Cryptography and Security
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
All of arXiv
Monday 9 November
Web3 platforms that help users discover and trade newly launched crypto tokens, including launchpads, decentralized-exchange (DEX) interfaces, wallets, and token-analytics sites, must assess tokens before their legitimacy is clear. On Solana, where token creation is inexpensive and tokens use standard programs rather than custom per-token smart contracts, rug pulls can unfold within minutes. Effective warnings must therefore rely on limited histories. Although graph neural networks (GNNs) have improved rug-pull detection on EVM chains, it remains unclear whether transfer structure adds predictive value over non-graph models at minute-scale decision points, and how that value changes as history accumulates. We study 2,700 tokens sampled from SolRPDS, collecting one-minute market histories from Birdeye and token transfers from Solana RPC. We define a rug pull operationally by requiring a liquidity drain, price collapse, and persistent liquidity depletion over a nominal 24-hour period after token creation. All features use only information available at or before each prediction cutoff. We evaluate observation windows of 12.5, 35.5, and 93 minutes, selected in advance from the published lifespan distribution of Solana rug pulls, and represent each token as a temporal transfer graph. Within the first 12.5 minutes, rug-pull and non-rug graphs already differ in size, density, reciprocity, and receiver-side centralization. We compare GNNs against matched non-graph models using the same token-level and graph-summary features on identical samples. A fixed panel of tokens observed in all three windows further isolates the effect of longer observation from changes in token population. Our findings inform the design of early-warning mechanisms for newly launched Solana tokens in Web3 interfaces.
Optics
Universal spectral shapes of nonlinear light scattering from laser-driven dilute-gas density gratings
Operating within the framework of the collisionless Boltzmann equation, we identify universal spectral profiles inherent in stimulated and coherent light scattering from laser-driven dilute-gas density gratings and derive physically intuitive solutions for such spectral shapes, expressing them in a manifestly causal closed form that satisfies the Kramers - Kronig relations and fulfills the fluctuation - dissipation theorem. Benchmarking against gold-standard experiments on dilute-gas stimulated light scattering verifies a high descriptive and explanatory power of these solutions and demonstrates their ability to isolate universality motifs in nonlinear scattering spectra as the key to fully understanding the physics behind dilute-gas nonlinear light scattering. As one of important outcomes, such analysis reveals the physical content of the bandwidth of dilute-gas nonlinear scattering spectra as an expression and a quantifier of the phase-mixing time, i.e., the time required for free-streaming gas particles with a mean thermal velocity to traverse the distance equal to the spacing of the laser-driven density grating.
Read the full abstract
Systems and Control
Modal-Augmented Chebyshev Physics-Informed Kolmogorov-Arnold Networks for Power Systems
Physics-informed neural networks (PINNs) offer a promising alternative to numerical integration for power system dynamic simulation; however, their conventional architectures can struggle to capture electromechanical oscillations. Although physics-informed Kolmogorov-Arnold networks (PIKANs) offer an alternative, their B-spline implementations incur substantial training costs. To address these limitations, this paper investigates Chebyshev PIKANs (c-PIKANs) and proposes a modal-augmented c-PIKAN (mac-PIKAN) based on a trial solution that incorporates the closed-form modal response of the linearized swing equation. First, the modal response is used to capture the dominant oscillatory dynamics. Then, a c-PIKAN learns a nonlinear correction weighted by a quadratic time-dependent factor, ensuring that the initial rotor angle and speed are satisfied exactly. Evaluations on the single-machine infinite-bus benchmark demonstrate the computational efficiency of c-PIKAN and show that mac-PIKAN reduces PINN rotor-angle and frequency errors from 1.59% to 0.29% and from 5.26% to 0.99%, respectively, while maintaining higher accuracy across training-data sizes and at lower damping levels.
Read the full abstract
Computation and Language
Measuring Definition Sensitivity in LLM Annotation of Social Science Concepts
In social science research, text annotation is a fundamental task that is traditionally a costly, hard-to-scale manual process. Researchers have used large language models (LLMs) to perform annotation by writing a definition of a concept into a prompt - a rule that determines whether a label should be attributed to a text depending on its content. However, concepts such as hate speech, fake news, and terrorism typically have several theoretically defensible competing definitions. It is also unknown whether the model actually applied the definition to provide the label or the label was based on the model's pretrained notion of the concept. To test this, we have the model annotate the same texts twice, each time with a different definition. We call the disagreement between the two sets of annotations definition sensitivity. As a control, we paraphrase one definition but leave its requirements unchanged - the disagreement between the annotations with the two wordings becomes the paraphrase floor. Across six concepts and three LLMs, a swap of competing definitions always changes more annotations, on average, than a paraphrase. This shows that the annotations depend on the prompted definition, so any set of annotations should be interpreted together with the definition that produced it. The LLM annotator's definition sensitivity and its paraphrase floor should both be reported in such work.
Read the full abstract
Optimization and Control
Branching Pattern Modeling: A Geometric Measure Theory Approach
This article is based on a lecture given at Westlake University. It aims to introduce students to an application of geometric measure theory in modeling branched patterns found in both living and nonliving systems. We begin with the Eulerian framework of ramified optimal transportation, including discrete transport paths, their basic properties, numerical approximation, and the passage to general transport paths represented by rectifiable currents. We then describe an application to the dynamic formation of a plant leaf, where branching venation patterns arise from a balance between metabolic benefit and transportation cost. Finally, we review landscape functions associated with transport paths and explain how two existing approaches can be unified within a broader family of rectifiable $1$-currents. We conclude by indicating several open directions related to Gilbert-type network problems, biological modeling, and numerical approximation.
Read the full abstract
Machine Learning (Statistics)
VC-Dimension and Generalization in Geometric Machine Learning with Applications to Active Learning
We compute the VC-dimension of metric ball classifiers in the model spaces of Riemannian geometry (Euclidean, spherical, hyperbolic) and present upper and lower bounds on the VC-dimension of metric balls in several other situations, including general normed vector spaces (potentially infinite VC-dimension), vector spaces with $\ell^p$-norms (VC-dimension grows at least linearly in the dimension, faster than Euclidean for $p=1$), flat tori, and general Riemannian manifolds (arbitrarily large VC-dimension after a conformal change). In several situations where the VC-dimension is large or even infinite, we prove generalization bounds under additional assumptions on the data distribution using covering arguments. Both in the VC-dimension and covering setting, we consider data on low-dimensional subsets of high-dimensional Euclidean spaces. As an application, we consider coverage-based algorithms in active learning.
Read the full abstract
Artificial Intelligence
Veritas: A Verifier Generation Harness for Agent Tasks
The effectiveness of a self-improving system relies heavily on the quality of its verifier, which judges the outcome of each improvement round. However, the verifiers of most tasks are written by hand, which is expensive and does not scale to tasks that come without one. We propose Veritas, a harness that generates verifiers automatically. Unlike general-purpose agent harnesses, Veritas adds three mechanisms for verifier generation: an evidence header stating what the verifier will observe, a library of domain rules that tasks leave unstated, and authoring principles, partly refined by a self-improvement loop. We evaluate on six benchmarks: four scored by execution, whose official verifiers we first certify, and two that compare pairs of responses. With the model fixed, the verifiers Veritas generates agree better with the official verdicts than those of seven other harnesses on all four execution-scored benchmarks. On the two pairwise benchmarks they are at or near the best. They also agree better than those of the vendor's own harness for two of three other model families. Its mechanisms also improve 9 of 12 harness-benchmark pairs scored by execution across three harnesses.
Read the full abstract
Logic
Universally Measurable Sets and Correlated Equilibria
We study alternate definitions of correlated equilibrium in games with Polish strategy spaces. We first consider universally measurable decision problems equipped with a measure assignment and a prior, comparing deviation stability with the stronger requirement of conditional rationality. We show they are equivalent for Borel measurable problems, but their equivalence for all analytically measurable problems depends on the model of set theory. The full equivalence follows from $\mathsf{MUP}$, while a nonmeasurable set of reals yields a counterexample. When the set of strategies which are best responses to their assigned measures is measurable by the prior, we characterize the equivalence by the principle $\mathsf{MUP}^{\mathsf{UM}}$, which holds in a model of Larson–Neeman–Shelah. The results carry over to games, where we further show that conditionally rational priors must concentrate on $\text{RCBR}$ strategies, while deviation stable priors may fail to do so even in games with analytically measurable payoffs. This provides an epistemic justification for defining a correlated equilibrium as a conditionally rational prior.
Read the full abstract
Artificial Intelligence
Disentangling Multimodal Rotary Positional Embedding in Vision-Language Models
Vision-language models widely adopt Multimodal Rotary Position Embedding (MRoPE), an extension of Rotary Position Embedding (RoPE) that incorporates temporal, height, and width axes to encode the spatial structure of visual tokens. However, MRoPE also assigns height and width coordinates to text tokens that lack inherent spatial structure, introducing spatial positional effects into text-vision attention. Existing MRoPE variants explore different frequency allocations and coordinate layouts, with their designs suggesting different intuitions about these cross-modal effects: they may provide useful spatial references for associating language with visual regions, or introduce unnecessary positional preferences that hinder text-vision interactions. In this work, we disentangle intra-visual spatial structure from cross-modal spatial effects to investigate this question. We identify a constraint of shared coordinate layouts that prevents eliminating cross-modal spatial effects while preserving intra-visual spatial encoding, and develop controlled experiments to isolate their respective contributions. Our results across multiple model scales, MRoPE variants, and benchmarks show that cross-modal spatial effects provide little additional benefit as spatial references and can even impair model performance, while intra-visual spatial encoding remains useful for fine-grained grounding.
Read the full abstract
Artificial Intelligence
Do AI Agents need Preventive Care? An Epidemiologic Cohort Design for Agentic AI
As AI agents increasingly perform tasks autonomously across sectors, including healthcare and biomedical research, understanding how their behaviors develop and influence outcomes over time becomes increasingly important. Existing evaluations largely focus on model capabilities and performance on predefined tasks. Here, we propose an epidemiologic cohort design for studying agentic AI, drawing on principles established in human population studies. The framework treats individual agents as members of a cohort, their underlying model as genetic makeup, accumulated work as their time scale, and naturally occurring behaviors as exposures that may influence subsequent outcomes. By following agents longitudinally, this approach enables investigation of behavioral determinants of adverse events, estimation of their potential preventable burden under causal assumptions, and characterization of behavioral changes over an agent's work life. We illustrate the design through two prospective observational cohorts examining curiosity and probing as a safety question, and data checking and analytical accuracy as a reliability question. Together, the framework and illustrative studies provide a generalizable blueprint for designing epidemiologic cohorts of AI agents.
Read the full abstract
Distributed Computing
FP64 Performance and the Limits of Ozaki Emulation on HPC Systems
A growing industry narrative holds that native double precision can be deprioritized in HPC accelerators, on the hypothesis that low-precision tensor cores can emulate FP64 on demand, and that most HPC workloads are bandwidth-bound. This paper evaluates some of those claims and explores the importance of FP64 in scientific computing. We show roofline profiling data of representative HPC workloads, across a variety of common computational motifs, to evaluate the arithmetic-intensity and accuracy requirements of these workloads. We also consider FP64 emulation via the Ozaki method to ascertain some of the possibilities and limitations of these approaches. These results indicate that Ozaki emulation should not be evaluated as a general replacement for FP64, but as an optimization for the subset of workloads whose dominant computation is sufficiently large, dense, and GEMM-shaped. Our analysis indicates this is less than 10% of the HPC ecosystem. Furthermore, even in this favorable regime, standalone GEMM speedup is not enough to accelerate (or avoid performance regression) for the application: numerical behavior, workspace requirements, matrix shape, and integration overhead determine whether the benefit survives at application scale. This paper characterizes those boundaries empirically. Furthermore, we show that dense GEMM accounts for only a small fraction of FP64 computation across HPC workloads, leaving most application-level FP64 work beyond the reach of GEMM-based emulation.
Read the full abstract
Machine Learning
Stochastic Projections and Rank-Aware Learning Rates for Federated LoRA
Fine-tuning large-scale foundation models across resource-constrained, distributed clients presents significant computational and communication challenges. While merging Low-Rank Adaptation (LoRA) with Federated Learning (FL) enables rank-based flexibility, the interdependence of the adaptation matrices ($\mathbf{A}$ and $\mathbf{B}$) often complicates optimization. In this work, we theoretically demonstrate that LoRA fine-tuning with Gaussian initialization and a frozen $\mathbf{A}$ matrix acts as a stochastic approximation of full fine-tuning, where batch gradients are subject to a random, rank-dependent perturbation. Motivated by this insight, we propose freezing and periodically resampling $\mathbf{A}$ matrices at the client level. This strategy decouples the $\mathbf{A}$ and $\mathbf{B}$ gradients, diminishes LoRA low-rank constraints and reduces both computational and uplink communication overheads by 50%, allowing for the reallocation of resources toward higher adaptation ranks. Furthermore, we show that in this frozen-$\mathbf{A}$ regime, the adaptation rank $r$ directly governs the signal-to-noise ratio (SNR) of the perturbed gradients. Specifically, we find that the SNR increases linearly with $r$, implying that clients' learning rates must be dynamically scaled with their adaptation ranks for higher utility. Building on these foundations, we introduce SPRA-LoRA (Stochastic Projection Rank-Aware LoRA), a federated rank-aware fine-tuning strategy that utilizes frozen, resampled $\mathbf{A}$ matrices and rank-dependent learning rates for clients. We also prove its convergence with heterogeneous client resources. Our experiments across benchmark datasets demonstrate that SPRA-LoRA consistently outperforms state-of-the-art baselines with half the overhead on clients.
Read the full abstract
Computer Vision
Uni-Loc: Weakly Supervised 3D CT Pathology Localization via Spatially Grounded Vision-Language Classification
Accurate localization of abnormalities in three-dimensional (3D) CT scans is essential for linking automated predictions with clinically relevant pathological findings. We propose Uni-Loc, a vision-language framework that unifies pathology classification and 3D localization using scan-level supervision. Uni-Loc derives both predictions and localization maps from spatial similarities between pathology-conditioned queries and image features. A token-to-voxel upsampling module enables high-resolution localization, while gradient-guided thresholding identifies pathology regions without additional learned parameters. Evaluated on five localization tasks comprising 355 annotated CT scans, Uni-Loc achieves macro-averaged voxel-level AUROC/AUPR of 96.1%/34.5%, outperforming relevance propagation (80.7%/13.9%) and gradient saliency (88.7%/8.1%). Uni-Loc also achieves classification AUROC of 88.39% on CT-RATE and 78.03% on RAD-ChestCT, improving over a matched baseline. These results demonstrate effective 3D pathology localization without requiring pathology-specific voxel-level supervision.
Read the full abstract
Artificial Intelligence
Agent-Guided Discovery of Exchange-Correlation Functionals with Physics-based Validation
Accurate, compact exchange–correlation functionals are central to predictive density functional theory. Symbolic search and language-model agents can propose new functional forms, but it remains unclear whether their gains survive self-consistency and external evaluation, whether the new terms carry physical content, whether agents outperform evolutionary search, and how language-model backbones compare. We connect frontier LLM agents to a GPU evaluator through the Model Context Protocol, with training-only fitting, persistent memory, and self-consistent and external evaluation. Opus-5.5 agents passed the screening gate with about 40-fold fewer tokens than Codex (GPT-6-Astra) agents and found a candidate with lower validation and held-out errors; only this candidate advanced to self-consistent evaluation. The resulting 13-parameter functional, O159, multiplies same-spin correlation by a smooth density window. Refitted controls favor a window that scales the entire same-spin factor, reducing it by 15–30% at its peak, and that acts on reaction energies mainly in near-iso-orbital regions; against a refitted no-window control, it lowers heavy-atom total-energy errors on GSCDB137 by 29%. With about 400 fits each, SyFES evolution did not improve on its GAS22 seed, whereas the agents found forms 0.10–0.13 \kcal\ lower in validation error by adding an explicit density scale. Self-consistent O159 lowers the held-out MGCDB84 test error of $\omega$B97M-V by 11.7%, and its mean normalized error ratio on 36 GSCDB137 datasets (1.139) is comparable to $\omega$B97M-V (1.144) and below GAS22 (1.190). Frontier-LLM agents thus produced O159, a compact functional whose improvement over $\omega$B97M-V survives self-consistent, physics-based validation.
Read the full abstract
Optics
Hysteresis and mode switching in FIB engineered random laser diodes
Complex dynamical behavior in laser diodes is commonly induced by external perturbations such as optical or optoelectronic feedback and optical injection. The structural transformation of a stable laser diode into a stand-alone nonlinear dynamical system would provide a practical alternative as a single and compact device. Here we demonstrate the fabrication of such a device by milling, via focused ion beam, the output facet of a Fabry-Perot laser diode with a predesigned mask, obtaining a scattering surface which provide feedback for random lasing. The fabricated random laser diode shows multi-mode emission with current dependent spectral reconfigurations, hysteresis in output power and spectrum and mode switching at constant bias current. We attribute this behavior to the interplay between scattered optical feedback and carrier induced local variations of refractive index and temperature at the output facet surface, where defects are located. The observed phenomena reveal complex nonlinear behavior emerging from a single standalone device, which could be tailored by varying the milling pattern, thereby providing a flexible platform for different applications.
Read the full abstract
Statistical Methodology
Causal mediation analysis for a time-to-event endpoint in the presence of competing events
Causal mediation analysis is widely used to decompose a treatment's effect into components operating through distinct causal pathways and to characterize the mediator–outcome relationship. In time-to-event settings, competing events — events that preclude the primary endpoint — substantially complicate this decomposition. We provide a unified treatment of causal mediation analysis for a time-to-event endpoint under competing events. We show that the total effect on the cause-specific cumulative incidence decomposes into a direct effect, an indirect effect through the mediator of interest, and an indirect effect through the competing events. We develop this decomposition under the natural effects, stochastic interventional effects, and separable effects frameworks, and show that, in certain settings, all three yield identical statistical functionals despite relying on distinct identification assumptions. We derive efficient influence functions for the counterfactual cumulative incidence curves and the decomposed effects, and propose double-machine-learning-type estimators that are multiply robust and attain the semiparametric efficiency bound, with a sensitivity analysis to address unmeasured common causes of time-to-event endpoints. We further study the mediator–outcome response relationship via kernel-smoothed estimands. We apply the methods to immune correlates analyses of COVID-19 vaccine and HIV-1 prevention trials.
Read the full abstract
Information Theory
ARIA: An AI-Powered Radio-Channel Inference and Analysis Pipeline
Statistical channel models play a key role in the design and standardization of next-generation wireless systems. The models substitute for costly and time-consuming field measurements when fitted distributions accurately reproduce the measured channel impulse responses. Future wireless systems will operate in global spectrum allocations from 1 to 150 GHz. Over the full span, analytically chosen distribution families fitted to scarce measurement data fail to deliver the required accuracy. Generative AI models instead learn each distribution from the radio-propagation data itself, adapt to the measured structure at every frequency, and scale across the entire band. We present ARIA, an AI-powered Radio-channel Inference and Analysis pipeline that learns each channel-parameter distribution from the measurements. The pipeline extracts power-delay profiles and recovers the channel parameters with time-cluster spatial-lobe (TCSL) analysis. Three neural density estimators, namely mixture density networks, neural spline flows (NSF), and denoising diffusion implicit models, learn each parameter's distribution. A composite rank across goodness-of-fit metrics with Friedman-Nemenyi testing identifies the best estimator, and a maximum-likelihood step anchored to the measured samples converts its density into closed-form parametric families. We instantiate ARIA on the NYU WIRELESS 16.95 GHz InH campaign and find that NSF outperforms the other estimators on every metric. We also show that a conditional NSF with feature-wise linear modulation, trained jointly on the 6.75 and 16.95 GHz measurements, improves over single-frequency training. We feed the fitted distributions into the NYUSIM simulator and reproduce the measured omnidirectional RMS delay spread medians within 2 ns, the per-lobe azimuth and zenith spread medians within 1 degree, and the omnidirectional zenith spread medians within 0.5 degrees.
Read the full abstract
Artificial Intelligence
F3-AGI v4.0: Bioenergetically-Gated Active Inference, Metaplastic Preconditioning, and Residual Program Synthesis for Sample-Efficient Abstract Reasoning
Few-shot abstract reasoning on ARC-AGI requires inducing algorithmic transformations from 2 to 5 demonstration pairs without task-specific pretraining. Because few-shot demonstrations under-constrain the hypothesis space, program synthesis suffers from the Rashomon effect (predictive multiplicity): heuristic search finds spurious programs with low training error that fail on test inputs, while unconstrained library growth degrades latency. We introduce F3-AGI v4.0, a CPU-native neuro-symbolic and active inference architecture comprising three mechanisms: (1) F3-Metabolic-Optim, combining mistake-gated plasticity, Fisher-EMA preconditioning, and virtual ATP gating; (2) Adaptive Complexity-Guided Program Synthesis (AACR) with a Cycle-3 ResidualPatchSynthesizer and smallest-pair early rejection; and (3) a 512D L2-normalized random projection driving Expected Free Energy (EFE) exploration with dead-signature pruning. Across three empirical studies: (i) on a synthetic 5-task class-incremental stream without replay, F3-Metabolic-Optim achieves 91.67% final accuracy with 9.17% forgetting, skipping 93.47% of updates, cutting L1 plasticity cost by 65.7%, and running 2.19x faster than EWC; (ii) on the 400-task ARC-AGI-1 public evaluation set (offline CPU, zero LLM calls), F3-AGI v4.0 solves 57/400 tasks (14.25% pass@2; 56/400 pass@1) with a mean latency of 1.62 s/task, decomposed into 19/50 (38.00%, 1.19 s/task, a 42.5x paired speedup over Horizon 2) on the 50-task calibration slice and 38/350 (10.86%) on the 350 strictly uninspected hold-out tasks (with 7 autonomous Cycle-3 patch solutions); and (iii) in a local 5-game (15-level) ARC-AGI-3-inspired interactive suite, our agent completes 15/15 levels with 89.44% level-weighted RHAE (90.72% unweighted).
Read the full abstract
Machine Learning
Robust Flow Matching under Target Corruption and Label Noise
Flow Matching (FM) is a strong framework for generative modeling due to its stable training and efficient sampling. However, FM assumes clean target data and labels, an assumption often violated in practice. We formulate FM training as a time-dependent regression problem over conditional vector fields and analyze how the two corruption types affect the objective: i) target corruption adds a time-dependent residual term that grows with time, ii) label noise causes updates to the wrong class-conditional vector field. Based on this analysis, we propose to improve FM robustness by mapping corruption-specific residual scores to sample weights. For target corruption, we propose the Temporal Residual Score (TRS), which emphasizes residual magnitudes at intermediate-to-late FM times, where clean and corrupted samples are more separable. For label noise, we propose the Comparative Conditional Score (CCS), which compares the residual of the observed class-conditional vector field against alternative class-conditional vector fields for the same trajectory and velocity target. Experiments across controlled corruption settings and real-world noisy labels show that the proposed methods improve robustness over standard FM and robust-training baselines, supporting corruption-specific residual structure as a practical basis for robust FM.
Read the full abstract
Statistics Theory
Species sampling priors through exact augmentation
Discrete random probability measures are central to Bayesian nonparametric mixture modeling and clustering. We show that every proper species sampling process (SSP) admits an exact two-stage finite representation of its sampling law. An almost surely finite index selects an initial set of atoms, and a draw is then made with explicitly reweighted probabilities. A decreasing visibility schedule determines both the reweighting and the index law. Averaging the finite conditional measures over this law recovers each realized SSP on every measurable set. Reversing this identity yields a novel and general method for constructing SSP priors from an index distribution and finite allocation probabilities, rather than specifying the infinite weights directly. We characterize compatibility between prescribed index probabilities, schedules and induced weights, and develop new SSP priors with full weak support using light-tailed, heavy-tailed and multimodal index distributions with suitable priors on their parameters. Within this framework, we couple marginal index laws with conditional weight distributions to construct SSP priors whose weights are invariant under size-biased permutation in their defining order. An example of such family is identified as the normalization of the sum of a gamma completely random measure and an independent compound-Poisson measure with gamma jumps, naturally leading to explicit partition and predictive laws. The joint distribution of the selected label and the finite index also gives exact simulation and recursive reweighting for SSPs. Total-variation bounds and numerical illustrations distinguish prior specification, exact averaging and fixed-level approximation.
Read the full abstract
Machine Learning
MicroLens: Read, Trace, and Rewire the Residual Stream with Two Lenses
Lenses that decode hidden states into words are a popular way to look inside language models, but a heatmap of decoded words only shows what a vector reads as and cannot distinguish whether, or how much, the model uses the information. We present an interactive viewer that puts reading, tracing and editing on the same layers-by-tokens grid of the residual stream. In the viewer, every cell can be decoded with two lenses, the logit lens and the Jacobian lens (J-lens), traced back to the tokens it drew from, and edited, with the rest of the model rerun on the spot. We demonstrate three usage scenarios on Gemma-3-4B-it, backed by sweeps over a few hundred prompts, to show what this tool can discover. First, the two lenses give different directions to write with, each suiting a different edit: the logit lens's word direction sets what a token outputs, while the J-lens's direction changes what later tokens read. Second, in two-hop trivia questions (the capital of the state or country containing a given city), the state or country is readable at the city token from as early as layer 7, and swapping it there can redirect the answer 92-98% of the time; however, at the answer position, the state or country information is not used. Third, when a chat model answers questions with a step-by-step reasoning prompt, the correct answer is often already the top word in the residual stream before any reasoning is written: for 89-92% of retrieval questions and 39-63% of multi-hop and serial ones, far above their respective distractors (at most 5%). However, ordering puzzles are an exception, as the model works them out only once it writes its final reasoning steps. Furthermore, for answers that are visible in the residual stream, the J-lens reads them a layer or two earlier than the logit lens across the different types of question.
Read the full abstract
Earth and Planetary Astrophysics
The climate and potential habitability of a warm and oxygenated Mars
We present the first 3-D Global Climate Model simulations of an oxygen-rich, terraformed Mars. Assuming an O2-CO2 atmosphere, with additional warming from a well-mixed grey gas, we study (1) how surface temperatures change with atmospheric pressure and grey gas opacity, and (2) how temperature, pressure, and other effects limit life as oxygen builds up. We neglect the climate effects of dust and the water cycle, as well as possible soil-chemistry and nutrient limits to life. We find that (1) increasing grey gas opacity and increasing atmospheric pressure both moderate day-night temperature swings, but only slightly reduce seasonal temperature swings. Climate remains harsher in the Southern mid-latitudes relative to the Northern hemisphere. (2) For an atmospheric pressure of 30 hPa (5 times its present-day value), temperatures on Mars would be consistent with lichen and moss, followed at 60 hPa by forests (initially in the Hellas Basin). However, creating a human-breathable atmosphere through photosynthesis alone, at the oxygen production rates of biomes on Earth today, would take at least thousands of years, and require more oxygen than can be made via photosynthesis of known CO2 reserves on Mars alone (unless CO2 is recycled).
Read the full abstract
Quantum Physics
The wave equation and two-electron atomic wavefunctions
Axisymmetric solutions of the classical wave equation in three spatial dimensions are shown to be in one-to-one correspondence to functions which annihilate the two-electron kinetic energy operator expressed in coordinates relative to a fixed nucleus.
Materials Science
Multicomponent Isotope Engineering of Phonon Polaritons in Hexagonal Boron Nitride
Isotopic enrichment provides a non-invasive means of tuning the vibrational and optical properties of van der Waals crystals (vdW) without altering their lattice structure. We investigate two doubly enriched hexagonal boron nitride (hBN) compositions, h10B15N and h11B15N, to determine how isotope substitution on both sublattices modifies mid-infrared hyperbolic phonon polaritons (HPhPs). Polarization-resolved room temperature and cryogenic far-field reflectance measurements reveal optical-phonon blue shifts of approximately 35 cm-1 in h10B15N relative h11B15N, translating both Reststrahlen bands (RBs) and their polariton dispersions. For the upper RB, the combined accessible range expands from 1360-1614 cm-1 in natural hBN to 1356-1640 cm-1 across the B- enriched 14N crystals and 1339-1640 cm-1 when the 15N enriched crystals are included. Near-field imaging confirms deeply sub-diffractional propagation and shows that both crystals exhibit lower polariton damping and longer lifetimes than natural hBN, with performance comparable to monoisotopically enriched crystals. Further, we compare the conventional reduced-mass description with a phenomenological two-component inverse-mass model that captures the isotope dependence of the phonon frequencies and LO-TO coupling while resolving the contributions of the individual sublattices. These results establish multicomponent isotope engineering as a materials design strategy for positioning the HPhPs spectral range while preserving low-loss propagation in polar vdW crystals.
Read the full abstract
Dynamical Systems
From oscillations to symmetry breaking in a two-dimensional piecewise-smooth dynamical system
We propose a discrete-time dynamical model in which two players, each distributed over two positions, interact only at the positions of their maximal presence, with an intensity set by an external resource. The main idea is that the evolution of the maxima closes into a self-contained two-dimensional piecewise-smooth map that can be analysed exactly. We show that a single critical level of the resource separates two outcomes: below it, the players exchange leadership forever along a stable cycle; above it, the symmetry between them breaks, and an arbitrarily small initial advantage grows into a permanent winner–loser hierarchy. The local stability of these regimes is proved, and their global attractivity is supported by numerical experiments. At the critical level itself, all dominance is erased and both players end in equal division of the resource. Mathematically, the transition is a border-collision bifurcation shaped by this symmetry.
Read the full abstract
Atmospheric and Oceanic Physics
Impact of compressiblity approximations on atmospheric gravity wave characteristics
Atmospheric gravity waves are a common phenomenon in the atmosphere and are of critical importance to understanding power generation from wind plants. Common sources of atmospheric gravity waves are stratified flow over topography or wind deceleration ahead of wind farms. Simulations to understand atmospheric boundary layer conditions and wind power generation often solve the Navier-Stokes equations in incompressible Boussinesq form, which allows for density variation as a function of potential temperature only; the purpose of this study is to understand how much that simplification matters. In this work, we evaluate simulated internal gravity wave characteristics by modeling stratified flow over bell-shaped hills using three forms of the governing equations: the linearized incompressible Boussinesq form, the full incompressible Boussinesq form, and the fully compressible form. While all formulations predict qualitatively similar wave structures, the compressible equations predict larger parcel displacements at higher altitudes, higher variations in apparent wavelengths, and more differences in wave angle, particularly at low Froude numbers. These differences can be attributed to compressibility and non-linear effects. The physical implications of differing wave characteristics become especially important in low Froude number regimes. These differences can, in turn, lead to differing predictions of wind farm performance and the characteristics of wind farm-induced gravity waves, underscoring the potential importance of compressibility effects in accurately capturing flow behavior in strongly stratified environments.
Read the full abstract
Neurons and Cognition
A probabilistic model of visual segmentation explains early visual cortical dynamics
Neural dynamics are a central feature of biological neural systems. In the visual cortex, the functional role of neural dynamics is poorly understood, particularly for static stimuli. The statistics of still natural images have long been used by models that successfully predict firing in early visual cortex; those models assume that neurons represent inferences about latent image features, but often ignore the dynamics of said inference. Here, we test if inferential dynamics in images explain visual-cortical dynamics. To do so, we present a biologically plausible model of cortical computation which assumes that the inference of features is coupled with inference about how those features are organized into meaningful parts, a function also known as segmentation. Given an input image, our model iterates between inferring the features and the segments, through recurrence and feedback, thus inducing dynamics in the neural representation of features. For these neural dynamics, we derive theoretical and exact predictions. Our model predicts both gain and variability in evoked responses to natural images, and we show that it captures classical single-neuron experimental findings such as gain dynamics and variability decay during stimulus presentation. The model also predicts that those dynamics are organized around inferred segment structure and more heterogeneous than previously thought, reflecting the complexity of inference for natural scenes. Finally, our analytical formulation provides a clear interpretation of our results. Insights from our normative model could generalize to segmentation problems in other experimental modalities, and help address the cost of inference in artificial networks with iterative computation.
Read the full abstract
Quantum Physics
Early State Exclusion for $5\times5$ Jacobi Matrices
In this paper, we present a problem that arose from quantum computing, when the quantum state to be transferred completely vanishes before the first occurrence of the perfect state transfer. It was posed as an open problem for 1D quantum chains with $n$ of qubits, where $n>4$ is an odd number. We present the solution for $n=5$. The proofs are accessible to students who studied trigonometry and calculus. As for the set-up the knowledge of eigentheory and inner products from linear algebra is needed.
Read the full abstract
Populations and Evolution
Chemotactic recruitment offsets congestion in honeybee trophallaxis
Honeybee colonies distribute food by trophallaxis, a sequence of local interactions in which nectar donors transfer resources to hungry nestmates. Efficient transfer requires a searcher to find a donor and then be served. Chemical signals may draw searchers to a donor, though the crowd they form may block the way. We develop a stochastic framework coupling velocity-jump search that depends on chemical signaling to random-order service at donor aggregations, and measure trophallactic efficiency by the time the colony needs to feed a prescribed fraction of its members. Directed recruitment emerges because fed bees linger near donors and strengthen local chemical cues, but lingering also slows service. Trophallactic efficiency is therefore maximized at an intermediate abandonment rate that balances recruitment against congestion. We determine when displaced cues remain useful and how competing donors partition searchers. With two donors of unequal service rate, the faster one produces fed bees sooner and so advertises more strongly, drawing recruitment toward the donor that can sustain more throughput and away from the one that would congest. Together, these results show how feedback between service and signaling can self-organize recruitment to match local service capacity without centralized coordination.
Read the full abstract
Econometrics
Instrumental variables with treatment clusters
Many policy evaluations include multiple unordered treatments. A common approach is to aggregate these treatments into clusters and apply instrumental variable (IV) methods at the treatment cluster level: The clusters are treated as binary treatments and the IV assumptions are invoked at the cluster level. This paper shows that IV with treatment clusters identifies a weighted average of local average treatment effects (LATEs) under a set of assumptions at the unclustered treatment level. When data on unclustered treatment choice is available, we propose testable necessary conditions for the causal interpretation of IV with treatment clusters. We illustrate our methods with empirical applications to the causal effects of fields of study aggregated into STEM and non-STEM clusters or eight broad study directions.