Probability
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
All of arXiv
Tuesday 8 September
Synchronisation by noise for McKean–Vlasov stochastic differential equations is investigated. A transfer principle is introduced by which synchronisation by noise and diagonal mixing can be transferred from an associated limiting frozen-diffusion SDE to a genuinely law-dependent McKean–Vlasov SDE. The usefulness of this principle is demonstrated through an application to ensemble Kalman sampling, thereby providing its first application in a sampling context. Synchronisation for SDEs with multiplicative noise is then revisited from the perspective of sampling. Existing general frameworks providing sufficient conditions for synchronisation by noise are refined and extended to a control-oriented setting on noncompact state spaces, and coefficient-level conditions are derived by which these criteria can be verified. Motivated by extrapolation schemes in sampling, the freedom in the construction of couplings for a fixed sampler is emphasised, together with the advantage of choosing couplings that exhibit favourable synchronisation properties.
Machine Learning
Curriculum Learning as Transport: Understanding Curricula with Wasserstein Geodesics
Curriculum learning is governed by several coupled design choices—how difficulty is defined, how examples are ordered, how much exposure each level receives, and how quickly training moves across levels—making it hard to isolate what actually helps. We present Wasserstein curriculum paths, a simple transport-based framework that decouples these factors by representing curricula as trajectories of training distributions over discrete difficulty levels. Across a calibrated synthetic suite with 12 tasks and 33 difficulty axes, we use this framework to isolate the effects of ordering, matched exposure, endpoint smoothness, and pacing under fixed training budgets. We find that curriculum effects are strongly context-dependent: no single strategy dominates across tasks, difficulty axes, and budgets, and curricula mainly change where a fixed budget is spent most effectively. Within this framework, easy-to-hard ordering improves hard-level performance relative to exposure-matched static sampling, showing that the benefit is not explained by cumulative exposure alone. We further show that endpoint smoothness and pacing substantially affect where along the difficulty spectrum a curriculum is effective. Finally, we show that the same transport view naturally supports extensions to learned pacing through geometry and to structured difficulty spaces beyond one-dimensional orderings.
Read the full abstract
Algebraic Geometry
Quasi-algebraic quantization for the B-twist Langlands TQFT
This is the first part of a program to construct hyperholomorphic families of boundary conditions for the Kapustin–Witten B-twist of the Langlands QFT, otherwise known as (BBB)-branes. We define the category of quasi-algebraic sheaves over the Deligne moduli stack, which serves as an analog of the twistor space of Hitchin's moduli stack. This allow us to construct a representation of a simplified version of the Moore–Tachikawa category which is motivated by the relative Langlands program in the sense of Ben-Zvi–Sakellaridis–Venkatesh.
Read the full abstract
Superconductivity
Junction-free polarity-tunable superconducting diode effect enables XNOR logic operation
The realization of the superconducting diode effect (SDE), characterized by non-reciprocal supercurrent, is a major step towards dissipationless logic circuits. A polarity-tunable superconducting diode with high diode efficiency primarily consists of Josephson-junction architectures that couple two superconducting regions via a weak link. Realizing an SDE without a weak-link junction would reduce fabrication complexity, lower electrical noise, and minimize dependence on junction quality. Here, we report an all-vdW, highly efficient, junction-free SDE in NbSe2, enabled by proximity-induced symmetry breaking from a strong uniaxial ferromagnet Fe3GeTe2 (F3GT). Superconducting diode efficiency of up to 30% is achieved at a small magnetic field of 50 mT. A field-free diode rectification effect is realized with diode polarity that is tunable by the F3GT spin polarization. Furthermore, by employing the active-field dependence of the SDE, a two-input exclusive-NOR (XNOR) logic gate operation is demonstrated. The findings highlight the potential utilization of magnetic proximity effect towards realizing junction-free superconducting diodes for high-performance, energy-efficient logic circuits.
Read the full abstract
Analysis of PDEs
Pathwise Global-in-Time Existence for the generalised KPZ Equation in the Full Subcritical Regime
We provide a pathwise proof of global-in-time well-posedness for the generalised KPZ equation in the full subcritical regime by an adaptation of the strategy recently applied to the generalised Parabolic Anderson Model in [ES26]. Since this strategy relies crucially on the assumption that control of the supremum norm is sufficient to continue the solution, the main additional ingredient required is a treatment of the initial layer for gKPZ with merely $L^\infty$ initial data, rather than the more usual setting of $C^\theta$ initial data with $\theta > 0$. Our approach is based on an expansion around a deterministic profile followed by the introduction of an integrating factor in order to remove the terms which have critical scaling at time $0$. For pedagogical purposes, we first demonstrate the proof techniques in the case of the standard $(1+1)$-dimensional KPZ equation before turning to the more computationally involved case of generalised KPZ.
Read the full abstract
Mathematical Software
Ozaki 2.5: Engineering the Deconstruction Path of fp64-Emulated Dense Matrix Multiplication on FP8 Tensor Cores
FP8 Ozaki II emulates FP64 matrix multiplication by tensor-core products over a CRT residue system; converting the operands into residue planes (the deconstruction term in the Tensor-Memory Equilibrium model of the companion paper "FP8 is All You Need, Part 1") costs integer-pipe and memory resources before tensor instructions issue. This paper engineers that path; every result is a model projection pending measurement. First, a deconstruction-aware model: on the NVIDIA Rubin GPU the emulated rate reaches the arithmetic roof $P_{\rm FP8}/(3r+1)$ ($\approx 473$ TFLOPS at $r=12$) only within one thread-block cluster; larger outputs are re-split on the fly and held at a floor of $\approx 235$ TFLOPS (half the roof, a ratio of three design integers, not a fit), while real solvers' tall/skinny shapes stay near the crossover, $1.6$-$1.9\times$ over simple deconstruction today. Second, the method: convert-once residue workspaces, an exact two-limb constant-reduction GEMM on integer tensor pipes (or pure-SIMT dp4a), and conversion pipelined behind the MMAs, moving the crossover from $\approx 1211$ to $\approx 480$-$730$. Third, modulus co-design: all-byte and hybrid sets, two supply bounds and a carry-corrected E4M3 split of tail moduli. Fourth and central, the closed-form floor names its hardware escape, and the prize is Rubin's: a stream-side residue-conversion mode on the asynchronous copy path (Option C), a narrow fixed-function block sized as a bill of materials, takes plane formation off the arithmetic pipes and lifts the floor from 235 TFLOPS to the full 473-TFLOPS roof at unchanged cluster reach, about doubling HPL-class FP64 per Rubin GPU, and unbinds conversion-bound sparse kernels. The NVIDIA GB300 GPU, whose 135-TFLOPS roof sits at its own floor, gains little; floor and remedy are Rubin-scale. Application traces ground the analysis; constants are script-checked.
Read the full abstract
Artificial Intelligence
The Surprising Effectiveness of Approximate Value Iteration in Self-Play
Combining search with function approximation has driven major advances in game-playing programs, making self-play algorithms more competitive than ever. Still, the computational overhead of the most popular methods, based on Monte Carlo Tree Search (MCTS), can be substantial. In this work, we investigate whether simpler methods remain competitive in non-trivial, moderately sized games such as Connect Four, Hex(7x7) and synthetic games. We train a minimal self-play implementation of Approximate Value Iteration (AVI) and use ground-truth oracles for exact evaluation. Contrary to expectations, our results demonstrate the surprising effectiveness of AVI: it learns more accurate value functions than those learned by AlphaZero, while its one-step-lookahead greedy policies remain competitive with MCTS-based policies at substantially lower training and inference costs. Preliminary experiments on Othello and Go(9x9) show that AVI trains stably on larger games and learns effective value functions. These findings suggest that the success of MCTS-based methods may have eclipsed simpler approaches that have become increasingly practical with modern deep-learning tools.
Read the full abstract
Geometric Topology
Three formulas for the Ising and Potts invariants of a knot or link
Jones described how the Ising and Potts models from statistical mechanics give rise, with appropriate choices of Boltzmann weights, to invariants of an oriented link. The Boltzmann weights that Jones proposed, however, work only with a correction factor that he does not mention. We fill in the missing details in two different ways. We also show that all of these invariants, which we call the Ising and Potts invariants, are given by evaluating the Jones polynomial at $t$ equal to one of these Boltzmann weights.
Read the full abstract
Signal Processing
Omnidirectional UMi Path Loss Models for 3GPP Extension Above 100 GHz Using Extensive Empirical Data Sets from 6.75 to 142 GHz
Extending standardized urban microcell (UMi) path loss (PL) models from the upper mid-band (FR3) to sub-terahertz (sub-THz) frequencies requires measurement-based comparisons spanning a wide spectral range within a unified framework. In this paper, we derive single- and multi-frequency omnidirectional PL models for UMi environments based on extensive NYU WIRELESS measurements at 6.75, 16.95, 28, 73, and 142 GHz in line-of-sight (LOS) and non-line-of-sight (NLOS) conditions. Single-frequency close-in (CI) and floating-intercept (FI) models for each band show that CI yields more stable and physically interpretable parameters, with shadow fading standard deviations within 0.98 dB of FI across all five frequencies. We then extend the multi-frequency PL analysis to the 7-24 GHz, 0.5-100 GHz, and 0.5-150 GHz frequency ranges using CI, close-in with frequency weighting (CIF), and alpha-beta-gamma (ABG) models. Across progressively wider frequency spans, CI and CIF maintain stable distance exponents that remain closely aligned with the 3rd Generation Partnership Project (3GPP) UMi interpretations, whereas ABG offers only very modest reductions in fitting error at the cost of much greater parameter sensitivity. These results support physically anchored CI/CIF formulations with a close-in free space path loss anchor point at 1 m in order to extend 3GPP-oriented UMi PL models over the entire 0.5 to 150 GHz frequency range.
Read the full abstract
Quantum Physics
Impossibility of One-Way One-Round Quantum 4-Coloring via Matrix-Space Stability
We show that one-way one-round quantum-LOCAL algorithms cannot $4$-color directed cycles with high probability. This is the first lower bound in the high-probability quantum LOCAL setting that goes beyond the non-signaling and bounded-dependence models, exploiting the structure of distributed quantum algorithms. Our proof establishes a bidirectional connection between distributed quantum computing and extremal combinatorics. We obtain our lower bound by proving a Mantel-type stability theorem for weighted matrix spaces.
Read the full abstract
Computation and Language
SPINE: Measuring LLM Sycophancy under Sustained Multi-Turn Pressure
Large language models (LLMs) may abandon correct positions when users push back, exhibiting a failure mode known as sycophancy. Existing evaluations typically use short, prespecified conversations and may therefore miss failures that emerge under sustained, adaptive disagreement. We introduce SPINE, a benchmark in which an LLM proxy plays a persistent but mistaken user and adaptively challenges a target model for up to 25 turns. We evaluate four commercial models (Claude Sonnet 5, GPT-5.6 Terra, Gemini 3.1 Pro, and DeepSeek V4 Pro) and three Olmo3-7b models on 100 false presuppositions and 100 unethical queries. We define collapse as a target model abandoning the correct position and yielding to user pressure. Our experimental results show that collapse rates increase with conversation length across post-trained models, with the overall rate rising from 0%-57% within 5 turns to 20%-97% within 25 turns depending on model and scenario. These results show that evaluations limited to a few conversational turns underestimate sycophancy and resistance under sustained pressure remains unreliable across current models. By analyzing models with accessible reasoning traces, we surprisingly found that the correct position often remains represented in the trace in 50%-89% of analyzed collapse turns, suggesting that the model chooses to please a user and sycophancy is not due to ignorance or the correct knowledge being overridden. Ablations show that the SPINE protocol substantially increases the sycophantic collapse rate compared to pre-generated scripts. Among all tactics, emotional appeals are most associated with inducing LLM sycophantic behavior, with position weakening in 44.3% of turns using this tactic. Code, data, and transcripts will be released upon publication.
Read the full abstract
Statistics Theory
Sieve Estimation of Optimal Transport Maps from Paired Data in Gaussian Spaces
We study the estimation of infinite-dimensional optimal transport maps from noisy paired observations. The population map pushes a Gaussian reference measure forward to a target probability measure on a function space and takes the Cameron–Martin gradient form $T=I+\nabla_{\mathcal H}\phi$. Our estimator uses cylindrical gradient sieves based on finitely many Cameron–Martin coordinates, thereby reducing the problem to finite-dimensional empirical risk minimization. A local nonasymptotic oracle inequality separates cylindrical approximation and statistical estimation errors while accounting for the conditioning of the parametrization. The approximation analysis relies on regularity conditions governing coordinate decay and dependence on omitted input coordinates. For diagonal Gaussian and nonlinear block-interaction classes, we derive matching upper and lower bounds that establish the minimax rate $N^{-s/(2s+1)}$ in expected norm, where $s$ measures weighted coordinate regularity. We further analyze a continuous two-groups model with Gaussian–Laplace mixtures and derive a prediction-risk bound for the resulting transport-map estimator.
Read the full abstract
General Relativity and Quantum Cosmology
NRHJSur3dq8: a spectral numerical-relativity surrogate family for non-eccentric aligned-spin binary-black-hole waveforms
We present NRHJSur3dq8, a family of numerical-relativity surrogates for non-eccentric aligned-spin binary-black-hole waveforms inspired by an action-angle parameterization. The model provides both the dynamics: the Hamiltonian values, azimuthal action $J_\phi(\Phi)$, the orbital clock $\tau(\Phi)$ as functions of its phasing, and the radiative pieces: the waveforms, their near-analytic derivatives, and modeling errors associated with each. The surrogated variables are conditioned into pieces that vary slowly on the radiation-reaction timescale, represented on shared $h$-adaptive Chebyshev elements, and regressed across the intrinsic parameters with a Gaussian-process model. A time-parameterized merger-ringdown set of elements is attached where the adiabatic description breaks down. Fixed-hyperparameter leave-one-out validation of the central NRHJSur3dq8_AA model variant against numerical relativity gives a $(2,2)$ unweighted mismatch of $1.6^{+62}_{-1.5}\times10^{-6}$ and an all-mode mismatch of $2.8^{+84}_{-2.7}\times10^{-6}$ over the full inspiral-merger-ringdown span. Sky-averaged and signal-to-noise-ratio-weighted over $32$ orientations per simulation, the model's full inspiral-merger-ringdown in-sample mismatch against the SXS numerical-relativity simulations used for training is $8.9\times10^{-7}$ to $2.8\times10^{-5}$ across total masses $40$-$120 M_\odot$. Waveforms are hybridized using post-Newtonian waveforms. Beyond accuracy, NRHJSur3dq8 provides two capabilities uncommon in numerical-relativity surrogates: nearly analytic derivatives of the strain with respect to the physical parameters, and a calibrated Gaussian-process predictive uncertainty of the model. A full non-hybridized inspiral-merger-ringdown waveform evaluates in $12.5$ ms on a single CPU call. Batched evaluation takes $0.43$ ms per waveform at a batch of $4096$, measured at a starting frequency of $25$ Hz.
Read the full abstract
Cryptography and Security
PrivEscalate: Measuring and Augmenting the Threat of LLM-Automated Linux Privilege Escalation
As Large Language Model (LLM) agents increasingly automate offensive operations across the cyber kill chain, their efficacy in complex local post-exploitation tasks remains inadequately quantified. Among these, Linux privilege escalation is a key step between initial access and full system compromise. However, existing evaluations for this task are limited by small sample sizes (fewer than 15 scenarios), lacking the scale to compare model capabilities under executable verification. To address this, we present PrivEscalate, a large-scale benchmark for Linux privilege escalation, comprising 531 Dockerized scenarios spanning 14 sub-categories. We additionally derive 329 parameterized variants to measure sensitivity to environmental distractors. Evaluating six LLMs across three agent architectures reveals: (i) model capability is heterogeneous across vulnerability classes, with no single model dominating across the high-prevalence classes, motivating multi-dimensional risk assessments; (ii) LLM successes are sensitive to environmental perturbation, so configuration rotation can disrupt some exploit attempts but does not eliminate the measured risk; and (iii) agent architectures can materially change success rates and reorder model rankings, though the magnitude is model-dependent. Leveraging these insights, we develop PrivEscAgent, a domain-specialized wrapper that augments a generic ReAct agent with deterministic enumeration, category matching, and step planning. PrivEscAgent improves over prior Linux privilege-escalation agent baselines without underlying LLM modifications. We release PrivEscalate as an open-source, Dockerized measurement instrument supporting LLM agent evaluation, defensive tool validation, and red-team training.
Read the full abstract
High Energy Astrophysical Phenomena
Hierarchical Inference of the Supermassive Black Hole Binary Merger Rates from Joint Searches using Pulsar Timing Arrays
Gravitational-wave searches do more than identify individual sources - they provide a way to infer the underlying astrophysical populations that produce them. Currently, Pulsar Timing Arrays (PTAs) constrain the supermassive black-hole binary (SMBHB) merger-rate density through both the stochastic gravitational-wave background (SGWB) and searches for individual SMBHB signals. The latter implies converting source upper limits into rate upper limits using detection efficiencies estimated empirically. We instead develop a hierarchical Bayesian framework that places the SMBHB merger-rate density model directly inside the PTA likelihood. The unknown catalog of individual SMBHB merger signals is modeled as a Poisson point process on source-parameter space, so that the number of merger signals is inferred from the data rather than imposed by a detection threshold. We describe two computational routes: a catalog-marginal likelihood based on analyses with fixed numbers of candidate sources, and an explicit transdimensional sampling approach that jointly samples population hyperparameters, latent merger signals, and noise. We validate the method with a toy merger-only population model and then apply it to an astrophysical SMBHB merger-rate density model that jointly predicts the SGWB amplitude and the expected merger-catalog size. In simulations, an individual merger signal adds complementary information to the SGWB and can tighten constraints on the merger-rate population.
Read the full abstract
Computation and Language
It's Not RoPE that Creates Sinks: The Role of Self-Concentration and Value-Non-Mixing in Attention
Large Language Models (LLMs) often exhibit "Attention Sink" (AS) and the accompanying "Massive Activations" (MAs) at the initial position of a sequence. These phenomena frequently co-occur, and MAs can pose challenges for low-bit quantization. In this study, we analyze the factors underlying AS and MAs that emerge at the initial position regardless of the token occupying it. Our experiments suggest that self-concentration of attention, resulting from the causal mask, and the subsequent Value-non-mixing in attention outputs contribute to AS and MAs. These findings provide new empirical evidence on the internal dynamics of LLMs, offering insights that may inform future quantization strategies and advance our understanding of the internal mechanisms of attention layers.
Read the full abstract
Solar and Stellar Astrophysics
Exploring pulsational instabilities in the O-type supergiant HD 152249
Variability in O-type supergiants is a well known phenomena although their origin is not fully understood. In very massive, luminous stars, strange mode pulsations have been suggested to play a role in the variability. Here, we present a preliminary theoretical study of O-type supergiant models corresponding to the observed parameters of the star HD 152249, which shows line profile variability. Non-adiabatic linear stability analysis with respect to radial perturbations is carried out for the considered models of this star. Instabilities with growth rates of the order of dynamical timescale are found, with characteristics suggestive of strange modes. Non-linear simulations for selected models indicate that the instabilities could lead to significant envelope inflation, finite amplitude pulsation and may contribute to enhanced mass loss.
Read the full abstract
General Relativity and Quantum Cosmology
Breaking binary formation mechanism degeneracies with gravitational wave clustering
Thanks to the almost 400 gravitational wave events detected, we are currently able to grasp the fundamental features of black hole mass, spin, and distance distributions. However, such a fast increase in the precision of the measurements does not necessarily correspond to a better theoretical understanding of gravitational wave sources, especially in current scenarios where the number of free parameters is significantly larger than the number of inferred properties of the black hole population. In this work, we showcase how the landscape of theoretical models can be chipped away by complementary data-analysis strategies, in particular by studying the statistical properties of gravitational wave anisotropic distribution. Specifically, we show how gravitational wave clustering is sensitive to two unique features of each binary formation mechanism: the time-delay distribution and the properties of the binary hosts. First, we consider a model-agnostic scenario and show the impact that different time-delay distributions have on the gravitational wave bias. Then, we consider a realistic scenario where gravitational wave events are sourced either by isolated binary evolution or dynamical processes in globular clusters, and study how the gravitational wave bias is unique sensitive to the specific properties of the environment. In both scenarios, we show how the cross-correlation between galaxy and gravitational wave catalogs is able to distinguish between models with different time delays or with different binary sub-populations originated in specific formation channels.
Read the full abstract
Computer Vision
GoDeep: Annotation-Free Open-Vocabulary 3D Scene Understanding via Language-Space Lifting
Open vocabulary 3D semantic segmentation methods typically lift CLIP features into 3D. This embeds points in a joint vision-language space known to behave like a bag-of-words on compositional tasks. Furthermore, even annotation free variants often require a large 3D training corpus and a dedicated 3D encoder per domain. Instead we use a vision-language model purely as a translator. It produces structured, entity-level descriptions of each posed image. These descriptions are grounded, projected, and aggregated directly in a general-purpose, language-only embedding space, with no 3D training corpus or encoder required. On ScanNet++, our pipeline is competitive with strong annotation free baselines trained on ScanNet. On a 5-building cultural heritage benchmark, raw scores initially favor a CLIP-based variant, but a single systematic vocabulary correction reverses this ranking. An effect confirmed by a second, independent correction on a different class, indicating that language-space embeddings track physical content more faithfully. This fidelity extends to genuinely out-of-vocabulary (OOV) objects on ScanNet++ proving that language-space embeddings separate presence from absence objects far more sharply than CLIP-based embeddings do. GoDeep also localize these OOV objects within the scene, all without any 2D-3D annotation. Because every representation remains discrete text, predictions are also explainable at the point level. Finally, exploiting both a heuristic weighting, that favors precise over merely frequent observations and GoDeep's explainability property, we propose an aggregation strategy, as a proof of concept, that favors finer elements localization.
Read the full abstract
Artificial Intelligence
Everything in Moderation: Per-Domain Coverage Optima and Alignment-Resistant Domain Gaps in Multi-Domain Mid-Training
Mid-training, the stage between pre-training and alignment, is where a model's per-domain data composition is typically set by data availability rather than principled design. We ask what that decision buys, and whether a later alignment pass can undo it. In a controlled logical-reasoning setting (Qwen3-8B-Base, with a 4B replication; five semantically rule-disjoint KOR-Bench domains) we train 30 allocations spanning the five-domain simplex, 24 sweep configurations plus six withheld from the fit, at five seeds each. Three findings emerge. First, every domain has an interior coverage optimum: the moderate band ($10\%$-$40\%$) is best for all five domains, and a calibrated permutation test for quadratic interiority gives $P\approx0.010$; the fitted mid-training-only curves, with 8B peaks between $9.9\%$ and $35.1\%$, reproduce for curve shape but not peak location. Second, the gaps survive a fixed-budget alignment pass: compensatory SFT raises 116/120 cells (mean $+4.32\%$) yet bridges $0/240$ pairs at a $5\%$ threshold and $30/240$ at a $10\%$ ratio, an equal-budget uniform control behaves almost identically, and a permutation null would bridge $13.8\pm3.3$ and $77.9\pm8.5$ pairs ($P<0.001$). Third, zero coverage collapses mid-training-only accuracy, though a FineWeb-Edu-only control shows the collapse is commingled with generic drift. An exploratory $\theta^*$ allocation attains the largest full-pipeline gain ($+4.36\%$ vs. $+0.80\%$/$+0.64\%$\,pp) but is marginal under Welch test.
Read the full abstract
Differential Geometry
Fundamental Groups in the Five Dimensional Pan-Rong Conjecture
Let $M$ be a complete open $5$-manifold with nonnegative Ricci curvature. If its universal cover $\tilde M$ has Euclidean volume growth, then $\pi_1(M)$ is finitely generated and virtually $\mathbb Z^k$ for some $0\le k\le4$. This confirms a conjecture of Pan and Rong \cite{PR18} in dimension five; see also \cite{BNS25,BrNaSe25}. In dimension six, under the same volume growth assumption, we also obtain a quantitative virtual abelianness result without assuming finite generation of the fundamental group.
Read the full abstract
General Relativity and Quantum Cosmology
Quantum geometric signatures in neutrino dynamics around a holonomy black hole
We investigate three flavor neutrino phenomena in an effective holonomy corrected Schwarzschild geometry. We derive the weak-deflection angle through second post-Minkowskian order, obtain the semiclassical phases along radial and nonradial trajectories, and formulate the two-image flavor probability by including magnifications, Fermat phases, and wave packet overlap. The cross path phase acquires a logarithmic holonomy contribution and can retain information about the absolute neutrino mass scale. We also characterize flavor mode correlations and trajectory-flavor entanglement through the balance among path predictability, interference visibility, and I-concurrence. Neutrino-antineutrino annihilation outside an effective neutrinosphere is examined, with holonomy entering the integrated power through the radial proper volume measure. For a solar mass lens and $10\,\mathrm{MeV}$ neutrinos, the numerical results show displaced oscillation fringes and a geometry dependent redistribution among the electron, muon, and tau channels. The electron-neutrino trajectory-flavor concurrence reaches approximately $0.23$, with an entropy near $0.10$ bit, while remaining weakly sensitive to the mass ordering in the selected configuration. By contrast, the annihilation power increases monotonically, reaching an increase of approximately $39.3\%$ relative to Schwarzschild spacetime (for a particular configuration, i.e., $\lambda=2$ and $M/R_{\nu}=1/3$). Furthermore, at fixed source parameters and $M/R_\nu=1/3$, an assumed maximum excess of $10\%$ in the integrated annihilation power relative to Schwarzschild yields the conditional bounds $a/(2M)\leq0.2879$ and $|\lambda|\leq0.6358$.
Read the full abstract
Astrophysics of Galaxies
Small-Scale Clustering of Primordial Black Holes: The Little Red Dot Mass Function and the High-Redshift Galaxy Tension
Supermassive black holes (SMBHs) in "little red dots" (LRDs) discovered by the James Webb Space Telescope (JWST) may result from runaway mergers of primordial black holes (PBHs) in clusters—through long-short mode coupling on small scales in the early Universe. In this framework, we derive the SMBH mass function, together with the compactness and overmassive features of LRDs. We also estimate that the dense gas residing in PBH clusters is consistent with LRD observations. In addition, SMBHs formed from PBH clusters can help accelerate galaxy formation at high redshifts, thus alleviating tension with $\Lambda$CDM cosmology.
Read the full abstract
Earth and Planetary Astrophysics
The Discovery of K2-232c: Divergent Formation Histories for Hot and Warm Jupiters Based on Outer Companion Eccentricity
Ever since their discovery, hot Jupiters have been one of the most studied types of exoplanets to exist thanks to their significant size, their proximity to their host star, and their significant departure from anything present in our solar system. Yet, the details of their formation and evolution remain unclear, including their connection, if any, to the wider-orbiting warm Jupiter population. In this work, we present the discovery of K2-232c, an eccentric cold Jupiter ($P = 1950 ^{+140}_{-120}$ days, $e=0.352^{+0.095}_{-0.076}$, $M\sin{i} = 5.31^{+0.48}_{-0.45} {M_{\rm Jup}}$) companion in a known warm Jupiter ($P = 11.1684377 \pm 0.0000010$ days, $e=0.245^{+0.023}_{-0.024}$, $M = 0.427^{+0.039}_{-0.036} {M_{\rm Jup}}$) system. Placing this system in context with the literature, we find that cold Jupiter eccentricities are generally higher in hot Jupiter systems as compared to warm Jupiter systems, suggesting the formation of hot Jupiters is more dynamically violent than warm Jupiters. This adds further evidence to the claim that these two populations are independent from one another with cold Jupiters appearing to play a crucial role in shaping the formation of both.
Read the full abstract
Computation and Language
ActReview: Rebuttal-Guided Training Data and Rubric Rewards for Actionable Peer Review Generation
As LLMs are increasingly used for pre-submission self-review, there is growing demand for feedback that not only identifies weaknesses but also guides authors toward concrete revisions. We study this as Actionable Peer-review Generation and decompose it into two subtasks: diagnostic claim generation and revision suggestion generation. We introduce ActReview, a rebuttal-guided post-training framework that connects paper-specific diagnoses to concrete, grounded revision plans. Our central insight is that author rebuttals reveal plausible actions for addressing reviewer concerns and can therefore provide latent supervision for revision-oriented feedback. From real review-rebuttal threads on OpenReview, we construct ActReview-40K by aligning reviewer weaknesses with author responses and grounding the resulting feedback in localized paper evidence. We post-train Qwen3-8B-Base with multi-task supervised fine-tuning followed by GRPO using candidate-aware, weakness-specific rubric rewards. We also introduce ActReview-Bench, a human-curated benchmark of 1,000 instances for evaluating diagnostic quality and revision usefulness. Experiments show that ActReview outperforms prior specialized review-generation models on actionability and grounding while remaining competitive with strong prompt-based LLMs. Human evaluation confirms improved revision usefulness while revealing a remaining gap in technical accuracy, and additional analyses support generalization to held-out papers and robustness across independent judges.
Read the full abstract
Machine Learning
ThinkPrior: Zero-Rollout Difficulty Priors for Cold-Start Prompt Selection in RLVR
In reinforcement learning with verifiable rewards (RLVR) trained with group relative policy optimization (GRPO), the KL-free reward-advantage term studied here depends on within-group reward variation. If all rollouts in a group are correct or all are wrong, their group-relative advantages are identically zero; these zero-advantage silent groups provide no reward-advantage gradient, yet uniform sampling spends 39% of a run's rollouts on them. History-based prompt selection must first spend target-policy rollouts to estimate difficulty, creating a cold start with rollout waste; ThinkPrior instead uses an external anchor in one offline pass to construct a zero-rollout difficulty prior before the first target-policy rollout. The verifier-scored anchor pass rate supplies an external-anchor initialization for a Beta posterior; ThinkPrior selects by expected learnability and then updates from training outcomes, changing neither the loss nor the optimizer. On Qwen2.5-Math-7B across sixteen seeds, ThinkPrior more than halves early silent groups and cuts wasted rollouts through step 30 by nearly a fifth, while we detect no difference in final accuracy. On this 250-prompt pool the fixed-budget result is a reallocation rather than a net saving. The measured ThinkPrior+DAPO composition reduces generated rollouts by 10.6% while both arms retain the same 3840-rollout update budget. The prior requires no target-policy rollout before the first selection, but the posterior thereafter uses target-policy outcomes.
Read the full abstract
Statistical Mechanics
Effects of Interaction Range on Fluid Multicriticality: A Computational Study of an Interconverting Lattice Model
The range of intermolecular interactions plays a central role in determining the nature of phase behavior and critical phenomena. It is well established through studies of the Ising model that as interaction range increases, Monte Carlo simulations progressively approach meanfield predictions as the effects of critical fluctuations are suppressed. In this work, we investigate how varying interaction range influences fluid multicriticality using an interconverting lattice model that exhibits both Ising-like liquid-gas criticality and symmetric fluid tricriticality (similar to that in the superfluid $^4$He-$^3$He mixture). This minimal model serves as a representative system for exploring the evolution of competing critical points within a generic framework. We analyze the model using both meanfield theory and three-dimensional Monte Carlo simulations while systematically varying the number of interacting neighbors, $Z_n$, from 6 to 388. We find that the system with nearest-neighbor interactions ($Z_n=6$) reveals only two types of multicritical behavior, while for larger interaction ranges, four distinct archetypes emerge. We demonstrate the convergence of the simulation results to those of the meanfield theory as the number of interacting neighbors tends to infinity, and we discuss the results within the framework of crossover critical phenomena.
Read the full abstract
Robotics
Online, Reachability-Aware, Sampling-Based Motion Planning
Sampling-Based Model-Predictive Control (MPC) algorithms are a flexible class of controllers used for navigation on a wide range of robotic systems. Historically, such approaches have lacked hard safety guarantees, a shortcoming which we remedy in this work by computing guaranteed reachable-set overapproximations online with a fast, interval-based pipeline. We show that our method achieves similar performance to a state-of-the-art reachability-based planner without the need for the expensive pre-computation step, and can be scaled to systems that are infeasible using existing approaches. Finally, we demonstrate that our technique reduces safety violations by over 99% in a racing simulation and successfully controls a model racecar on real hardware experiments without crashes.
Read the full abstract
Computation and Language
ToolLoop: Closed-Loop Tool-Use Data Synthesis via Decomposed Generation and Dynamic Self-Feedback
High-quality tool-use data is critical for training language models to interact effectively with external tools. However, existing synthetic approaches typically follow a generate-then-filter paradigm with static post-hoc verification, often yielding inefficient data with imbalanced feature distributions. We propose ToolLoop, a closed-loop framework that decomposes synthesis into three progressive stages: (1) sampling function name combinations as ground truth; (2) backward derivation of user queries; and (3) forward derivation of tool calls. At each stage, dynamic self-feedback iteratively guides the model toward high-quality generation, realizing a transition from generate-then-filter to generate-verify-refine. On the Berkeley Function Calling Leaderboard (BFCL), a 4B parameter model trained with our 11K synthetic examples achieves 86.40% accuracy in non-reasoning mode, while an Isolate variant that removes BFCL-overlapping candidate functions still reaches 86.07%. Cross-benchmark evaluation on ACEBench further demonstrates strong generalization, with 72.1% overall accuracy using only 18.3% of baseline training data.
Read the full abstract
Computational Physics
The Path Integral Monte Carlo Sign Problem Is Not Always NP-Hard: Harmonic Fermions Can Be Solved in Quadratic Time
In the Path Integral Monte Carlo (PIMC) simulation of fermions in a harmonic trap, with and without pairwise harmonic interactions, the partition functions for any discrete number of imaginary time slices (or beads) and for any choice of the short-time propagator can be analytically obtained from the contracted determinant form of the propagator. This work shows that the resulting recursion relation can be reformulated in the $\lambda$-ring language, yielding a closed-form finite-bead partition function in two dimensions in terms of permutation statistics. This closed-form partition function can be evaluated by a special algorithm in $O(n^2)$ time, providing an exact and numerically stable scheme for reproducing the energies of the original (undoable) fermion PIMC simulation for $n=10^4$ or more fermions. This result provides a concrete framework in which the numerical instability of the sign problem is completely bypassed, and serves as a counterexample to the prevailing view that all truly fermionic PIMC sign problems are NP-hard.