arXiv preprints from January 1, 2026 through September 23, 2026 — 01:22:10 EST
Due to the prevalence of large language models (LLMs), key-value (KV) cache reduction for LLM inference has received remarkable attention. Among numerous works that have been proposed in recent years, layer-wise token pruning approaches, which select a subset of tokens at particular layers to retain in KV cache and prune others, are...
We introduce a principled learning to optimize (L2O) framework for solving fixed-point problems involving general nonexpansive mappings. Our idea is to deliberately inject summable perturbations into a standard Krasnosel'skii-Mann iteration to improve its average-case performance over a specific distribution of problems while...
In recent years, self-supervised representation learning for skeleton-based action recognition has advanced with the development of contrastive learning methods. However, most of contrastive paradigms are inherently discriminative and often struggle to capture the variability and uncertainty intrinsic to human motion. To address this...
We estimate risk premia in the cross-section of cryptocurrency returns using the Giglio-Xiu (2021) three-pass approach, allowing for omitted latent factors alongside observed stock-market and crypto-market factors. Using weekly data on a broad universe of large cryptocurrencies, we find that crypto expected returns load on both...
Hint-based faithfulness evaluations have established that Large Reasoning Models (LRMs) may not say what they think: they do not always volunteer information about how key parts of the input (e.g. answer hints) influence their reasoning. Yet, these evaluations also fail to specify what models should do when confronted with hints or...
We study an Aganagic-Vafa brane supported on a special Lagrangian submanifold $\mathcal{L}$ in a non-compact toric Calabi-Yau threefold $\mathcal{X}$. From the perspective of geometric engineering, the Aganagic-Vafa branes give rise to a special class of half-BPS codimension-two defects in 5d $\mathcal{N}=1$ supersymmetric field...
The Lifshitz formula and methods of its preparation in the literature are considered. It is shown that in Lifshitz's work itself, this formula is given without a consistent conclusion. Moreover, the approach to the conclusion proposed in this work does not allow us to obtain it. The most general conclusion of this formula can be the...
We present StdGEN++, a novel and comprehensive system for generating high-fidelity, semantically decomposed 3D characters from diverse inputs. Existing 3D generative methods often produce monolithic meshes that lack the structural flexibility required by industrial pipelines in gaming and animation. Addressing this gap, StdGEN++ is...
We investigate superconducting transmission lines as a novel platform for realizing a quantum fluid of microwave photons in a propagating geometry. We predict that the strong photon-photon interactions provided by the intrinsic nonlinearity of Josephson junctions are sufficient to enter a regime of strongly interacting photons for...
We study completion of partial matrices with nonnegative entries to matrices of nonnegative rank at most $r$ for some $r \in \mathbb{N}$. Most of our results are for $r \leq 3$. We show that a partial matrix with nonnegative entries has a nonnegative rank-1 completion if and only if it has a rank-1 completion. This is not true in...
A search is performed for lepton-number-violating $B^-\to D^{(*)+}μ^-μ^-$ decays, using data collected by the LHCb experiment in proton-proton collisions at a center-of-mass energy of 13 TeV, corresponding to an integrated luminosity of 5.4 fb$^{-1}$. No significant signal is observed, and upper limits are set on the branching...
We investigate the structure of damped two-dimensional perturbations in unstable plane-parallel shear flows with an inflection point. In inviscid flows within the stable wavenumber region $k$, no regular eigenmodes exist -- the frequency spectrum $ω$ consists of a continuous set of singular van Kampen modes with real frequencies....
We study an optimal claim reporting problem in a bonus-malus setting. We assume, that the insurance contract consists of two regimes, where reporting a claim leads to a transition to a higher-premium regime, whereas remaining claim-free for a prespecified time period results in a shift to the lower premium regime. The insured can...
Consensus protocols are crucial for a blockchain system as they are what allow agreement between the system's nodes in a potentially adversarial environment. For this reason, it is paramount to ensure their correct design and implementation to prevent such adversaries from carrying out malicious behaviour. Formal verification allows...
Thermodynamic profiles from X-ray and millimetre observations of galaxy clusters are often compared under the simplifying assumptions of smooth, spherically symmetric intracluster medium. These approximations lead to expected discrepancies in the inferred profiles, which can provide insights about the cluster structure or cosmology....
In future nuclear fusion reactors, even a small fraction of fusion-born energetic particles (EP) about 100 times hotter than the thermal bulk species, contributes substantially to the kinetic pressure and therefore affect the MHD equilibrium, mainly via the Shafranov shift. In this work, we perform first-principles numerical...
As intelligent agents become more generally-capable, i.e. able to master a wide variety of tasks, the complexity and cost of properly evaluating them rises significantly. Tasks that assess specific capabilities of the agents can be correlated and stochastic, requiring many samples for accurate comparisons, leading to added costs. In...
We provide precise asymptotics for the number of antichains in the poset $\{0,1,2\}^n$, answering a question of Sapozhenko. Finding improved estimates for this number was also a problem suggested by Noel, Scott, and Sudakov, who obtained asymptotics for the logarithm of the number. Key ingredients for the proof include a...
We develop a unified charge-dependent solar modulation model by solving the three-dimensional Parker transport equation, incorporating a realistic wavy heliospheric current sheet to treat drift effects self-consistently. Using a local interstellar spectrum from GALPROP constrained by Voyager data, we fit the model to time-resolved...
As Natural Language Generation (NLG) dominates modern NLP, scalable evaluation remains a critical bottleneck. Consequently, LLM-as-a-judge (LaaJ) adoption has accelerated rapidly, appearing in more papers than human evaluation in 2025. This pivotal shift motivates a critical analysis of current evaluation practices. Overcoming the...
Synthetic data generation is an appealing tool for augmenting and enriching datasets, playing a crucial role in advancing artificial intelligence (AI) and machine learning (ML). Not only does synthetic data help build robust AI/ML datasets cost-effectively, but it also offers privacy-friendly solutions and bypasses the complexities of...
Pit thermal energy storage (PTES) provides large-scale thermal storage capacity in district heating systems, supporting flexibility on both daily and seasonal scales. Most existing large-scale energy system studies on PTES do not account for temperature differences between storage and the network. Neglecting these temperature...
Multimodal Large Language Models (MLLMs) rely on strong linguistic reasoning inherited from their base language models. However, multimodal instruction fine-tuning paradoxically degrades this text's reasoning capability, undermining multimodal performance. To address this issue, we propose a training-free framework to mitigate this...
Traditional two-dimensional risk matrices (heatmaps) are widely used to model and visualize likelihood and impact relationships, but they face fundamental methodological limitations when applied to complex infrastructures. In particular, regulatory frameworks such as NIS2 and DORA call for more context-sensitive and system-oriented...
One of the oldest flow visualization techniques is through multiple pathlines generated by the movement of seeding particles spatially distributed in the flow. In the computerized era, particle images are used in quantitative measurements, such as particle image and particle tracking velocimetry (PIV and PTV). Here, we present several...