arXiv preprints from January 1, 2026 through September 22, 2026 — 08:53:30 EST
Tool-Integrated Reasoning (TIR) empowers large language models (LLMs) to tackle complex tasks by interleaving reasoning steps with external tool interactions. However, existing reinforcement learning methods typically rely on outcome- or trajectory-level rewards, assigning uniform advantages to all steps within a trajectory. This...
We disprove the natural domain extension of the Berger--Coburn heat-flow conjecture for Toeplitz operators on the Bargmann space and identify the failure mechanism as a gap between pointwise and uniform control of a Gaussian averaging of the squared modulus of the symbol, a gap that is invisible to the linear form $T_g$. We establish...
Vision-Language Models (VLMs) create a severe visual feature bottleneck by using a crude, asymmetric connection that links only the output of the vision encoder to the input of the large language model (LLM). This static architecture fundamentally limits the ability of LLMs to achieve comprehensive alignment with hierarchical visual...
The ongoing Euclid mission aims to measure spectroscopic redshifts for approximately two million galaxies using the H $α$ line emission detected in near-infrared slitless spectroscopic data from the Euclid Deep Fields (EDFs). These measurements will reach a flux limit of $5\times 10^{-17}\,{\rm erg}\,{\rm cm}^{-2}\,{\rm s}^{-1}$ in...
Diffusion models have shown remarkable empirical success in sampling from rich multi-modal distributions. Their inference relies on numerically solving a certain differential equation. This differential equation cannot be solved in closed form, and its resolution via discretization typically requires many small iterations to produce...
Recent advances in end-to-end autonomous driving show that policies trained on patch-aligned features extracted from foundation models generalize better to Out-of-Distribution (OOD). We hypothesize that due to the self-attention mechanism, each patch feature implicitly embeds/contains information from all other patches, represented in...
The dynamic trees problem is to maintain a tree under edge updates while supporting queries like connectivity queries or path queries. Despite the first data structure for this fundamental problem -- the link-cut tree -- being invented 40 years ago, our experiments reveal that they are still the fastest sequential data structure for...
We study a semi-asynchronous client-server perceptron trained via iterative parameter mixing (IPM-style averaging): clients run local perceptron updates and a server forms a global model by aggregating the updates that arrive in each communication round. The setting captures three system effects in federated and distributed...
We summarize recent work to develop an effective theory of muon-to-electron conversion, based on a complete set of low-energy effective operators that are developed from a systematic expansion in velocities and momenta. The expansion effectively factors rates into sums of particle physics and nuclear physics terms, where the former...
While spin squeezing has been traditionally considered in all-to-all interacting models, recent works have shown that it can also occur in systems with power-law interactions, enabling direct tests in Rydberg atoms, trapped ions, ultracold atoms, and nitrogen-vacancy (NV) centers in diamond. For the latter, Wu et al. Nature 646 (2025)...
Deploying large language models in long-horizon, goal-oriented interactions remains challenging because similar entities and facts recur under different latent goals and constraints, causing memory systems to retrieve context-mismatched evidence. We propose STITCH (Structured Intent Tracking in Contextual History), an agentic memory...
Issue resolution, a complex Software Engineering (SWE) task integral to real-world development, has emerged as a compelling challenge for artificial intelligence. The establishment of benchmarks like SWE-bench revealed this task as profoundly difficult for large language models, thereby significantly accelerating the evolution of...
Federated learning enables multiple parties to jointly train learning models without sharing their own underlying data, offering a practical pathway to privacy-preserving collaboration under data-governance constraints. Continued study of federated learning is essential to address key challenges in it, including communication...
Concept-based explanations quantify how high-level concepts (e.g., gender or experience) influence model behavior, which is crucial for decision-makers in high-stakes domains. Recent work evaluates the faithfulness of such explanations by comparing them to reference causal effects estimated from counterfactuals. In practice, existing...
We investigate late-time cosmic acceleration in $f(R,L_m)$ gravity driven by nonlinear matter contributions, focusing on the class $f(R,L_m)=R/2+c_1 L_m+c_n L_m^{n}+c_0$ with the explicit choice $L_m=ρ_m$ and an uncoupled radiation sector. We analyze two realizations: (i) Case A: $f(R,L_m)=R/2+βρ_m^{n}+γ$, where $γ$ acts as a vacuum...
We exploit the variational and Hamiltonian structures of quantum hydrodynamics with spin to unfold the correlation and torque mechanisms accompanying spin-orbit coupling (SOC) in electronic motion. Using Hamilton's action principle for the Pauli equation, we isolate SOC-induced quantum forces that act on the orbital Madelung--Bohm...
Nitinawarat and Narayan proposed a perfect secret key generation scheme for the so-called \emph{pairwise independent network (PIN) model} by exploiting the combinatorial properties of the underlying graph, namely the spanning tree packing rate. This work considers a generalization of the PIN model where the underlying graph is...
Our study examines how generative AI (GenAI) influences performance, creative self-efficacy, and cognitive load in architectural conceptual design tasks. Thirty-six student participants from Architectural Engineering and other disciplines completed a two-phase architectural design task, first independently and then with external tools...
The rotating shallow water equations (RSWE) are a mainstay of atmospheric and oceanic modeling, and their wave dynamics has close analogues in settings ranging from two-dimensional electron gases to active-matter fluids. While recent work has emphasized the topological character of RSWE wave bands, here we develop a complementary...
To increase the voltage multiplication factor in a small-sized direct-acting electron accelerator DIN-2K with an inductive energy storage and a plasma opening switch, it is necessary to ensure an increase in the rate of change of the current and its amplitude during the POS opening for the purpose of obtaining an explosive electron...
Dicke states serve as a critical resource in quantum metrology, communication, and computation. However, unitary preparation of Dicke states is limited to logarithmic depth in standard circuit models and existing constant-depth protocols require measurement and feed-forward. In this work, we present the first unitary, constant-depth...
The phase mode of a superconductor's order parameter encodes fundamental information about pairing and dissipation, but is typically inaccessible at low frequencies due to the Anderson-Higgs mechanism. Superconducting samples thinner than the London penetration depth, however, support a gapless phase mode whose dispersion can be...
Traditional assessment methods collapse when students use generative AI to complete work without genuine engagement, creating an illusion of competence where they believe they're learning but aren't. This paper presents the conversational exam -- a scalable oral examination format that restores assessment validity by having students...
Modeling complex dynamical systems under varying conditions is computationally intensive, often rendering high-fidelity simulations intractable. Although reduced-order models (ROMs) offer a promising solution, current methods often struggle with stochastic dynamics and fail to quantify prediction uncertainty, limiting their utility in...
Future motion representations, such as optical flow, offer immense value for control and generative tasks. However, forecasting generalizable spatially dense motion representations remains a key challenge, and learning such forecasting from noisy, real-world data remains relatively unexplored. We introduce FOFPred, a novel...