Peer Revenue Past citations are no guarantee of future returns

Papers

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

All of arXiv

Monday 14 September

General Mathematics

Admissible qx+1 Sequences, Semiconvergents, and Rational Catalan Numbers

Let $q\geq3$ be odd and put $m_r=\lfloor r\log_2q\rfloor$. We count parity sequences with exactly $r$ entries equal to $1$ for which the coefficient in the $qx+1$ iteration first drops below $1$ at step $m_r+1$. Their number $a_q(r)$ also counts the corresponding residue classes modulo $2^{m_r+1}$. By classifying increasing integer sequences according to the first index at which they exceed a prescribed bound, we obtain a recurrence which, for $q=3$, establishes the five- or six-term block pattern conjectured in 2015, allowing a shorter final block. We establish the sharp inequalities \[ \frac1r\binom{m_r-1}{r-1}\leq a_q(r)\leq\frac1r\binom{m_r}{r-1} \] Equality in the lower (respectively upper) bound occurs exactly when $\{r\log_2q\}$ is smaller (respectively larger) than all preceding fractional parts $\{j\log_2q\}$, $1\leq j<r$. These indices are characterized by denominators of convergents and semiconvergents from below and above; at these indices $r\geq2$, the values $a_q(r)$ are rational Catalan numbers. We also characterize the dependence on $q$ and prove, for each fixed odd $q\geq3$, that $a_q(r)=\Theta_q(r^{-3/2}B_q^r)$, where $B_q=(\log_2q)^{\log_2q}/(\log_2q-1)^{\log_2q-1}$. For $q=3$, this proves Hikawa's conjectured polynomial factor for the number of finite binary words of given weight for which the length of every nonempty prefix is smaller than $\log_2 3$ times the number of its ones.
Read the full abstract

Computer Vision

Authority-Preserving Evaluation of Medical Vision-Language Assistants

Medical vision-language models can propose how urgently a skin lesion should be reviewed, but the local service retains authority to accept or replace that proposal under referral policy, capacity, and locally held patient context. Proposal quality and selected-action quality are therefore distinct evaluation targets, and benchmark evidence transfers between them only when local review preserves the expected action score. We introduce AuthEval, a logging and evaluation framework that records both actions, scores the selected action under declared local criteria, and, where feasible, scores the declined proposal under the same rule. It reports the resulting authority gap only when the record supports it. Because the gap is the product of the proposal-change rate and the mean score change on changed cases, that rate alone determines neither its magnitude nor its sign. On ISIC 2019, with MedGemma and simulated local review, two constraint regimes with similar change rates produced an optimistic image-equal gap under capacity ($+0.744$ simulator units) but no detectable gap under safety. The declared evaluation unit also mattered: under mixed constraints the gap reversed from $+0.374$ to $-0.206$ when weighting shifted from image to lesion-aware cluster. AuthEval thus clarifies whether a study's records support claims about the model, the workflow, or both.
Read the full abstract

All authors Flint Xiaofeng Fan, Cheston Tan, Yew-Soon Ong, Roger Wattenhofer

Disordered Systems and Neural Networks

Towards the spin-glass transition in finite dimensions via blue percolation

We prove blue percolation in the Chayes-Machta-Redner representation of the zero-field Edwards-Anderson spin glass in finite dimensions. For fixed symmetric iid couplings with $0<m=\mathbb{E}|J|<\infty$, the first blue-percolation inverse temperature $\beta_{\mathrm{b}}(d)$ satisfies $2dm\,\beta_{\mathrm{b}}(d)\to1$ as $d\to\infty$, in periodic joint limits. For symmetric $\pm1$ couplings, both overlap signs percolate at explicit temperatures in every dimension from 7 to 12. That proof reveals the overlap field, shows that it dominates a weakly coupled Ising field, and bounds a weighted second moment of open paths; in dimensions 7 to 9 the paths make short lateral excursions, and an interaction-matrix criterion controls the second moment. These proofs are computer-assisted: finitely many inequalities between rigorously bounded quantities are decided in exact rational arithmetic. For exponential-moment disorder, we also establish a percolating regime with exactly equal infinite-blue-sector densities and finite spin-glass susceptibility. In these periodic limits, persistent density imbalance would imply distinct spin-flip-related Gibbs states and ordinary spin-glass order: positive spatial-overlap variance, infinite spin-glass susceptibility and a cusp of the replica-coupling pressure, without requiring finite-blue cancellation. Proving imbalance remains open.
Read the full abstract

Disordered Systems and Neural Networks

A Cross-Entropy Audit (CEA) for Consciousness Tests Moved from Brain to Large Language Model

Brain tests of consciousness are often applied to large language models (LLMs). Here consciousness means access: whether information, such as a heard sound, becomes available for report and use. A brain recording is in microvolts and an LLM's internal activations are arbitrary numbers, so the two share no common measurement unit. A brain test of this kind compares competing statistical models of how responses vary over repeated trials. The test's decision names the best-fitting model. The decision is wrong when the test names two states for responses that vary smoothly, or the reverse. No method we reviewed measures how often a transferred test is wrong. We call such a method a transfer audit and propose one, the cross-entropy audit (CEA). First, our audit holds back part of the trials and fits each model to the remaining trials. It then scores how well each fitted model predicts the held-back responses. Comparing two models gives a score difference in nats per trial. That difference does not depend on whether the recording is in microvolts or in activations, so brain and machine are scored on one scale. Second, the audit writes down five choices that such tests make for brain data: the competing models, their fitting, what counts as one independent sample, what the uncertainty interval describes, and the decision rule. Each may need to change on a machine. Third, the audit runs the test on simulated data built to follow one model, so the right decision is known, and counts wrong decisions. In our example, the two models of a two-state test, carried over unchanged to an LLM, found two states in 989 of 12,000 smoothly varying simulated datasets. The audited version of the test found two states in none of them. It detected 599 of 600 two-state datasets. The audited version's uncertainty interval, however, was too narrow at six of the twelve smoothly varying settings and is being repaired.
Read the full abstract

2609.22300 · 14 Sep · 51 pages, 3 figures, 7 tables; Supplementary Materials (42 pages) as an ancillary file. v3: the article as submitted to Entropy; new title; the method is named the cross-entropy audit (CEA); a brain-side matched comparison and same-data comparisons with published two-state tests are added; claims narrowed. Data and code: https://doi.org/10.5281/zenodo.23513442

Robotics

When Does Test-Time Physical Diagnosis Pay?

When a robot faces unfamiliar physical conditions, a common approach is to collect evidence about what changed and adapt. For such diagnosis to improve behavior, six ordered empirical conditions must hold: a meaningful reference, identifiability of the physical condition, use of the acquired evidence, decision value, selection value over a fixed alternative, and safe realization. We test this chain in controlled and public environments. In our controlled environments the frozen policy chooses correctly on every decision, but which evidence it buys is fixed by how the panels assign decisions to evidence strata. On three mechanisms absent from training, the frozen decoder fails its registered transfer gate. On all 96 decisions of the transfer panel, its full-trace input lies far outside its training range, and on such inputs it returns one answer whatever the trace contains. All 15 of its errors fall in the 30 decisions from which cheaper evidence was removed by design, a split hidden by aggregate accuracy. None of three primary public tasks passes its registered qualification gates. Before attributing a failure to how a frozen model uses evidence, evaluation should therefore check whether its deployment inputs lie inside its training range.
Read the full abstract

General Mathematics

Kernel-Checked Frontier Certificates for Erdős Problem 414

Let $h(n) = n + \tau(n)$, where $\tau$ counts divisors. Erdős and Graham asked, after Spiro, whether the orbits of any two positive integers under $h$ eventually share a point (Problem 414 on Bloom's list). The problem is open, and a finite computation cannot close it; nothing about it has been checked by a proof kernel, and the statement in the formal-conjectures repository carries a sorry. We replay what can be certified through the Lean 4 kernel under an axiom gate (axioms exactly propext, Classical.choice, Quot.sound; no sorry; no native_decide). Three elementary facts recorded by Li, namely that coalescence is an equivalence relation, $\tau(n)$ is odd exactly for squares, and $\tau(n)$ is bounded by $2\sqrt{n}$ so that orbits skip no square annulus $[k^2, (k+1)^2)$, are formalized, and his frontier lemma is formalized in a window form: every orbit started below a level $N$ has a point within $\lceil 2\sqrt{N} \rceil$ below $N$. The frontier lemma becomes a certificate: the $\lceil 2\sqrt{N} \rceil$ values $\tau(m)$ just below $N$ and the orbit points above $N$ until the orbits through the crossing points have merged. Its soundness is a kernel-checked theorem, and the certified rungs state that every pair of positive integers below $N$ coalesces for $N = 10^5, 10^6, 10^7, 10^8$; the certificate at $10^8$ has 44,530 entries rather than $10^8$. Every ledger and every quoted count was produced by two programs sharing no code that agree by hash. We record what is known about a non-coalescing pair and measure the exit-set sizes Li bounds. Nothing here is a proof of the conjecture.
Read the full abstract

Number Theory

Iwasawa theory of class field towers

Let $p$ be an odd prime and let $K_\infty/K$ be a $\mathbb{Z}_p$-extension. For each finite layer $K_n$, let $G_n$ be the Galois group of its maximal unramified pro-$p$ extension. We study the augmentation and Zassenhaus filtrations of the groups $G_n$ as $n$ tends to infinity. Assume that the classical Iwasawa $\mu$-invariant is positive. If $d_n$ denotes the minimal number of generators of $G_n$, then there is an integer $\delta>0$ such that $d_n=\delta p^n+O(1)$, while the minimal number of relations is $O(p^n)$. For each fixed $m\geq2$, as $n\to\infty$, we show that the $m$-th augmentation quotient has the same leading asymptotic as that of a free pro-$p$ group on $d_n$ generators. We also obtain corresponding asymptotics for the Hilbert series, the exponential growth rate of the augmentation filtration, and the Zassenhaus quotients.
Read the full abstract

Combinatorics

Close Divisors of Typical Integers:The Ford–Green–Koukoulopoulos Conjecture

For an integer $k\geq2$, let $\alpha_k$ be the supremum of the real numbers $a$ for which almost every integer $n\geq2$ has divisors $d_1<\cdots<d_k\mid n$ satisfying $d_k\leq d_1\bigl(1+(\log n)^{-a}\bigr).$ Let $\mathcal A\subseteq\N$ be the logarithmic random set in which the events $m\in\mathcal A$ are mutually independent and $\Pp(m\in\mathcal A)=1/m$ for every $m\geq1$. For a finite set $B\subseteq\N$, write $\Sigma(B)=\sum_{b\in B}b, \Sigma(\varnothing)=0$ and $m(B)=\max_{s\in\Z}\#\{C\subseteq B\mid\Sigma(C)=s\},$ and define \[ \beta_k=\sup\left\{c<1\,\middle|\,\lim_{D\to\infty}\Pp\bigl(m(\mathcal A\cap(D^c,D])\geq k\bigr)=1\right\}. \] Ford, Green and Koukoulopoulos proved $\alpha_k\geq\beta_k/(1-\beta_k)$ and conjectured that equality holds for every fixed $k\geq2$. In this paper, we confirm their conjecture. More precisely, for every fixed $a>\beta_k/(1-\beta_k)$, almost every integer $n\geq2$ has no divisors $d_1<\cdots<d_k\mid n$ satisfying $d_k\leq d_1\bigl(1+(\log n)^{-a}\bigr)$. We also correct local errors in their paper [Invent. Math. 232 (2023), 1027–1160], concerning the finite-subflag reduction, the residual-sum count, the moment estimate and the lattice adjustment. These corrections preserve the entropy-threshold comparison used in our proof.
Read the full abstract

Computer Vision

On The Robustness-Resolution Tradeoff In Temporal Quantization Of Event Streams

Event pipelines often discretize asynchronous timestamps before learning. This step looks harmless, but its stability depends directly on temporal resolution. We study this dependence at the representation level. We first show that hard temporal binning is discontinuous: an arbitrarily small timestamp shift near a boundary can move unit event mass between bins. We then define a class of nonnegative, mass-preserving, resolution-faithful continuous encoders and prove that every encoder in this class has global L1 sensitivity at least 2/Delta, where Delta denotes bin width. Linear two-bin interpolation attains this limit. Local support and first-moment preservation also make it unique. Experiments on SHD, N-MNIST, and DVS128 Gesture support the analysis. Across uniform timestamp budgets, linear interpolation lowers mean representation drift by 47-72% while keeping clean accuracy nearly unchanged. On DVS Gesture, it produces zero prediction flips across all tested budgets and three seeds. On SHD, measured drift follows 1/Delta with R^2 = 0.992.
Read the full abstract

Signal Processing

Machine Learning for Underwater Optical Wireless Communication Systems: A Comprehensive Survey

Underwater Optical Wireless Communication (UOWC) has emerged as a promising technology for high-speed underwater data transmission, offering significantly higher bandwidth and lower latency compared to acoustic and Radio Frequency (RF) technologies. However, the underwater medium introduces severe impairments that degrade link performance and limit communication range. The growing complexity of these challenges has increased research interest in Machine Learning (ML) and Deep Learning (DL) approaches, which offer powerful tools for channel modeling, signal processing, and system adaptation in ways that conventional analytical methods struggle to achieve. This survey provides a comprehensive review of ML methods applied across the full UOWC system pipeline, covering channel modeling, transmitter design, receiver detection and equalization, link alignment, and emerging applications, including wireless power transfer, semantic communication, object detection, optical sensing, and localization. Finally, we discuss open challenges and future research directions to motivate further work in this rapidly growing field.
Read the full abstract

Computer Vision

Validating, Not Sampling: Region-Level Robustness of Vision-Language and Vision-Language-Action Models

Vision-language models (VLMs) and vision-language-action models (VLAs) are increasingly deployed in real-world applications. There, a small perturbation to the recorded camera image may change a decision significantly. However, existing benchmarks for these models only sample perturbations, which does not guarantee the absence of a failure in the untested region. We present the first robustness validation of six VLMs (drawn from the Gemma, InternVL, LLaVA, and Qwen families) and five VLAs (drawn from the GR00T, OpenVLA, and $\pi$ families) over entire continuous regions of photometric and geometric image perturbation: brightness shifts, camera rotations, and their composition. To this end, we build on the validation framework H$^2$V and introduce H$^2$V-M, a margin-aware convergence rule that makes validation affordable at the 32B parameter scale. We demonstrate that H$^2$V-M outperforms H$^2$V by an order of magnitude in model queries and that it finds counterexamples faster than random sampling while providing soundness guarantees. Our VLM and VLA robustness validation shows that robustness is mostly dependent on the perturbation type, rather than the model, and that VLMs are more robust to large camera rotations than VLAs. For VLAs, even perturbations as small as $\pm1^\circ$ can change the commanded action in many cases. We also show that robustness depends more on model family than on model size.
Read the full abstract

All authors Bogdan Aron, Christopher Brix, Benedikt Brückner, Yanghao Zhang, Panagiotis Kouvaros, Alessio Lomuscio

Networking

Toward Mission-Critical ISAC: Reliable Energy-Aware Coordination in UAV Swarms

Unmanned aerial vehicle (UAV) swarms deployed in mission-critical applications must simultaneously track a mobile aerial target and maintain reliable data links. However, active integrated sensing and communication (ISAC) operation imposes a dual energy burden on propulsion and transmission, threatening mission continuity through premature battery depletion. In this paper, we propose a two-tier UAV swarm architecture in which mission UAVs (MUAVs) execute cooperative ISAC for mobile aerial target tracking while dedicated charging UAVs (CUAVs), equipped with solar harvesting panels, replenish low-battery MUAVs via aerial UAV-to-UAV wireless power transfer (WPT). We formulate the joint minimization of the cooperative posterior Cramér-Rao bound (PCRB) over MUAV trajectories, per-slot sensing-communication time splits, WPT scheduling and admission, and CUAV rendezvous trajectories, subject to minimum uplink rate, dual-tier energy causality, WPT proximity, collision-avoidance, and speed constraints, yielding a non-convex mixed-integer program (MIP) that, to the best of our knowledge, is the first to jointly couple cooperative ISAC sensing quality with aerial WPT and dual-tier energy management. To solve it efficiently, we propose Receding-Horizon Alternating Optimization (RHAO), a four-block per-slot algorithm that decomposes the problem into: charging admission via the Hungarian algorithm, MUAV trajectory and time-split via successive convex approximation (SCA), CUAV rendezvous, and WPT power allocation, with monotone convergence guarantees. Simulation results demonstrate that RHAO reduces the mean PCRB by 16.8 times over a fixed-time-split baseline and 5.4 times over a static-trajectory scheme, while the aerial WPT subsystem sustains all MUAVs above the energy-critical threshold throughout the full mission horizon.
Read the full abstract

Computer Vision

Beyond the Survey: A Systematic Empirical Study of Detection and Association in Visual MOT

This paper presents a comprehensive experimental evaluation and detailed analysis of state-of-the-art multi-object tracking algorithms, with an emphasis on quantifying the individual contributions of detection and association components to overall tracking performance. Unlike existing surveys that primarily offer theoretical categorizations or taxonomies of tracking methods, our work adopts a rigorous experimental perspective grounded in publicly available implementations, providing practical guidance for researchers and practitioners in method selection and system design. We introduce a unified pipeline diagram that consolidates the core components across the two main branches of visual multi-object tracking: tracking-by-detection and end-to-end deep learning paradigms, and systematically analyze the object detection, feature extraction, and data association modules. Through extensive empirical studies on standard benchmarks, including MOT16, MOT17, MOT20, SportsMOT, DanceTrack, and CrowdTrack datasets, we reveal critical insights: (1) detection quality dominates association strategy performance, with detector improvements yielding more than 10% gains compared to less than 5% from refined association strategies; (2) modern deep learning detectors paired with specialized re-identification models significantly outperform joint detection and embedding approaches; and (3) transformer-based end-to-end methods exhibit greater robustness to detection quality variations but at a substantial computational cost. Our findings from extensive experiments provide key insights into component-level effects in MOT, particularly the dominant influence of detection quality relative to association, while offering practical insights for designing and optimizing MOT systems under varying performance and robustness requirements. Code and experimental setups are available at github.com/linh-gist/VisualMOT.
Read the full abstract

All authors Linh Van Ma, Juhua Hu, Wei Cheng, Unse Fatima, Moongu Jeon

Artificial Intelligence

AgentPProf: Semantic Profiler for Long Horizon AI Agents

AI agents increasingly orchestrate long-running activities with users, tools, and system resources for days and weeks. To improve agent quality, safety, and cost efficiency, developers need to determine where failures happen, what triggers unsafe effects, and which tasks consume the most budget, then optimize those tasks. In systems software, profiling answers similar questions by aggregating resource consumption and attributing it to responsible code paths to identify hotspots. Yet existing agent observability tools focus on per-execution debugging and tracing rather than cross-run, long term profiling, making these questions difficult to answer at scale. Agent observability needs profiling, not only debugging, but profiling agents is challenging: the responsible entities are task intent like diagnose authentication, compare branches rather than code paths, and lack stable identifiers for aggregation. We propose a semantic operation stack model that adapts profiling to agent trajectories. Uniform operations represent all activities, and operation stacks replace the runtime call stack, enabling hierarchical attribution at different granularities. We observe that an agent's task occupies a contiguous span and decomposes into subtasks, so we introduce recursive operation segmentation, which recursively splits trajectories at task boundaries. AgentPProf is a profiler that aggregates agent trajectories into pprof-compatible profiles, enabling flame graph visualization and analysis. AgentPProf reaches 0.764 $B^3$ F1 against human annotations on CodeTraceBench. On three problem-localization benchmarks, the profile raises MAP by up to 56%, demonstrating that it effectively attributes resources, locates problems, and helps optimize token cost at practical profiling cost. AgentPProf is available at https://github.com/eunomia-bpf/agentsight.
Read the full abstract

All authors Yusheng Zheng, Chaokun Chang, Yu Mao, Tianyuan Wu, Yuxi Huang, Tao Ma, Wenan Mao, Shuyi Cheng, Andi Quinn, Wei Wang

High Energy Physics (Phenomenology)

Testing Higgs-Coupled Minimal Dark Matter with Solar Neutrinos after the LZ High-Recoil Event

The LUX-ZEPLIN (LZ) Collaboration has reported a nuclear-recoil candidate at $248$ keV, motivating interpretations in terms of endothermic dark matter (DM). We investigate this signal in Higgs-coupled minimal dark matter (HC-MDM), where the same electroweak interaction responsible for the terrestrial recoil also induces DM capture in the Sun and a potentially observable high-energy neutrino signal. We independently determine the mass splittings required to reproduce the LZ event and perform an improved calculation of Solar capture and post-capture evolution. Our treatment includes natural isotope mixtures, exact inelastic kinematics, nuclear form factors, finite-temperature nuclear motion near kinematic closure, orbital cooling through loop-induced elastic scattering, and an explicit capture–annihilation equilibrium test. The resulting Solar predictions are numerically stable and robust against the astrophysical and nuclear variations considered. For the two thermal benchmarks within the released IceCube DM mass grid, $3_M2_D$ and $5_M4_D$, the predicted annihilation rates exceed the mass-matched $W^+W^-$ IceCube response by factors of approximately $2.2\times10^3$ and $91$, placing both standard-halo solutions under strong tension within the $W^+W^-$ response mapping used here. For the heavier representations, whose masses lie above the published $10$ TeV IceCube grid, we construct a response-based extrapolation to higher masses. The $22$ TeV $7_M6_D$ benchmark remains in strong tension with IceCube even under our conservative response-loss extrapolation, while the $48$ TeV $9_M8_D$ benchmark also remains above the extrapolated response, although much closer to the sensitivity boundary. The $82$ TeV $11_M10_D$ benchmark is response dependent, whereas the $130$ TeV $13_M12_D$ benchmark remains below the extrapolated IceCube sensitivity.
Read the full abstract

Formal Languages and Automata

The triple rendezvous time of a synchronizing automaton can be floor(4n/3)

For every $n\ge9$, we construct a strongly connected synchronizing automaton with two input letters and $n$ states whose shortest word merging some three distinct states has length $\lfloor4n/3\rfloor$. One letter is a permutation; the other has an image of size $n-1$. We give complete transition maps and an explicit word attaining the bound. For the lower bound, we assign an integer to each unordered pair of states and prove that applying either input letter can decrease it by at most one. For $n\not\equiv0\pmod3$, the same construction contains a specified state whose shortest merging word, with its partner chosen freely, has length $\lfloor4n/3\rfloor-1$.
Read the full abstract

Representation Theory

On the homological conjectures for Artin algebras

We prove that the Auslander–Reiten conjecture, the generalised Nakayama conjecture, the Nakayama conjecture, and each of the two Tachikawa conjectures are globally equivalent for Artin algebras over any fixed commutative artinian ring. For finite dimensional algebras over a fixed field, we also prove that the derived invariance of the Nakayama conjecture is equivalent to the truth of the Nakayama conjecture for all such algebras, answering a question raised by Chen and Xi.
Read the full abstract

2609.19172 · 14 Sep · 10 pages. Version 2: Title changed; Rene Marczinzik joins as co-author. New results include the global implication from Tachikawa's first conjecture to the Auslander–Reiten conjecture, and the equivalence of the Nakayama conjecture with the Chen–Xi conjecture on derived invariance of finiteness of dominant dimension. Comments welcome!

General Relativity and Quantum Cosmology

Schwarzschild solution from a Raychaudhuri-Huygens relation

Null-congruences encode the causal structure of spacetime, describing a phase space of radiation channels, whose expansion scalar $\theta$ satisfies the Raychaudhuri equation. In spherical symmetry, the expansion scalar of wave fronts of area $A=4\pi r^2$ satisfies $\theta = A^\prime/A = 2/r$. About a mass $M$, this carries an encoding area $A_E = \lambda \varphi \ell_p^2 = \lambda R_g r \le A$ in a UV-IR consistent coupling $\ell_p^2= G\hbar/c^3$ to its Compton phase $\varphi = (mc/\hbar)r \propto 1/\hbar$. Here, $\lambda = 8\pi $ is fixed by the Newtonian limit at large distances and the saturation limit $A_E=A$ at horizon about $M$ based on the non-local extension of the equivalence principle in Rindler spacetime, without using the Einstein equations. In spherical symmetry, this recovers the Schwarzschild metric from a geometric embedding of $p=A_E/A$, which is directly observable in gravitational lensing. The result shows a non-thermal origin of the scaling dimension of two of gravitational phase space, previously derived based on the second law of thermodynamics by Bekenstein (1973), with possible implications for cosmological spacetime.
Read the full abstract

Artificial Intelligence

Regularized Emphatic Temporal-Difference Learning: Stability under Constant Stepsizes

Emphatic temporal-difference learning (ETD) stabilizes the expected off-policy TD update and changes its projection geometry, but neither property determines constant-stepsize sampled dynamics. We construct an ergodic two-state counterexample in which the ETD mean map contracts while the sampled product has a positive top Lyapunov exponent. Regenerative-cycle analysis separates this sign from the infinite variance of the follow-on trace. We introduce regularized emphatic TD (RETD), a normalized first-order post-shock repair that leaves the trace and importance ratios unchanged, stores the emphatic TD signal in a leaky scalar state, and releases a delayed correction. RETD's raw equilibrium is an affine shift of the ETD equilibrium; single- and two-regularization readouts recover the ETD fixed point exactly. We prove almost-sure convergence for harmonic diminishing stepsizes and a conditional constant-stepsize moment-contraction result from a Markovian random-product bound. RETD has certified negative exponents on the two-state construction and one Baird point, whereas the positive Baird ETD sign remains numerical. Paired 10,000-run experiments validate both separations, fixed-point recovery, a nonmonotone stability region, and task dependence. RETD changes post-shock dynamics; it does not reduce the shared follow-on-trace variance.
Read the full abstract

All authors Xingguo Chen, Zhaohui Wu, Jinguo Ye, Chao Li, Shangdong Yang, Guang Yang, Skylar Liang, Wenhao Wang

K-Theory and Homology

On motivic cohomology of commutative C*-algebras

Consider the ring of complex-valued continuous functions on a compact Hausdorff space $X$. Cortiñas–Thom computed its nonpositive algebraic $K$-theory, confirming Rosenberg's expectation. We give a motivic refinement of their result: For $i\geq 2j\geq 0$, the $i$th integral motivic cohomology group of weight $j$ is naturally isomorphic to the $i$th Betti cohomology group of $X$.
Read the full abstract

Genomics

Democratizing Clinical Tumor Whole Genome Sequencing: 18-hour End-to-end Analysis via Trillion-parameter Large Language Models Locally Deployed on Consumer-grade Hardware

Whole genome sequencing (WGS) is essential for precision oncology, yet its clinical adoption remains limited by prohibitive computational costs and multi-day turnaround times. This work presents a fully localized low-resource framework enabling stable deployment of a trillion-parameter biomedical LLM on a single consumer-grade RTX 4060 laptop with 32GB system memory and 8GB VRAM, as well as on routine clinical workstations in general hospitals, completing the entire tumor-paired WGS workflow from raw FASTQ input to clinical-grade full-variation-spectrum report output. Under standard 30X depth configurations, our implementation finishes a single tumor-paired WGS analysis within 18 hours, achieving 99.62% F1 score for somatic variant detection with over 99.9% concordance to the industrial-standard A100 cluster pipeline, fully meeting clinical oncology accuracy requirements. Quantitative profiling shows adaptive heterogeneous memory scheduling accounts for 71% of total execution time, while model optimization introduces less than 9% of total detection error. This work is the first engineering implementation of trillion-parameter biomedical LLM-driven clinical-grade genomic analysis on consumer-grade hardware, breaking the industry paradigm that trillion-scale genomic LLMs require hundred-thousand-dollar GPU clusters and multi-day turnaround, establishing a low-resource pathway for global primary medical institutions to adopt whole-genome precision oncology at zero additional cost.
Read the full abstract

Machine Learning (Statistics)

Stability-Constrained Approximation in Spline KANs: Exact Layer Balancing and Budget-Compatible Saturation

Deep spline superposition networks face a tension between approximation order and stability across depth. We study approximation under a hard layerwise Lipschitz budget, and organise it around two quantities: the factorisation stability complexity of a given deep factorisation, and the budget-compatible approximation complexity of a discretisation operator. First, we solve exactly the finite-depth diagonal balancing problem for a fixed chain of nonnegative envelope matrices: the optimal uniform layer budget equals $\|M_{L-1}\cdots M_0\|_{\infty\to\infty}^{1/L}$, attained by an explicit one-pass minimiser, for rectangular layers, with a complete treatment of degeneracies and non-attainment. The optimum can be arbitrarily larger than the Lipschitz constant of the network itself, because passing to envelopes destroys sign cancellation. Second, we give a constructive spline discretisation theorem preserving the budget up to a controlled slack, with an explicit grid threshold. Conversely, for linear spline-valued operators that preserve the budget exactly, we prove budget-compatible minimax lower bounds on classes constrained simultaneously in the first and third derivative norms – a constraint pair that is forced by the problem and that rules out the usual scaling escapes. Finally, we show that the corresponding layer errors need not cancel under composition: for every operator of the class there is a stable depth-$L$ tower realising a constant fraction of the accumulated error, so the linear-in-depth accumulation of the upper bound is not a proof artefact.
Read the full abstract

2609.17619 · 14 Sep · 33 pages. Ancillary code reproduces the numerical constants of Appendix B. Complements arXiv:2604.26444, which constructs deep KAN representations with controlled layer-wise Lipschitz product; here the allocation of that product across layers is solved exactly, and the cost of preserving it under spline discretisation is bounded from both sides

Cosmology and Nongalactic Astrophysics

FOSSIL: A future mission for CMB spectral distortion measurements

We present FOSSIL, a proposed space mission which is designed to deliver unprecedented absolute spectroscopic measurements of the sky intensity over a broad frequency range (50GHz to 2THz) to accurately measure chemical-potential mu-type, Compton y-type, and Compton relativistic y-type CMB spectral distortions. FOSSIL will reach monopole sensitivities representing a gain of roughly three orders of magnitude over COBE/FIRAS. These capabilities will allow us to constrain primordial density fluctuations on scales k~(10 to 10^4)Mpc^-1, test inflationary models, probe dark matter annihilation/decay/interactions, constrain primordial black hole seed scenarios, and pin down feedback models in galaxy evolution and structure formation by measuring the cosmic thermal energy content. Furthermore, its 130 spectral channels will enable legacy astrophysics, including precise characterisation of the Cosmic Infrared Background, line intensity mapping during cosmic noon, and detailed Galactic dust mapping.
Read the full abstract

All authors N. Aghanim, B. Maffei, J. Aumont, A. Beelen, B. Borgo, E. Bozzo, J. Chluba, X. Coulon, S. Couturier, F. Cuttaia, B. Cyr, P. De Bernardis, C. De Jabrun, J.-J. Diaz-Garcia, D. Eckert, G. Fabbian, L. Ferrari, F. Finelli, T. Gascard, J. E. Gudmundsson, A. Kogut, G. Lagache, M. Loquet Le Gall, S. Martin, A. I. Silva Martins, A. Monfardini, C. O'Sullivan, L. Pagano, D. Paoletti, G. Pisano, M. Remazeilles, V. Reveret, J. A. Rubino-Martin, V. Sauvage, G. Savini, L. Spencer, A. Tartari, L. Terenzi, N. Trappe, D. Watts, B. Winter, E. Baker, E. S. Battistelli, R. Battye, J. L. Bernal, A. Besnard, F. R. Bouchet, M. Braglia, C. Burigana, M. Calvo, S. Caminade, M. Citran, D. Clements, W. Coulton, E. Courau, G. De Zotti, E. Di Valentino, H. Dole, G. Domenech, S. Evangelista, S. Farrens, M. Frailis, K. A. Glasscock, V. Hervier, E. Kovetz, H. Liu, T. Louis, F. Madec, A. S. Maniyar, C. J. A. P. Martins, S. Masi, L. Maurin, F. Moll, L. Morelli, S. K. Naess, A. Nicola, F. Pace, N. Ponthieu, V. Poulin, E. Prele, N. Sanchez, W. Qin, L. Rodriguez, D. Scott, F. Simon, J. Singal, T. Slatyer, E. M. Teixeira, L. Thiele, M. Tristram, T. Trombetti, L. Vacher, S. Vinzl, A. Zacchei, E. Allys, K. Aryan, J. M. Casas Gonzales, O. Lahav, C. Sidiropoulos

Theoretical Economics

Reputation without a Control Group

An adviser who warns that a task is difficult may become harder to evaluate when her advice is followed more thoroughly. We study a long-lived adviser and successive short-lived workers who choose between standard and intensive implementation. Standard implementation reveals whether the warning was correct and gives the next worker an opportunity to acquire cost-saving practical knowledge. Intensive implementation protects the project but produces only occasional evidence about the adviser. We construct a stationary sequential equilibrium in which the adviser initially accepts an informative implementation, withholds the next project after her reputation improves, and resumes recommendations once inherited know-how has been lost. The interruption is chosen because it changes the successor's implementation decision: preserving know-how reverses the adviser's preference at the relevant history. All realized evidence remains public. Weak subsequent evidence eventually ends the low-ability adviser's protection, but this can take a long time.
Read the full abstract

General Relativity and Quantum Cosmology

A closer look at Viaggiu Holographic Dark Energy with Statefinder Hierarchy and Fractional Growth Diagnostics

We employ the statefinder hierarchy along with the fractional growth parameter to investigate the recently proposed noninteracting Viaggiu holographic dark energy (VHDE) model with the future event horizon as the infrared cut-off. By adjusting different values of the VHDE model parameter $\delta$, we illustrate the evolution trajectories of several important parameters, prticularly, the statefinder hierarchy $S_3^{(1)}$, $S_3^{(2)}$, $S_4^{(1)}$, $S_4^{(2)}$ and the fractional growth parameter $\epsilon$ against the redshift $z$ and the fractional energy density of matter, $\Omega_m$. Together with the growth rate of matter perturbations, we present the statefinder hierarchy that defines a composite null diagnostic (CND), which can differentiate dynamical dark energy from $\Lambda$CDM. The joint use of the statefinder hierarchy and the fractional growth parameter offers a powerful diagnostic tool for probing VHDE, effectively resolving the degeneracy between different parameter choices within the model. We find that employing a CND pair, rather than a single diagnostic, significantly enhances the diagnostic power for the VHDE model. Finally, we perform an $\omega_d-\omega_d'$ analysis and, in addition, assess the squared sound speed $v_{s}^{2}$ to determine the model's dynamical stability under small perturbations.
Read the full abstract

General Relativity and Quantum Cosmology

Prescription Dependence in Nonextensive Bose–Fermi State Counting near Black-Hole Horizons

Nonextensive Bose and Fermi state is well studied with counting near black-hole horizons in a one-function, nonextremal, static, spherically symmetric class. The calculation is organized around a $q$-deformed grand ensemble whose total partition function is a $q$-product of single-level partitions. This identifies the bosonic occupation-number sum as the primary level partition and allows us to separate two operations that are often conflated: replacing that sum by a geometric-series form and applying the ordinary reciprocal-log identity to a $q$-logarithm. A finite-energy tunneling prescription yields a leading near-horizon state-counting term controlled by the surface gravity and a horizon-response factor. Entropy matching then gives an energy scale proportional to the Hawking temperature; the geometry factor cancels from normalized nonextensive corrections and from the Bose–Fermi ratio. We derive the first-order coefficients analytically and evaluate the unexpanded $q$-integrals numerically. The linear approximation is accurate to better than one percent when $q$ lies within roughly two percent of unity. All prescriptions recover the extensive limit, but the exact bosonic level sum and the geometric approximations give opposite signs for the leading deformation of the conventional $7/8$ Bose–Fermi relation. The result isolates a robust geometry-factorized structure and quantifies a distinct $q$-algebra systematic uncertainty.
Read the full abstract

Data Structures and Algorithms

Low-Degree Polynomial Approximation of the Cross-Polytope

We determine the degree-distortion tradeoff for polynomial approximation of the $d$-dimensional cross-polytope $B_1^d$. For every $1\le t\le d$, every globally nonnegative degree-$2t$ form that is positive away from the origin has multiplicative sandwich distortion at least $(2e)^{-1/2}\sqrt{d/t}$, while an explicit sum-of-squares (SoS) form achieves distortion at most $(2e)^{1/2}\sqrt{d/t}$. Hence both the nonnegative-form and SoS optima are $\Theta(\sqrt{d/t})$, and degree $\Theta(d)$ is necessary and sufficient for constant distortion. The lower bound is representation-free. Averaging over signed permutations and evaluating on flat points of the $\ell_1$ sphere reduces every candidate to a support-size profile $V(k)=k^{-2t}Q(k)$ with $\deg Q\le t$ and $Q(0)=0$. After the substitution $u=1/k$, Lagrange interpolation shows that this degree budget cannot keep the profile nearly constant across $d$ support scales. A matching SoS construction averages even powers of sign-vector facet normals and reduces the upper bound to a Rademacher moment. We also isolate a weighted reciprocal-grid lemma, derive consequences for $\ell_p$ balls, and contrast the polar cube. For a general symmetric polytope, weighted facet powers yield a one-sided certificate whose boundary-floor objective is concave and whose worst-direction oracle reduces to convex dual-norm problems; at degree two, its optimizers recover classical optimal design and the John ellipsoid. This is an oracle-model certificate optimization, not an end-to-end complexity result or a characterization of the full SoS optimum.
Read the full abstract

Computer Vision

DualCount: Structurally Consistent Density and Point Modeling for Zero-Shot Object Counting

Zero-shot object counting aims to estimate the number of objects specified by a text query without category-specific training. Recent approaches primarily rely on density regression or detection-style instance prediction. While effective, density-based models often suffer from spatial ambiguity and background leakage due to weakly regulated mass allocation, leading to fragmented or part-biased representations that increase counting error in complex scenes. In this work, we propose an instance-aware dual-decoder framework that structurally couples density and point representations for zero-shot object counting. Instead of treating density estimation as independent pixel-wise regression, we interpret it as a structured mass allocation problem over a latent set of object instances. Predicted instance centers induce a soft instance-wise decomposition of the density map, upon which we enforce two geometric constraints: (1) per-instance mass conservation, ensuring each object contributes approximately one unit of density mass, and (2) center-of-mass alignment, encouraging each density component to concentrate around its corresponding predicted center. These constraints introduce instance-level geometric consistency and lead to more accurate mass allocation, thereby reducing counting error. Extensive experiments on FSC-147, PUCPR+, and CARPK show that our approach consistently reduces counting error and establishes new state-of-the-art performance in zero-shot object counting.
Read the full abstract

Sound

Timbre Analysis of the Hulusi, a Southwestern Chinese Free-Reed Instrument, using Machine Learning

The hulusi is a wind instrument that was invented in Yunnan Province, China, and has become tremendously popular in recent years. It consists of a mouthpiece, a gourd, and three bamboo tubes, all with free reeds made of copper. The main bamboo tube in the middle has seven finger holes. In this instrument, the pipe length, not the free reed's eigenfrequency, determines the instrument's pitch, unlike, for example, with the Western accordion or the blues harp. In this study, a machine learning model implemented in the COMSAR framework (https://github.com/ifsm) was used to investigate the timbre characteristics of the hulusi to cluster different instruments and pitches. The measured hulusi pitches C, B, A, G, and F were analyzed according to seven psychoacoustic features, among which only the spectral centroid, sharpness, and fractal correlation dimension are shown to form pitch clusters. These timbre features were used to train Kohonen self-organizing maps (SOMs) for clustering. Brightness and sharpness analysis revealed that the highest pitches were less bright and less sharp than mid- and low-range pitches were. Furthermore, the fractal correlation dimension, which mainly determines the chaoticity of the initial transients, was the best-clustering timbre feature for the hulusi, with the highest pitches showing the least chaoticity. This result is supported by defining a cluster quality index for the SOMs.
Read the full abstract

Algebraic Geometry

Transitive automorphism groups of maximal curves

Let $q$ be an odd prime power and let $X$ be a maximal curve of genus at least two over $\F_{q^2}$. We prove that the automorphism group of $X$ acts transitively on its rational points if and only if $X$ is the Hermitian curve, or $q=5$ and $X$ is the unique $\F_{25}$-maximal curve of genus $3$, namely the maximal Klein quartic. Together with the result of Giulietti and Korchmáros in characteristic two, this gives the classification for every prime power $q$. In particular, the Hermitian curve is the only maximal curve of genus at least two whose automorphism group is doubly transitive on its rational points.
Read the full abstract