Peer Revenue Now accepting positions

Papers

Fresh from arXiv, typeset properly. Highlighted words trade on the market.

All of arXiv

Saturday 7 November

Machine Learning

DPDP: Diffusion Policy-Steered Diffusion Planner for Offline Reinforcement Learning

In offline reinforcement learning, diffusion models fill two roles with complementary strengths. Typically guided by a critic, a diffusion policy is value-aware and favors high-return, behavior-supported actions, but it is short-sighted: it never lays out the states it intends to visit. A diffusion planner is far-sighted, synthesizing such a long-horizon trajectory, but it is only behavior-driven at execution, realizing each planned transition using only offline behavior.Each is strong exactly where the other is weak. We introduce DPDP to bring them together at the planner's execution step, where a planned transition becomes an action. Reading the planner's inverse-dynamics executor as a posterior over actions: the environment's transition likelihood relates an action to the planned next state, while the behavioral prior determines its weight under the offline data. Replacing this prior with a value-optimized diffusion policy preserves the planned transition likelihood and favors valuable actions that can serve the plan. We target the resulting clean-action distribution in two ways. A training-free sampler composes policy and inverse-dynamics scores at matched action noise levels and applies Langevin correction; this route uses a shared action noise schedule and an approximately common terminal prior. A preference objective instead amortizes the transfer into a single diffusion executor. Across D4RL, DPDP improves the planner it builds on, staying competitive with reactive diffusion policies on locomotion and leading on long-horizon navigation.
Read the full abstract

Machine Learning

Hydro-Dataset: A Multi-Modal Open-Source Pipeline for Nordic Hydrological Research and Foundation Model Benchmarking

Despite world-class snow monitoring by the Swedish Meteorological and Hydrological Institute (SMHI) and Finnish Meteorological Institute (FMI), no station-anchored, multi-parameter snow benchmark dataset exists for the Nordic boreal domain. We present Hydro-Dataset v1.1, a reproducible open-source pipeline that harmonizes 582 SMHI meteorological observation stations across nine hydrological winters (2015/16-2023/24) into a CF-1.9 NetCDF product with 1.8~million annotated station-days. Six derived snow parameters, such as Snow Water Equivalent (SWE), snow phenology, melt rates, degree-day melt factor, rain-on-snow flags, and freeze-thaw cycles, supply benchmark-ready labels for hydrological foundation model training. ERA5-Land cross-validation (Pearson R=0.78, bias=-262mm) confirms observation-anchored SWE provides a more reliable benchmark than reanalysis for machine learning training.
Read the full abstract

Computer Vision

Sensitivity-Gated Accumulator Caching for On-Device Learning of Low-Bit Hand Networks

Small hand-tracking networks are deployed as 4- and 8-bit codes, and they should keep adapting to a user's camera after deployment, on the device and within roughly the memory of the model itself. Quantization-aware training cannot do this cheaply, because it keeps a 32-bit master copy of every weight together with its optimizer state, about 96 bits per weight. Training the codes directly with stochastic rounding needs no copy but injects rounding noise at every step, and the methods that remove this noise keep a residual buffer for every weight. We introduce sensitivity-gated accumulator caching (SGAC), which keeps residual state for only a small and changing set of weight groups. A set-associative cache of 8-bit accumulators collects the sub-step updates of the groups with the largest Fisher-weighted squared code step and writes them into the codes as whole steps; an evicted line is flushed by stochastic rounding, and a group without a line does not move until it is admitted. We prove that the write-back is unbiased and show, under a local quadratic model, that rounding a sub-step update costs loss of the same order as the update's own gain, so that rounding pays only where the expected gradient is large relative to the curvature. On a 383k-parameter hand-keypoint network, SGAC trains with 8.1 bits of state per weight. Scored on the 4/8-bit network that the device runs after adaptation, it has the highest mean accuracy of nine deployed low-bit networks on two simulated cameras. On twelve ImageNet-C corruptions it has the highest average of corrupted and clean accuracy of seven deployed methods, ahead of the PEQA-style, ECO-style and direct low-bit baselines with confidence intervals that exclude zero, and it matches an 8-bit accumulator kept for every weight with 56% of that state.
Read the full abstract

All authors Soroor Ghandali, Kanika Mamgain, Manjot Kaur, Akanksha Shukla, Shashira Guntuka

Mathematical Physics

The random matrix hard edge — applications, exact PDFs and large $N$ expansions

A hard edge of a Hermitian random matrix occurs when the eigenvalue density is strictly zero on one side of the edge, and nonzero on the other. Another possibility is that the eigenvalues occur in $\pm$ pairs, which creates a hard edge at the origin for the positive eigenvalues. Both review material and new results are presented for random matrix ensembles exhibiting a hard edge. The review material has two focuses: one is on applications of the random matrix hard edge, while the other is on exact eigenvalue probability density functions. The new results relate to the large $N$ expansion of correlation functions and conditioned gap probability generating functions with hard edge scaling for the Jacobi ensemble. For all even $\beta$ we are able to show that these quantities admit expansions in inverse even powers of the variable $N_{\rm hJ}':= N - 1/2 + (a + b + 1)/\beta$. For the $\beta =2$ Jacobi unitary ensemble, we can relate the leading (next leading) correction to the limit law to a particular second order (fourth order) differential operator applied to the latter. The analogous second order differential operator is found for $\beta=1,4$, leading then to its conjectured form for all $\beta>0$.
Read the full abstract

Analysis of PDEs

Higher-order Sobolev stability and temporal decay estimate for the tropical climate model with small $H^1$ data

We study the stability and temporal decay of solutions to the two-dimensional tropical climate model with Laplacian dissipation in both velocity equations and no thermal diffusion. For initial data in $H^s (\mathbb{R}^2)$ ($s \geq 1$) with sufficiently small $H^1$ norm, we establish a unique global solution satisfying uniform $H^s$ bounds and time-integrated estimates for the velocity and temperature. The smallness is independent of $s$, and the higher Sobolev norms of the initial data need not be small. Furthermore, without additional spatial integrability or negative-order regularity of the initial data, we prove that the $\dot H^s (\mathbb{R}^2)$ norms of the solutions are $o(t^{- s / 2})$ as $t \to \infty$. We also show that this decay rate is sharp: no additional decay factor tending to zero works for all such initial data.
Read the full abstract

Quantum Physics

Deterministically fabricated C-band single-photon sources based on quantum dot circular Bragg grating cavities

Telecom-wavelength quantum light sources are essential components for scalable fiber-based quantum networks, where on-demand emission of single photons is required for high secret key rates. Here, we demonstrate the deterministic integration of metal-organic vapor-phase epitaxy (MOVPE)-grown C-band InAs/InP quantum dots into circular Bragg grating resonators. The integrated emitters exhibit a radiative lifetime of $T_1 = (1400 \pm 50)~\mathrm{ps}$ and a spectral linewidth of $\Delta E_{\mathrm{FWHM}} = (41.0 \pm 4.0)~\mu\mathrm{eV}$. Under non-resonant excitation, we measure a post-selected single-photon purity of $(0.95 \pm 0.02)$ and a raw two-photon interference visibility of $V_{12\mathrm{ns}} = (35 \pm 2)\%$, which increases to $(70 \pm 18)\%$ under temporal post-selection. We discuss strategies to improve photon indistinguishability, including radiative-rate engineering, charge stabilization, and more selective excitation. Our results combine MOVPE growth, deterministic device integration, and telecom C-band emission, providing a route toward scalable semiconductor quantum light sources for fiber-based quantum networks.
Read the full abstract

All authors Monica Pengerla, Yury Berdnikov, Alexander Kosarev, Dominik M. Feder, Avijit Barua, Aris Koulas-Simos, Paweł Holewa, Sven Rodt, Imad Limame, Elizaveta Semenova, Stephan Reitzenstein

Machine Learning

Horizon-Scalable Q-Learning

Learning accurate off-policy state-action value functions (Q-functions) from offline datasets becomes harder as the task horizon grows. One-step temporal-difference (TD) learning bootstraps after every transition, so small per-step errors accumulate along the horizon, and early states can be badly misvalued even when the TD loss is small. Multi-step TD shortens this chain by following the dataset trajectory for several steps; however, since the behavior policy is often suboptimal, every followed action adds a bias, and no fixed number of steps suits a whole trajectory. In this paper, we propose Horizon-Scalable Q-Learning (HQL), a simple method that enables accurate long-horizon value learning by automatically balancing between one-step and multi-step returns. HQL extends a target through a dataset action only when the critic values it at least as highly as the best-of-N actions sampled from the BC policy, and otherwise falls back to a one-step target. We show that this rule is a principled off-policy correction that needs no action densities, which makes HQL compatible with expressive flow and diffusion policies. Experiments show that HQL avoids the failures of both one-step and multi-step TD on a combination-lock task across horizons, and outperforms prior horizon-reduction methods on long-horizon OGBench and OCBench tasks.
Read the full abstract

All authors Haoran Xu, Pragya Srivastava, Xinjie Liu, Siddhant Agarwal, Amy Zhang

Populations and Evolution

The forces at hand: social learning is permissive, not instructive, for the evolution of phenotypic plasticity

Evolution can favour a trait for two reasons: an instructive cause makes selection favour the trait; a permissive one lets it persist once selection no longer removes it. The distinction matters most for phenotypic plasticity and its replacement by inherited behaviour, genetic assimilation, the Baldwin effect. Social learning is credited with favouring plasticity on evidence that cannot tell the two apart: how much plasticity remains at the end of a simulation. Here we show, analytically and computationally, which role it plays in a minimal model of learning and evolution. Writing expected fitness as a function of plasticity, we find that environmental change is the only source of benefit from plasticity, and the social channel enters only its cost. Simulation confirms this: in a static environment selection against plasticity never reverses sign at any channel strength; when the environment changes, it reverses at a threshold the channel lowers, and without the channel the population collapses before reaching it. Social learning is permissive, not instructive: it supplies the means to be plastic, never the reason. For the debate over whether plasticity leads evolution, this is one case settled by measurement: environmental change directs, socially inherited information permits, and neither suffices alone.
Read the full abstract

Combinatorics

On the connected zero forcing polynomial of finite trees

Given a graph $\Gamma$ with vertex set $V(\Gamma)$ and a set of colored vertices S, the zero forcing rule is an iterative graph coloring process which states that a colored vertex with exactly one uncolored neighbor forces that neighbor to become colored. If repeated application of this rule eventually colors every vertex of $\Gamma$, the set $S$ is called a zero forcing set. Now, if the induced subgraph $\Gamma\vert_S$ is connected, then $S$ is called a connected zero forcing set. The minimum cardinality of such a set is the connected zero forcing number $Z_C(\Gamma)$. A connected zero forcing polynomial of a finite graph $\Gamma$ is a monic polynomial $Z_C^{\Gamma}(x) \;=\; \sum_{k} a_k x^k$, where $a_k$ counts the number of connected zero forcing sets of $\Gamma$ of size $k$, which we denote by $Z_C(\Gamma,k)$. In this paper, we find the connected zero forcing polynomial of any finite tree $T$. We then establish sharp lower bounds for each coefficient of the connected zero forcing polynomial, over all trees on $n$ vertices with $p$ many pendant vertices. We also identify a family of trees, which we call uniform trees, and we prove that for any tree on $n$ vertices and $p$ many pendant vertices, there exists a uniform tree on $n$ vertices and $p$ many pendant vertices, whose connected zero forcing polynomial bounds as an upper bound, the connected zero forcing polynomial of the tree coefficient wise. Again for $2\leq p\leq 4$, we establish a sharp upper bound for each coefficient of the connected zero forcing polynomial, over all trees on $n$ vertices with $p$ many pendant vertices.
Read the full abstract

Distributed Computing

Network Contention State Aware Dynamic Optimizer for Accelerator Collective Communication Libraries

Collective Communication Libraries (CCL) are critical for large-scale distributed AI/ML training, yet their configurations are determined by static heuristics that do not account for real-time network conditions. Our empirical study shows that these heuristics perform near-optimally in contention-free environments, but degrade substantially under background traffic. To address this, we present a cooperative multi-agent run-time optimizer for accelerator collective communication. Our approach employs a two-agent architecture: (i) detector- a Long Short Term Memory(LSTM) recurring neural network fed classifier head with a sequential probability ratio test that distinguishes sustained contention from transient spikes; and (ii) selector- a configuration selector built on a Dueling Double DQN (D3QN) with Conservative Q-Learning that adapts configuration parameters in response. We implemented this dynamic tuner over NCCL, a widely used CCL developed by NVIDIA. When evaluated on Chameleon Cloud's isolated bare-metal cluster, this dynamic tuning approach reduces AllReduce latency by 36% over NCCL and 32% over a fully converged AutoCCL. We further evaluated it across six real AI workloads spanning convolutional networks, transformers, and recommendation architectures. Our multi-agent tuner approach delivers up to 65% speedups on Delta cluster with NVIDIA A100 GPUs and Slingshot-11 fabric internode connectivity.
Read the full abstract

All authors Rajat Bisht, Tevfik Kosar, Anthony Skjellum, Gene Cooperman, MD S Q Zulkar Nine

Complex Variables

Holomorphic Isometric Embeddings of the Unit Disk into the Symmetrized Bidisk

We give a complete characterization of the holomorphic isometric embeddings $F:(\mathbb{D},k\,ds^2)\rightarrow (\mathbb{G}_2,ds^2)$ of the unit disk into the symmetrized bidisk with respect to the Bergman metric up to a scalar constant. We show that the isometric constant $k$ can take only the values $k=1$ and $k=3$. For $k=3$, we prove that $(2z,z^2)$ is the unique isometry up to a reparametrization. For $k=1$, we construct an isometric embedding which intersects the royal variety at a single point and does not admit a holomorphic lift to the bidisk via the symmetrization map. We further show that this isometric embedding is unique up to reparametrizations. Finally, none of the isometries are totally geodesic.
Read the full abstract

Computation and Language

Think or Remember: The Exchange Rate of Agentic Inference

Agents built on reasoning models resend their growing transcript at every call. Two transacting parties, not the model, decide what that costs: the client keeps or edits messages; the server decides whether earlier reasoning returns in the prompt and what its cache reuses. We define a cost model whose counts match an instrumented engine exactly; the architecture charges $c$ per token processed and $\beta$ per pair attended. Their ratio $P^*=c/\beta$ is an exchange rate between thinking and remembering: per token, processing dominates below this context length, a thinking regime, and attention above it, a memory regime. A policy change has a mechanical cost, incurred even if the agent writes the same tokens, and a behavioral one: it writes different tokens, changing cost and outcome. Serving alone can flip which policy is cheaper: on recorded Qwen3-8B ScienceWorld runs, dropping earlier reasoning costs 1.74 times keeping it when the server reuses generated states and 0.34 times when it reuses nothing. In live reruns, dropping it cuts correct HotpotQA answers by 9-23 percentage points on four models, at lower point-estimate cost. Comparing context policies means comparing systems: architecture and serving set the cost of the history kept and the work it induces.
Read the full abstract

All authors Juan Wisznia, Santiago Luis Groba Alonso, Matías Gómez Taddeo, Gonzalo Zabaleta, Joaquín Cullen, Luciano Del Corro

Machine Learning

An Analysis of Reinforcement Learning with Function Approximation

We address the problem of computing the optimal Q-function in Markov decision problems with infinite state-space. We analyze the convergence properties of several variations of Q-learning when combined with function approximation, extending the analysis of TD-learning in Tsitsiklis and Van Roy (1996) to stochastic control settings. We identify conditions under which such approximate methods converge with probability 1. We conclude with a brief discussion on the general applicability of our results and compare them with several related works.
Read the full abstract

2611.10981 · 7 Nov · This version of the paper corrects the one that was presented at the 25th International Conference on Machine Learning. We corrected condition (7) of Theorem 1 and its proof, as well as the importance-sampling update (11). We also made the assumptions of Theorems 1 and 2 more precise and revised the proof of Theorem 2, adding an appendix with the auxiliary results it uses

Systems and Control

An Interline Multiport DC Circuit Breaker with Power Flow Controller

Currently, the high-voltage multi-terminal DC (MTDC) grids face two major challenges. First, fast rise rate and lack of zero-crossing point in DC-side fault currents render AC circuit breakers inoperable. Second, DC line currents are passively determined by terminal DC voltages and line resistances, which inherently limits the capability of changing individual line currents for essential grid services such as power flow management and overloading prevention. To address these two limitations, DC circuit breakers (DCCBs) and DC power flow controllers (DCPFCs) have been developed. The conventional DCCBs are two-port devices installed in series with DC lines at their terminals in an MTDC converter substation, so that any fault current flowing through the line can be rapidly cleared. Additionally, a DCPFC can be installed at line terminals to adjust terminal voltages and power exchanges among adjacent lines. As a result, for a substation that connects to multiple lines, it needs to host multiple DCCBs and a DCPFC, which significantly increases the cost and footprint of both the substation and the MTDC grid. To this end, this work proposes an integrated equipment combining multiple DCCBs and a DCPFC, called the interline multiport DCCB with DCPFC, which can reduce the number of components and enhance the economic feasibility of MTDC networks. This paper presents the topology, operation principles, and component selection procedures of the proposed equipment, and compares it with the state-of-the-art technologies. Electromagnetic transient (EMT) simulation results validate the fault interruption and power flow control performance of the interline multiport DCCB with PFC.
Read the full abstract

All authors Maliha Ferdous, Zhi Jin Zhang, Qichen Yang, Maryam Saeedifard

Astrophysics of Galaxies

Tidally-induced vertical asymmetries in the LMC bar and disk

Past observations and simulations have shown that the bar in the Large Magellanic Cloud (LMC) may be tilted with respect to the disk plane, and that the outer LMC disk could be warped. These studies propose that the tilted bar and the warp in the LMC are induced by the Small Magellanic Cloud (SMC) pericentric passages. We analyze and quantify the effect of the SMC passage on the vertical position and kinematic asymmetries of both the bar and disk, as a function of different impact parameters or initial conditions. We used seven representative simulations from the KRATOS suite of $N$-body simulations of the LMC, SMC and Milky Way (MW) system for this analysis. To classify bar particles and analyze the bar and disk separately, we adopted a classification method based on persistent homology. To quantify the vertical asymmetries in the bar and disk of the LMC-like galaxy, we applied Fourier decomposition. We constructed Gaia DR3 mock catalogs to study whether the observed vertical asymmetries in Gaia DR3 are caused by astrometric uncertainties. We detect strong $m = 1$ vertical asymmetries in the bar and disk of the LMC-like galaxy for all interacting simulations after the pericentric passage of the SMC-like galaxy, indicative of a warped bar and disk. We also detect $m=2$ vertical asymmetries in the bar of the LMC-like galaxy for some simulations, indicative of a buckling bar. We conclude that the SMC passages induce vertical asymmetries in both the bar and the disk of the LMC, compatible to those measured in Gaia DR3. Tidal interactions naturally give rise to a warped and buckling LMC bar, and a warped LMC disk. Comparison between our results and Gaia DR3 suggest that models with a higher mass ratio between the SMC and the LMC may be favorable for reproducing the observed tilt of the LMC bar, and that the most recent SMC pericentric passage occurred at most 210 to 300 Myr ago.
Read the full abstract

All authors Shuyu Wang, Marie Schölch, Mercè Romero-Gómez, Óscar Jiménez-Arranz, Anthony G.A. Brown

Algebraic Geometry

Unramified case of Colliot-Thélène conjecture on quadratic forms

The Colliot-Thélène conjecture on quadratic forms asserts that any rationally isotropic nondegenerate quadratic form over a semilocal regular domain with $2$ invertible is isotropic. We prove the Colliot-Thélène conjecture for semilocal regular domains that are geometrically regular over a discrete valuation ring with $2$ invertible. In particular, this includes unramified regular local rings of mixed characteristic $(0,p)$, where $p\neq 2$.
Read the full abstract

Information Retrieval

DM-Call: Decision-Model-Guided Function Calling with Cumulative Verification Feedback

Reliable function calling requires Large Language Models (LLMs) to select appropriate functions and generate arguments consistent with user intent. However, syntactically valid calls may still be semantically incorrect due to inappropriate function selection or mismatched arguments. In this paper, we propose Decision-Model-Guided Function Calling (DM-Call), a framework that combines local schema validation with separate decision-model verification of function selection and argument correctness. For rejected candidates, DM-Call accumulates verification probabilities, binary judgments, and schema errors across attempts. A hybrid feedback representation further combines numerical probabilities with verbal correctness labels to guide iterative refinement without executing external tools. Extensive evaluation across BFCL and API-Bank (first-call) demonstrates DM-Call boosts function-calling accuracy versus direct generation, at modest additional cost.
Read the full abstract

Superconductivity

Superconductivity and Polar Metallicity in Atomically Thin Nitride Superlattices

The integration of polar and superconducting materials in the atomically thin limit typically results in a competition that suppresses the superconducting order. Here, we challenge this conventional view by employing first-principles calculations to investigate 1:1 superlattices composed of alternating atomically thin layers of TiN/AlN and {\delta}-NbN/AlN. Contrary to the expectation that dense insulating barriers would quench superconductivity, we find that the polar interface can coexist with a superconducting state. We observe that the transition temperature (Tc) of TiN is enhanced by nearly 2 K over its bulk value, while {\delta}-NbN maintains a relatively high Tc exceeding 20 K. We demonstrate that this resilience arises from the formation of a polar metal, where the strong internal electric fields of the AlN layers are partially screened by a charge redistribution within the nitride layers. This screening mechanism, along with the geometric confinement, creates a quasi-two-dimensional electronic environment, establishing these superlattices as a novel platform for exploring the coexistence of structural inversion-symmetry breaking and superconductivity in the two-dimensional limit.
Read the full abstract

All authors Huma Nawaz, Dawei Wang, C. S. Ting, Laurent Bellaiche

Statistical Methodology

Inference for Non-Probability Samples Using Error-Prone Auxiliary Information

Non-probability samples are a potential source of rich information for making statistical inferences. However, these datasets often suffer from quality issues, in particular selection bias. The data from a non-probability sample may be combined with auxiliary information from the population or from a reference probability sample to address this bias. The auxiliary information present on these other sources is generally considered to be error-free, however this is not always the case. In this paper, we present a novel semiparametric approach to estimate the participation probabilities of units on a non-probability sample while accounting for error-prone auxiliary information in the reference data source. The approach combines a logistic regression model for the participation probabilities with a nonparametric kernel regression model for the error-prone auxiliary variables. Through a simulation study and real-world data analysis we show that our proposed approach effectively mitigates selection bias and accounts for measurement error, while estimates from methods which ignore the measurement error still exhibit some bias. The consistency of the Inverse Probability Weighted estimator produced using the participation probabilities derived from our approach is demonstrated, and a bootstrap variance approach for the estimator is described.
Read the full abstract

All authors Lyndon Ang, Robert Clark, Bronwyn Loong, Anders Holmberg

Computation and Language

SWEET: Mitigating Spurious Feature Reliance in LLM Reasoning via Weight Editing

Reasoning in large language models can be disrupted by simple input perturbations (e.g. adding irrelevant distractors), suggesting reliance on spurious correlations. Identifying and suppressing the spurious features involved in long chain-of-thought reasoning remains challenging. We introduce Spurious-feature WEight EdiTing (SWEET), a training-free method that combines counterfactual interventions with mechanistic interpretability to identify and suppress spurious-features. SWEET extracts token- and sequence-level behavioral directions from activation contrasts between original problems and their counterfactual variants - capturing local and global patterns of spurious reasoning. It then localizes the responsible MLP neurons and attention heads and suppresses their contributions along these directions through weight edits. Using only 20 original-counterfactual pairs for direction extraction, SWEET improves accuracy by an average of 11.8 percentage points on perturbed inputs and 2 points on original inputs across five open-weight reasoning models and five reasoning benchmarks. It also reduces the relative gap between original and perturbed accuracies from 17.6% to 2.5%. Analyses identify two recurring forms of spurious reasoning, i.e. over-integrating irrelevant conditions and overthinking, and provide mechanistic evidence that SWEET improves reasoning robustness by suppressing both behavioral directions.
Read the full abstract

All authors Zhizhang Fu, Lucas Zhao, Michalis Korakakis, Mengru Ding, Lukasz Szpruch, Yue Zhang

Sound

Patchbay: a block model for the sonification of ADS-B aircraft data

Patchbay is a design for turning the data that an ADS-B receiver collects into sound by wiring typed blocks together. ADS-B reports are the position and altitude messages that aircraft broadcast and that a small ground receiver can collect. We first describe what such a receiver reports: signals that differ in shape (continuous, stepped, circular, categorical), in resolution, in who produces them, and in how long they stay put, which runs from about ten seconds to the whole time an aircraft is in view. From this we derive a small model. Four kinds of connection carry an aircraft set, a value, an event or audio, and 27 blocks select aircraft, read fields, decide when a note starts, shape values and make sound. One of these blocks makes the receiver act as a radar: scanning by bearing it is a line that rotates about the receiver, and scanning by range it is a ring that grows from the centre or shrinks inward. Each aircraft sounds at the moment the scan reaches it, and the scan's period is set in bars of a beat clock. The model does not depend on a browser. Our reference implementation is a browser editor, contained in one HTML file, that replays two recorded captures and records both an audio take and a MIDI take, a file of the notes played rather than of the sound. We say what is not built and place the work beside the closest systems we found, among them a radar-sweep sequencer for tangible objects from 2006 and a range-ring sonification of ADS-B aircraft published in 2026. Sweeps that trigger sound, and typed connections, both exist in earlier systems. To our knowledge, none of the sources we reached carries the aircraft in view, with every field attached, as one kind of connection in an editor.
Read the full abstract

Machine Learning

JUBILEE: A Benchmark Environment for Off-Policy Evaluation in Debt Collections

Off-policy evaluation (OPE) estimates the value a candidate decision policy would achieve, using only data logged by the policy already in production. In debt collections, unrecorded factors can influence both actions and outcomes. This paper presents JUBILEE, a benchmark environment built around a debt-collections simulator that generates million-loan panels with competing risks, rare recovery events, and hidden factors driving both outcomes and assignment. The evaluation includes an 80-run comparison of tabular and gradient-boosted estimators under two synthetic outcome mechanisms and a 190-run parameter grid. Independent policy simulations provide reference values with estimated Monte Carlo uncertainty for a synthetic reward score and for simulated cash. With hidden factors influencing action assignment, several tabular methods agree closely while underestimating the field-visit policy's score, an error that also appears with boosted models. Intervals from tabular doubly robust estimation usually contain the reference without hidden assignment dependence but mostly miss it when that dependence is present. Estimates from randomized decisions do not show the same large error, although their limited precision leaves policy choice uncertain. The decision consequences depend on how policy value is defined. The field-visit policy ranks last under the score and first under simulated cash, which excludes channel costs. Under cash, log-based estimators miss this best candidate in nearly every run with hidden assignment dependence in the paired study. JUBILEE supports reproducible study of these failure modes under specified collections mechanisms.
Read the full abstract

All authors Diego Silva Prado, Tiago Amaral Penteado de Jesus, Bruno Caldas de Souza, Cristiano Bevitori Maffia De Oliveira

Computation and Language

When Safety Becomes An Outlier: Understanding the Retention of LLM Safety Behaviors

Safety-aligned large language models often lose their safety behaviors after fine-tuning, even when safety data are included. While this fragility is well documented, its underlying cause remains unclear. In this paper, we propose that formulaic refusal strings can function as outlier modes: narrow, weakly grounded response patterns that are easy to fit but vulnerable to subsequent updates. Controlled experiments support this account: fixed refusals are acquired much like arbitrary constant strings, in contrast to prompt-grounded natural responses. Motivated by these findings, we introduce Naturalness-Enhanced Safety Targets (NEST), a safety alignment framework that improves safety retention by constructing natural, context-aware safety targets that keep safety responses within the model's existing output distribution and grounded in semantics. We instantiate NEST in both supervised and preference-based alignment settings, using perplexity under the original model as a proxy for distributional support. Across multiple models and alignment regimes, natural safety targets improve both immediate safety and its retention under continued fine-tuning without sacrificing helpfulness. These findings suggest a practical principle for more durable safety alignment: construct safety targets that draw on the model's existing language capabilities and remain grounded in the meaning of each request.
Read the full abstract

All authors Binchi Zhang, Hadi Abdullah, Yiwei Cai, Jundong Li

Atomic Physics

A versatile beamline for attosecond photoelectron and photon spectroscopy

We report on the development of a versatile attosecond beamline for ultrafast spectroscopy of atoms and molecules, as well as for the metrology of infrared laser pulses. Using a few-cycle pulse (approximately 5 fs) for driving the high-order harmonic generation process, we show that the modulation of the harmonic flux in the cut-off spectral region, induced by a weak perturbing field, enables complete electric-field characterization of the perturbing pulse. Phase-sensitive detection with a lock-in amplifier further reveals polarization-dependent effects, observed as modulations at twice the fundamental optical frequency, emphasizing the sub-laser-cycle electronic control inherent to the measurement technique. In addition, the beamline supports the implementation of two-color (extreme ultraviolet and infrared) phase-sensitive experiments using approximately 30 fs pulses and a velocity map imaging spectrometer (VMI). This configuration will enable the study of photoemission time delays in atomic and molecular systems. Together, these results establish the beamline as a robust and adaptable platform for probing attosecond dynamics across diverse experimental approaches.
Read the full abstract

All authors Ronak Narendra Shah, Jahanzeb Muhammad, Ianina Kosse, Marvin Schmoll, Samuel Bengtsson, Riccardo Mori, Dominik Hoff, Gerhard G. Paulus, Fabio Frassetto, Luca Poletto, Giuseppe Sansone

Populations and Evolution

Rule-table structure of life-like cellular automata: activity, asymmetry, sensitivity, and the propagation storage factorisation

What structure on the rule space of a cellular-automaton family exposes which rules can support which behaviours? For the 262,144 life-like rules – the two-dimensional outer-totalistic family containing Conway's Game of Life – the standard coordinate, Langton's activity parameter, is long known to be too coarse. I organise the family by three symmetry-derived rule-table coordinates: an unweighted activity, a birth-survival asymmetry (odd under the swap of birth and survival), and the Vichniac-Binder Boolean-derivative sensitivity in closed form, whose centre-bit component vanishes exactly on the centre-blind (B=S) submanifold. The cardinality coordinates lambda, delta are necessary but never sufficient to certify behaviour; the centre-bit sensitivity labels the B=S submanifold on which the paper's one exact exclusion holds. Combined with an exhaustive enumeration of every rule's compact gliders, still lifes and oscillators (three-by-three seed, period at most sixteen), they expose a hierarchical factorisation: the birth condition is a narrow gate deciding whether anything can propagate, the survival condition tunes storage, and every compact-glider-supporting rule is birth-light and survival-rich. One exclusion is exact: no B=S rule supports a compact glider at that footprint, proven by exhaustive enumeration and closing only for compact-footprint patterns – 36 B=S rules ship at four-by-four seed and 40 at five-by-five. Two dynamical results frame the whole: glider-capable rules that random sampling never expresses show that capability is certifiable from the table while its expression is only estimable from dynamics; and every uncatalogued rule producing a richly Life-like ecology turns out to be a Conway or HighLife variant on the neighbour counts its trajectories actually visit.
Read the full abstract

Differential Geometry

Taylor Formulas for H-structures

We give a uniform second-order Taylor formula for tensorial Riemannian $H$-structures in Riemannian normal coordinates whose initial frame is $H$-adapted. For a defining multi-tensor $\xi$ with intrinsic torsion $T$, the linear coefficient is $T_q\diamond\xi$, and the quadratic coefficient features the covariant derivative of $T$, its iterated diamond action, and curvature. The formula applies componentwise to tensors of arbitrary type and recovers the known $G_2$ and $\operatorname{Spin}(7)$ expansions. We also give the almost Hermitian specialisation and prove that vanishing of the complete second-order torsion contribution forces the intrinsic torsion to vanish at the centre of the coordinates.
Read the full abstract

Robotics

BAT-WAM: Bridging Action and Touch in World-Action Models for Dexterous Hands

Dexterous manipulation with world action models (WAMs) relies on hand configuration and contact information, yet joint-state and whole-hand tactile observations may be partially or entirely missing. RGB images provide scene context and hand-object geometry but are insufficient to reliably recover these missing modalities. We further observe a correlation between hand state and tactile information. Motivated by this observation, we introduce BAT-WAM, a framework that bridges visual, joint-state, and tactile representations for WAMs. BAT-WAM learns separate codebooks and decoders for the tactile and action streams. An RGB-conditioned continuous latent bridge connects the two streams for bidirectional learning, while visual information serves as continuous context. Under the HTBench evaluation protocol, BAT-WAM reduces the root mean squared error (RMSE) of tactile reconstruction by 34.73% under fully masked conditions and 38.84% under partially masked conditions, relative to an image-conditioned HandTouch checkpoint. For future tactile prediction conditioned on target-time RGB and historical actions, it reduces mean RMSE by 85.56% relative to OpenWAM with HandTouch. Real-robot experiments further show that BAT-WAM consistently outperforms related WAM baselines in dexterous manipulation. Generalization experiments on unseen objects and scenes also demonstrate substantial improvements over baseline models. These findings support joint modeling of visual context, hand state, and touch as a representation approach for multimodal WAMs.
Read the full abstract

All authors Dong Fang, Feixuan Huang, Yuanxin Zhong, Yuxiang Gao, Qi Song, Yunlong Wang, Xiaosong Jia, Yu-Gang Jiang

Cryptography and Security

PQC-Corpus: A Dataset of Linked Development Histories for Post-Quantum Cryptography

As post-quantum cryptography (PQC) standards evolve, developers must adapt implementations, tests, and protocols. Studying this work requires connecting code changes to problem reports and developer reviews. We present PQC-Corpus, a dataset of linked development histories from 179 public GitHub repositories, spanning cryptographic libraries and general-purpose projects. It connects issues and closing pull requests to commits and full patches, with comments and reviews. We combine Git histories with GitHub threads, reconstruct 23 pull-request commit lists truncated at 250 entries, and check file integrity, schemas, and links. Two annotators independently assessed sampled commits and issues ($\kappa=0.86$ and $0.93$), labeling about seven in ten commits and two in three issues PQC-related within these samples. We provide full and filtered views with removal ledgers, and quantify how filtering reduces links from issues to commits. A patch-free view supports inspection before downloading patches, while traversal scripts follow issues to code changes. We make the dataset and collector publicly available for studies of specification changes, hybrid interoperability, cryptographic review, and migration assistance.
Read the full abstract

Machine Learning (Statistics)

Bayesian transfer for offline Q-learning with multi-task Gaussian processes

Offline reinforcement learning can benefit from data collected in related tasks, yet indiscriminate knowledge transfer may lead to negative transfer under task mismatch and limited target-data coverage. We propose a target-aware Bayesian transfer framework for offline Q-learning based on multi-task Gaussian processes that jointly model task-specific Q-functions. Source tasks are assessed using target predictive evidence and maximum mean discrepancy (MMD), while posterior uncertainty and behavior-supported action sets guide conservative Bellman updates and policy extraction. To account for the dependence induced by Bellman bootstrapping, we distinguish Bellman innovations from bootstrap errors and establish stability and predictive source-selection results under explicit regularity conditions. We further derive selection-adaptive posterior contraction and offline policy-performance bounds, identifying the structural conditions under which source information can be pooled to achieve minimax-optimal rates in a shared-function Gaussian sequence experiment. Simulation studies with heterogeneous source tasks demonstrate improved target Q-function estimation and reduced material negative transfer compared with indiscriminate source pooling. An application to human decision-making data further shows higher average expected immediate payoff than both target-only learning and matched-size random source selection. These results highlight the benefits of combining target-aware source selection with uncertainty-aware conservative learning for reliable and sample-efficient offline reinforcement learning.
Read the full abstract

Pattern Formation and Solitons

Flat-Top Solitons as Acoustic Resonant Cavities

A two-dimensional flat-top soliton forms an acoustic resonator without external confinement. Its nearly uniform interior carries sound, while its moving edge selects a discrete sequence of radial wavelengths. For the cubic–quintic nonlinear Schrödinger equation, we connect this spectrum to bulk compressibility, radius, and line tension. Numerical simulations confirm the predicted resonances below the particle-emission threshold. The soliton itself provides the sound-carrying medium and the moving boundary, allowing its resonance frequencies to be predicted from its stationary properties.
Read the full abstract