in solution. Enhanced-sampling molecular dynamics simulations, based on a crystal structure of a GB1 variant reported in this work, reproduce these elevated barriers and reveal how the crystal restrains motions.},
  author       = {Becker, Lea Marie and Fu, Haohao and Tatman, Benjamin and Dreydoppel, Matthias and Kapitonova, Anna and Balazs, Daniel and Weininger, Ulrich and Engilberge, Sylvain and Chipot, Christophe and Schanda, Paul},
  issn         = {17554349},
  journal      = {Nature Chemistry},
  publisher    = {Springer Nature},
  title        = {{Aromatic ring flips reveal reshaping of protein dynamics in crystals and complexes}},
  doi          = {10.1038/s41557-026-02155-0},
  year         = {2026},
}

@misc{21145,
  abstract     = {Protein conformational energy landscapes are shaped not only by intramolecular interactions but also by their environment. In protein crystals and protein-protein complexes, intermolecular contacts alter this energy landscape, but the exact nature of this alteration is difficult to decipher. Understanding how the crystal lattice affects protein dynamics is crucial for crystallography-based studies of motion, yet its influence on collective motions remains unclear. Aromatic ring flips in the hydrophobic core represent sensitive probes of such dynamics. Here, we compare the kinetics of aromatic ring flips in the protein GB1 in crystals, in complex with its binding partner IgG, and in solution, combining advanced isotope labeling with quantitative NMR methods. We show that rings in the core flip nearly a thousand times less frequently in crystals than in solution. Enhanced-sampling molecular dynamics simulations, based on a new crystal structure, reproduce these elevated barriers and reveal how the crystal restrains motions. },
  author       = {Becker, Lea Marie and Schanda, Paul and Chipot, Christophe},
  publisher    = {Institute of Science and Technology Austria},
  title        = {{Additional Data for "Aromatic Ring Flips Reveal Reshaping of Protein Dynamics in Crystals and Complexes"}},
  doi          = {10.15479/AT-ISTA-21145},
  year         = {2026},
}

@inproceedings{22146,
  abstract     = {We study differentially private model training with stochastic gradient descent under learning rate scheduling and correlated noise. Although correlated noise, in particular via matrix factorizations, has been shown to improve accuracy, prior theoretical work focused primarily on the prefix-sum workload. That workload assumes a constant learning rate, whereas in practice learning rate schedules are widely used to accelerate training and improve convergence. We close this gap by deriving general upper and lower bounds for a broad class of learning rate schedules in both single- and multi-epoch settings. Building on these results, we propose a learning-rate-aware factorization that achieves improvements over prefix-sum factorizations under both MaxSE and MeanSE error metrics. Our theoretical analysis yields memory-efficient constructions suitable for practical deployment, and experiments on CIFAR-10 and IMDB datasets confirm that schedule-aware factorizations improve accuracy in private training.},
  author       = {Kalinin, Nikita and Andersson, Joel D},
  booktitle    = {7th Symposium on Foundations of Responsible Computing},
  isbn         = {9783959774192},
  issn         = {1868-8969},
  keywords     = {differential privacy, machine learning, matrix factorization},
  location     = {Cambridge, MA; United States},
  publisher    = {Schloss Dagstuhl - Leibniz-Zentrum für Informatik},
  title        = {{Learning rate scheduling with matrix factorization for private training}},
  doi          = {10.4230/LIPIcs.FORC.2026.2},
  volume       = {368},
  year         = {2026},
}

@article{22145,
  abstract     = {An in-operando electro-intercalation method for the on-chip synthesis of alkali-metal-intercalated materials and their Raman spectroscopic and transport characterization in ultrahigh vacuum (UHV) is developed. We apply this method to synthesize fulleride superconductors via Rb+ intercalation into a C60 film. During the intercalation, we monitor the stoichiometry via UHV-Raman spectroscopy and probe superconductivity via transport measurements. An increase of the superconducting transition temperature from 7.0 K to 14.5 K is observed when the stoichiometry is tuned from Rb2.7C60 to Rb3C60. In our experiment, an ionic Rb+ flux into the host material is induced by an applied electronic current via a Butler–Volmer-type mechanism. Electro-intercalation captivates through improved stoichiometric precision, the ability to smoothly vary stoichiometry via duration of current application, and the absence of a lower limit of the volume of the host material. It represents a powerful concept for the on-chip synthesis of intercalated materials, battery research, and beyond.},
  author       = {Shchukin, Konstantin P. and Gallego Lacey, Oliver N. and Coquinot, Baptiste and Jakowski, Jacek and Huang, Jingsong and Staudenmayer, Patrik and Falke, Yannic and Pandeya, Ram Prakash and Grüneis, Alexander},
  issn         = {1936-086X},
  journal      = {ACS Nano},
  keywords     = {fulleride, intercalation, alkali metal, superconductivity, Raman},
  number       = {24},
  pages        = {17360--17372},
  publisher    = {American Chemical Society},
  title        = {{On-chip tuning of superconductivity in fullerides via current-driven Rb+ intercalation}},
  doi          = {10.1021/acsnano.6c02466},
  volume       = {20},
  year         = {2026},
}

@article{22141,
  abstract     = {Molecular electrocatalysis is commonly interpreted through electronic descriptors, implicitly treating catalysts as mechanically passive during redox cycling. Yet, electron transfer often imposes structural demands on molecular scaffolds, raising the question of whether internal mechanical constraints can directly regulate access to reactive states and, in turn, catalytic outcomes. Addressing this question has remained challenging because mechanical effects are typically inseparable from changes in composition or electronic structure. Here, we achieve this separation by exploiting two constitutionally identical molecular catalysts whose only distinction is ligand geometry. This minimal geometric variation enables or suppresses intramolecular hydrogen bonding, thereby encoding distinct mechanical constraints that isolate molecular mechanics as a variable in redox accessibility. In the α isomer, molecular constraints impose a mechanically enforced barrier that severely limits access to the reactive redox state. This disrupts the temporal ordering of elementary steps, and diverts reactivity toward competing hydrogen evolution, eroding both selectivity and stability. In contrast, mechanical compliance in the β isomer enables facile access to the redox-active state, allowing CO2 activation to intrinsically outpace water activation and yielding CO selectivities exceeding 92%. Operando spectroscopy and real-time mass spectrometry, combined with computational simulation, directly resolve this mechanically gated reaction sequence as it unfolds. Molecular mechanics thus emerge as determinants that link electron flow to reaction sequencing and catalytic selectivity, revealing that constitutionally similar catalysts can be mechanically, and therefore catalytically, distinct.},
  author       = {Mendhe, Rahul Mahadeo and Christudas Dargily, Neethu and Kottaichamy, Alagar Raja and Dutt, Shifali and Sk, Mukaddar and Makri Nimbegondi Kotresh, Harish and Ottakam Thotiyl, Musthafa},
  issn         = {1520-5126},
  journal      = {Journal of the American Chemical Society},
  publisher    = {American Chemical Society},
  title        = {{Mechanical gating of redox access in molecular electrocatalysis}},
  doi          = {10.1021/jacs.6c02632},
  year         = {2026},
}

@article{22144,
  abstract     = {Most cells polarize and migrate in response to electrical fields. In this issue of Cell, Belliveau et al. identify TMEM154/Galvanin, a receptor that serves as a cellular antenna to sense electrical gradients and guide migration toward the cathode.},
  author       = {Riedl, Michael and Sixt, Michael K},
  issn         = {1097-4172},
  journal      = {Cell},
  number       = {13},
  pages        = {3845--3846},
  publisher    = {Elsevier},
  title        = {{A new sense for electrical fields}},
  doi          = {10.1016/j.cell.2026.05.038},
  volume       = {189},
  year         = {2026},
}

@article{22147,
  abstract     = {Let 1 ≤ k ≤ n and M be a random n × n matrix with independent uniformly random {±1}-entries. We
show that there exists an absolute constant c > 0 such that
P[rank(M) ≤ n − k] ≤ exp(−cnk).
This confirms a well-known prediction in the area, extending a result of Rudelson (who previously
proved this same result under the restriction k ≤ √n, via different methods).},
  author       = {Hunter, Zach and Kwan, Matthew Alan and Sauermann, Lisa and Sawhney, Mehtaab},
  issn         = {1687-0247},
  journal      = {International Mathematics Research Notices},
  number       = {12},
  publisher    = {Oxford University Press},
  title        = {{On random matrices with large corank}},
  doi          = {10.1093/imrn/rnag126},
  volume       = {2026},
  year         = {2026},
}

@article{22148,
  abstract     = {How the twin-arginine translocase (Tat) system transports fully folded substrate proteins across cellular membranes without disrupting membrane integrity has been a fundamental question in cell biology for decades. The Tat system, found in prokaryotes and plant organelles, recognizes a cargo signal peptide via a conserved twin-arginine motif. The multi-subunit Tat complex facilitates the proton-motive-force-dependent translocation process, yet its overall architecture has remained unknown. Here, we present the cryo-electron microscopy (cryo-EM) structure of the Escherichia coli (E. coli) trimeric TatB₃C₃ complex with bound substrate SufI, assembled in vivo. The complex adopts an unusual, wide-open, bowl-shaped architecture with a polar inner cavity. Unexpectedly, the cargo is engaged in a dual-contact mode: while the signal peptide binds inside one TatBC unit, the folded domain docks tightly onto an adjacent unit, possibly performing a proofreading function. This structure provides a mechanistic framework for substrate engagement and suggests the direct involvement of the entire Tat complex in substrate translocation.},
  author       = {Zhao, Ziyu and Sazanov, Leonid A},
  issn         = {1097-4164},
  journal      = {Molecular Cell},
  publisher    = {Elsevier},
  title        = {{Structure of E. Coli twin-arginine translocase (Tat) complex with bound cargo}},
  doi          = {10.1016/j.molcel.2026.05.026},
  year         = {2026},
}

@inproceedings{22007,
  abstract     = {Truncation of cryptographic outputs is a technique that was recently introduced in Baldimtsi et al. [Foteini Baldimtsi et al., 2022]. The general idea is to try out many inputs to some cryptographic algorithm until the output (e.g. a public-key or some hash value) falls into some sparse set and thus can be compressed: by trying out an expected 2^k different inputs one will find an output that starts with k zeros.
Using such truncation one can for example save substantial gas fees on Blockchains where storing values is very expensive. While [Foteini Baldimtsi et al., 2022] show that truncation preserves the security of the underlying primitive, they only consider a setting without preprocessing. In this work we show that lower bounds on the time-space tradeoff for inverting random functions and permutations also hold with truncation, except for parameters ranges where the bound fails to hold for "trivial" reasons.
Concretely, it’s known that any algorithm that inverts a random function or permutation with range N making T queries and using S bits of auxiliary input must satisfy S⋅ T ≥ Nlog N. This lower bound no longer holds in the truncated setting where one must only invert a challenge from a range of size N/2^k, as now one can simply save the replies to all N/2^k challenges, which requires S = log N⋅ N /2^k bits and allows to invert with T = 1 query.
We show that with truncation, whenever S is somewhat smaller than the log N⋅ N /2^k bits required to store the entire truncated function table, the known S⋅ T ≥ Nlog N lower bound applies.},
  author       = {Pietrzak, Krzysztof Z and Wang, Pengxiang},
  booktitle    = {6th Conference on Information-Theoretic Cryptography},
  isbn         = {9783959773850},
  issn         = {1868-8969},
  keywords     = {Time-Space Lower Bounds, Blockchains},
  location     = {Santa Barbara, CA, United States},
  publisher    = {Schloss Dagstuhl - Leibniz-Zentrum für Informatik},
  title        = {{Time-space tradeoffs of truncation with preprocessing}},
  doi          = {10.4230/LIPIcs.ITC.2025.4},
  volume       = {343},
  year         = {2025},
}

@article{22032,
  abstract     = {We prove that the focusing and defocusing continuum Calogero–Moser models are well-posed in the scaling-critical space L^2+(R). In the focusing case, this requires solutions to have mass less than that of the soliton.},
  author       = {Killip, Rowan and Laurens, Thierry and Visan, Monica},
  issn         = {2692-3688},
  journal      = {Communications of the American Mathematical Society},
  number       = {7},
  pages        = {284--320},
  publisher    = {American Mathematical Society},
  title        = {{Scaling-critical well-posedness for continuum Calogero–Moser models on the line}},
  doi          = {10.1090/cams/48},
  volume       = {5},
  year         = {2025},
}

@article{22036,
  abstract     = {We prove dispersive decay, pointwise in time, for solutions to the mass-critical nonlinear Schrödinger equation in spatial dimensions d= 1, 2, 3.},
  author       = {Fan, Chenjie and Killip, Rowan and Visan, Monica and Zhao, Zehua},
  issn         = {1432-1823},
  journal      = {Mathematische Zeitschrift},
  publisher    = {Springer Nature},
  title        = {{Dispersive decay for the mass-critical nonlinear Schrödinger equation}},
  doi          = {10.1007/s00209-025-03821-8},
  volume       = {311},
  year         = {2025},
}

@article{12311,
  abstract     = {In this note, we prove a formula for the cancellation exponent  kv,n between division polynomials  ψn  and  ϕn  associated with a sequence  {nP}n∈N of points on an elliptic curve  E  defined over a discrete valuation field  K. The formula greatly generalizes the previously known special cases and treats also the case of non-standard Kodaira types for non-perfect residue fields.},
  author       = {Naskręcki, Bartosz and Verzobio, Matteo},
  issn         = {1473-7124},
  journal      = {Proceedings of the Royal Society of Edinburgh Section A: Mathematics},
  keywords     = {Elliptic curves, Néron models, division polynomials, height functions, discrete valuation rings},
  number       = {5},
  pages        = {1646--1660},
  publisher    = {Cambridge University Press},
  title        = {{Common valuations of division polynomials}},
  doi          = {10.1017/prm.2024.7},
  volume       = {155},
  year         = {2025},
}

@article{12662,
  abstract     = {Modern machine learning tasks often require considering not just one but multiple objectives. For example, besides the prediction quality, this could be the efficiency, robustness or fairness of the learned models, or any of their combinations. Multi-objective learning offers a natural framework for handling such problems without having to commit to early trade-offs. Surprisingly, statistical learning theory so far offers almost no insight into the generalization properties of multi-objective learning. In this work, we make first steps to fill this gap: We establish foundational generalization bounds for the multi-objective setting as well as generalization and excess bounds for learning with scalarizations. We also provide the first theoretical analysis of the relation between the Pareto-optimal sets of the true objectives and the Pareto-optimal sets of their empirical approximations from training data. In particular, we show a surprising asymmetry: All Pareto-optimal solutions can be approximated by empirically Pareto-optimal ones, but not vice versa.},
  author       = {Súkeník, Peter and Lampert, Christoph},
  issn         = {1433-3058},
  journal      = {Neural Computing and Applications},
  pages        = {24669–24683},
  publisher    = {Springer Nature},
  title        = {{Generalization in multi-objective machine learning}},
  doi          = {10.1007/s00521-024-10616-1},
  volume       = {37},
  year         = {2025},
}

@inproceedings{20820,
  abstract     = {The high computational costs of large language models (LLMs) have led to a flurry of research on LLM compression, via methods such as quantization, sparsification, or structured pruning. A new frontier in this area is given by dynamic, non-uniform compression methods, which adjust the compression levels (e.g., sparsity) per-block or even per-layer in order to minimize accuracy loss, while guaranteeing a global compression threshold. Yet, current methods rely on estimating the "importance" of a given layer, implicitly assuming that layers contribute independently to the overall compression error. We begin from the motivating observation that this independence assumption does not generally hold for LLM compression: pruning a model further may even significantly recover performance. To address this, we propose EvoPress, a novel evolutionary framework for dynamic LLM compression. By formulating dynamic compression as a general optimization problem, EvoPress identifies optimal compression profiles in a highly efficient manner, and generalizes across diverse models and compression techniques. Via EvoPress, we achieve state-of-the-art performance for dynamic compression of Llama, Mistral, and Phi models, setting new benchmarks for structural pruning (block/layer dropping), unstructured sparsity, and quantization with dynamic bitwidths.},
  author       = {Sieberling, Oliver and Kuznedelev, Denis and Kurtic, Eldar and Alistarh, Dan-Adrian},
  booktitle    = {42nd International Conference on Machine Learning},
  issn         = {2640-3498},
  location     = {Vancouver, Canada},
  pages        = {55556--55590},
  publisher    = {ML Research Press},
  title        = {{EvoPress: Accurate dynamic model compression via evolutionary search}},
  volume       = {267},
  year         = {2025},
}

@inproceedings{20821,
  abstract     = {Modern deep neural networks exhibit heterogeneity across numerous layers of various types such as residuals, multi-head attention, etc., due to varying structures (dimensions, activation functions, etc.), distinct representation characteristics, which impact predictions. We develop a general layer-wise quantization framework with tight variance and code-length bounds, adapting to the heterogeneities over the course of training. We then apply a new layer-wise quantization technique within distributed variational inequalities (VIs), proposing a novel Quantized Optimistic Dual Averaging (QODA) algorithm with adaptive learning rates, which achieves competitive convergence rates for monotone VIs. We empirically show that QODA achieves up to a 150% speedup over the baselines in end-to-end training time for training Wasserstein GAN on 12+GPUs.},
  author       = {Nguyen, Anh Duc and Markov, Ilia and Wu, Frank Zhengqing and Ramezani-Kebrya, Ali and Antonakopoulos, Kimon and Alistarh, Dan-Adrian and Cevher, Volkan},
  booktitle    = {42nd International Conference on Machine Learning},
  issn         = {2640-3498},
  location     = {Vancouver, Canada},
  pages        = {46026--46072},
  publisher    = {ML Research Press},
  title        = {{Layer-wise quantization for quantized optimistic dual averaging}},
  volume       = {267},
  year         = {2025},
}

@article{20839,
  abstract     = {For every couple of Hausdorff functions ψ and φ verifying some mild assumptions, there exists a compact subset K of the Baire space such that the φ-Hausdorff measure and the ψ-packing measure on K are both finite and positive. Such examples are then embedded in any infinite dimensional Banach space to answer positively a question of Fan on the existence of metric spaces with arbitrary scales.},
  author       = {Helfter, Mathieu},
  issn         = {2308-1317},
  journal      = {Journal of Fractal Geometry},
  publisher    = {EMS Press},
  title        = {{Sets with arbitrary Hausdorff and packing scales in infinite dimensional Banach spaces}},
  doi          = {10.4171/jfg/177},
  year         = {2025},
}

@misc{20842,
  abstract     = {Probing the possibility of entanglement generation through gravity offers a path to tackle the question of whether gravitational fields possess a quantum mechanical nature. A potential realization necessitates systems with low-frequency dynamics at an optimal mass scale, for which the microgram-to-milligram range is a strong contender. Here, after refining a figure-of-merit for the problem, we present a 1-milligram torsional pendulum operating at 18 Hz. We demonstrate laser cooling its motion from room temperature to 240~microkelvins, surpassing by over 20-fold the coldest motions attained for oscillators ranging from micrograms to kilograms. We quantify and contrast the utility of the current approach with other platforms. The achieved performance and large improvement potential highlight milligram-scale torsional pendulums as a powerful platform for precision measurements relevant to future studies at the quantum-gravity interface.},
  author       = {Agafonova, Sofya},
  publisher    = {Institute of Science and Technology Austria},
  title        = {{Research Data for: 'One-milligram torsional pendulum toward experiments at the quantum-gravity interface'}},
  doi          = {10.15479/AT-ISTA-20842},
  year         = {2025},
}