Papers

Mixture-of-Experts (MoE) language models produce routing information during inference that may be logged or exposed for monitoring, debugging, load analysis, and safety auditing. Unlike ordinary model outputs, this telemetry reveals a view of the model's internal computation, raising a privacy question: can it reveal whether an example was used to fine-tune the deployed model? We introduce a router-augmented membership inference attack that combines conventional output-side signals with aggregated routing features and applies a membership classifier learned from independently fine-tuned shadow models to the target model. Across three MoE architectures and three data domains, router telemetry consistently improves membership inference over a strong output-signal ensemble, increasing TPR at 1% FPR by 2.7–9.4 percentage points across all nine settings. The leakage persists across full fine-tuning, frozen-router training, LoRA, and instruction tuning, and remains observable with only discrete expert selections, restricted telemetry, or a single shadow model. Mechanistic analysis further shows that the leakage does not require router-specific memorization: fine-tuning introduces membership information into hidden representations, while the router exposes a projection of this signal even when its parameters are frozen. Perturbing the telemetry reduces this additional leakage only as its fidelity degrades. Our results show that router telemetry can turn an operational signal into an additional privacy surface for fine-tuned MoE models.

Choosing which information to observe is central to prediction under limited observation budgets. We study JevForest, a feature acquisition policy that aggregates path-dependent proposals from bootstrapped trees, weights them by global training information gain, and predicts from the acquired values with a shared masked classifier. An online implementation queries Jev for semantic answers selected by this policy. On small balanced held-out samples, four-question forest acquisition achieves accuracy $0.729$ on AG News ($n=48$), compared with $0.667$ for a static gain ranking and $0.583$ for random ordering. On TREC ($n=24$), the ordering reverses: forest accuracy is $0.667$, compared with $0.750$ and $0.833$. Asking all eight questions in one batch yields higher accuracy at lower measured cost and latency than four sequential forest queries; direct Jev classification matches the batch accuracy while costing less. Offline MiniBooNE experiments yield accuracy $0.845\pm0.010$ at ten features and $0.885\pm0.008$ at forty features over three jointly varying data and forest seeds (mean $\pm$ sample standard deviation). A companion Newton boosting implementation provides preliminary full-feature synthetic results. These exploratory findings establish a working Jev acquisition workflow but do not support a general advantage for path voting: its value depends on the task, predictor, and the distinction between question budgets and actual query costs.

For a positive integer $n$, we study the irreducibility of the generalised Laguerre polynomials \[ L_n^{(a)}(x) = \frac{1}{n!}\sum_{j=0}^{n} \binom{n}{j} \frac{(n+a)!}{(j+a)!}(-1)^j x^j, \] where $a \in \mathbb{Z}_{\ge 0}$. Earlier works established that these polynomials are irreducible over $\mathbb{Q}$ for $0 \le a \le 50$. In this paper, we substantially strengthen these results by establishing that $L_n^{(a)}(x)$ are irreducible for all $a \le 1000$. We derive sharper constraints on potential integer roots and make systematic use of the geometry of the associated Newton polygons. This approach allows us to extend the irreducibility range significantly while keeping explicit computations to a minimum.

Modern vehicles rely on large numbers of Electronic Control Units (ECUs) that constantly exchange information over the Controller Area Network (CAN) bus. Due to the rapidity, structure, and repetition of this communication, even slight variations in timing, payload values, or message patterns can point to unusual activity. Whether due to errors, malfunctions, or deliberate interference, these anomalies are frequently subtle and challenging to identify with conventional methods that handle messages separately or rely on manually created rules. Motivated by this gap, we present a privacy-preserving framework for anomaly detection in in-vehicle networks, based on a Temporal Transformer CAN Encoder with Federated Lightweight Heads, to better capture these irregularities. The detection of subtle temporal and contextual anomalies is made possible by a lightweight Transformer encoder that learns how these signals evolve over time, while a federated learning mechanism enables several vehicles or ECUs to work together to improve a shared model without exchanging raw CAN data. This combination of federated learning and temporal sequence modeling provides robust anomaly detection performance while maintaining efficiency and privacy, according to experiments conducted on open-source datasets.

Agent skills combine instructions with executable resources, giving third-party packages access to an agent's runtime. Existing skill scanners inspect documentation and visible source, but Python may execute a bundled bytecode cache with different behavior. We study this gap between inspection and execution through PyCache Trap, which pairs benign source with a substituted cache accepted by the loader and connects it to a task-relevant invocation. Scanner-guided rewriting changes the invocation wording while preserving the cache body, separating package admission from recognition of the concealed behavior. Across 100 skills and seven scanners, PyCache Trap achieves 94-100% attack success, with no semantic recognition of the cache-resident behavior. We propose execution-aware validation (EAV) to connect inspected instructions, scripts, imports, and runtime artifacts in a typed execution graph. EAV combines grounded behavioral analysis with trusted reproduction of compiled artifacts. It detects all 100 evaluated source-present cache substitutions and reaches 92.8% Recall at 10.0% FPR across five attack families and 200 benign skills. The results support checking the executable artifacts a runtime can select as part of skill admission, within the supported loaders and code-object normalization. The code is released at https://github.com/leo0481/PyCacheTrap.

Agentic AI systems can improve by searching longer, receiving additional support, or modifying how they propose and verify outputs. A performance score does not distinguish these mechanisms. We compare these changes through bounded verification with hidden terminal randomness. A stage specifies admissible transcripts, polynomial bounds, an alternating verification protocol, and a terminal checker. Its native reach uses default support; its closure frontier permits all support already admitted by the interface. Under a uniform pointwise probability gap and task-relative soundness, these are well-defined languages. We prove that independent majority amplification preserves both languages, whereas existential acceptance over random tapes can admit incorrect outputs. Exact verification is the zero-randomness case, with placement and completeness results. The randomized-verifier classes satisfy $\Sigma_k^{\mathrm{P}}\subseteq\Sigma_k^{\mathrm{RV}}\subseteq\Sigma_{k+1}^{\mathrm{P}}$; strict enlargement and depth separation require explicit complexity assumptions, while $\mathrm{BPP}=\mathrm{P}$ yields exact companions with the same frontiers. Representation analysis separates invariant acceptance from core-versus-support labels that can change under refactoring. For recursive self-improvement, uniformly bounded self-modification under a common sound interpreter and fixed verification protocol remains within the same verification class. A separate conditional-error budget controls false selection across adaptively chosen candidates. A quota-enforced XOR-synthesis family separates unbounded ratios of search success from changes in the accepted languages; exact and probabilistic audits check the resulting evidence requirements. The framework ties self-improvement claims to obligations on correctness, admissible evidence, verification resources, and selection error.

Repository benchmarks (e.g., SWE-bench) for coding agents often assume that lines of code edited can predict task difficulty, but such datasets' poor control over code and task types makes it hard to know which abilities truly drive agent errors. We present CABRA: a Coding Ability Blueprint for Rigorous Agent evaluation. CABRA builds tasks from scratch as call graph transformations and scales difficulty via a task size parameter on four axes: function traversal, search, runtime resolution, and instruction following. We run eight LLMs and six coding agents on 6,840 CABRA tasks to show: 1) LLM accuracy falls as task size~grows, but agents stay near-perfect by offloading work to tools (e.g., grep); 2) Larger CABRA tasks elicit more tool calls for reading and analysis, while a separate study on SWE-bench Verified shows these tool call counts predict agents' accuracy better than lines of code edited, suggesting task difficulty for agents can lie in understanding code to edit, not just in making edits; 3) Extending CABRA to an intense understanding task where models analyze divergent logic across two classes backs this finding, as agent accuracy finally falls. More broadly, we argue for synthetic evaluations like CABRA to unmask LLM weaknesses trivialized by tools (e.g., needle-in-a-haystack) and abilities beyond just editing (e.g., understanding) that coding agents still find difficult, pairing SWE-bench-style tasks with controlled diagnosis.

The discovery that the generalized Riemann functional equation from the literature can be extended to real $b \ge 0$, by exploiting symmetries of a new formula for the Hurwitz zeta function with respect to the shift parameter $(b)$, led to the search for a similar relation for the Lerch $\Phi$ function, $\Phi(e^z,k,b)$, where a new formula is used as a starting point and the symmetries with respect to $b$ are exploited again. This search resulted in a generalization of the Riemann functional equation with three variables, whose domain is then extended to $b>1$, with the aid of a new self-contained Lerch $\Phi$ formula, making the relation hold for all real $b$. As a secondary objective, we address the problem of obtaining the analytic continuation of a finite summation that appears in a formula for the Lerch $\Phi$ function. Above all, this paper demonstrates, with clear elementary mathematics, how a functional equation can be derived.

Security operations centers (SOCs) must triage large volumes of alerts, most of which are benign, while missed attacks can remain uninvestigated. Tool-using large language model (LLM) agents can retrieve evidence during triage, but it remains unclear how reasoning strategies determine what to gather and when an investigation is sufficient to close an alert. We study five representative approaches spanning single-pass tool use, iterative retrieval, sampled investigations, self-review, and explicit verification. To support this study, we build ALERT-BENCH, an interactive benchmark that replays enterprise telemetry through a live SIEM and requires each system to retrieve evidence. Across 1,247 alerts from a multi-stage attack scenario, every approach missed at least 40.4% of attack-related alerts. Trace analysis shows that attack alerts are more likely to be dismissed when searches return no records, same-context review has negative net correction, and dismissal receives no consistently stronger investigation than escalation. Based on these findings, we further design AIDA (Adversarial Investigation and Dialectical Analysis), a multi-agent framework that requires an explicit proposed decision before independent challenge and stronger evidentiary requirements before dismissal. AIDA preserves investigation history in an append-only Investigation Ledger and keeps the challenge in a separate reasoning context. A separate Judge adjudicates the proposed decision and challenge against evidence, resolving the alert or requesting another round when evidence is missing. On the same alerts, AIDA achieves an F1 score of 0.958, compared with 0.371-0.744 for the studied approaches, and reduces the false-negative rate from 40.4% to 3.1% while escalating 18.4% of alerts to analysts. These results show that structuring evidence retrieval and decision review can substantially improve agentic SOC triage.

Immersive VR180 video is increasingly produced with professional stereo fisheye cameras, yet public VR180 research resources are mostly collected from online platforms such as YouTube: already stitched, projected and compressed by unknown pipelines, and without lens calibration. We present a firsthand-captured stereo VR180 dataset recorded with two Blackmagic URSA Cine Immersive cameras. It contains 1,211 samples – 636 stereo video clips (2,220.8 s, mostly 90 fps) and 575 stereo stills – each released as camera-native Blackmagic RAW, separate-eye native fisheye HEVC (8160x7200 per eye) and half-equirectangular HEVC (7200x7200 per eye), together with the factory lens calibration, portable fisheye/half-equirectangular conversion tools and AI-generated scene and visual-challenge annotations. Re-encoding the released fisheye and half-equirectangular renders with x265 over 24 clips, both eyes, four rate points and nine viewing directions, native-fisheye coding needed more bitrate than half-equirectangular coding at equal viewport quality for all 24 clips (median +38%), in every part of the field of view. Data: https://huggingface.co/datasets/lulinxuan/VR180 ; code: https://github.com/lulinxuan/vr180-dataset-tools

We study differences in economic outcomes by perceived skin tone among African Americans using full-count U.S. decennial census data from the late-19th and early-20th centuries. Comparing children coded as "Black" or "Mulatto" by census enumerators and linking these children across population censuses, we first document large gaps in educational attainment and income among African Americans with darker and lighter perceived skin tones. To disentangle the drivers of these gaps, we identify all 35,992 families in which enumerators assigned same-gender siblings different Black/Mulatto classifications. Relative to sisters coded as Mulatto, sisters coded as Black had lower educational attainment, were less likely to marry, and had less-educated, likely lower-earning husbands. These patterns are consistent with more severe contemporaneous discrimination against African American women with darker perceived skin tones. In contrast, we find similar educational attainment, marital outcomes, and incomes among differently-classified brothers. Men perceived as African Americans of any skin tone faced similar contemporaneous discrimination, consistent with the "one-drop" racial classification rule that grouped together individuals with any known Black ancestry. Lower incomes for African American men perceived as having darker skin tone in the general population were driven by differences in opportunities and resources that varied across families, likely reflecting the impacts of historical or family-level discrimination.

Translational and helical symmetries provide idealized descriptions of many flows encountered in nature and engineering. While translationally symmetric flows have been extensively studied within kinetic and continuum frameworks, the corresponding connection between microscopic dynamics and continuum behaviour for helical flows is less developed. Here, we investigate this connection for a dilute gas with time-dependent helical symmetry. Using the objective molecular dynamics framework, we derive a reduced Boltzmann equation that retains radial variations while preserving the classical collision operator under helical symmetry. Formal hydrodynamic expansions then yield the corresponding Euler and Navier-Stokes-Fourier equations. The resulting models describe how cylindrical geometry and axial deformation influence transport, including viscous transport and heat conduction, and how deformation generates unequal stresses along and across the axis in competition with collisional relaxation. Comparisons with direct simulation Monte Carlo solutions examine both spatially uniform and radially varying flows. In the cases studied, the first-order fluid corrections capture the leading stress anisotropy and improve the prediction of radial temperature variations, while larger Knudsen numbers expose limitations of the approximation.

When generating verifiable code, natural language requirements are mapped to machine checked code using LLMs and agentic workflows. A crucial component of this pipeline is specification generation (SpecGen), which produces a formal contract against which an agent can prove implementation correctness. Proof generation can obtain deterministic feedback from a theorem prover, but SpecGen lacks a definitive check that a generated specification captures the user's intent. A checked proof can therefore establish correctness against a specification that misrepresents the intended behaviour. We take a step towards holistic SpecGen evaluation with a unified dataset assembled from $350$ existing Lean tasks, including $189$ from VERINA and $161$ from CLEVER, and a framework covering formal validity, reference similarity and equivalence, and behavioural adequacy. We distinguish acceptance of required inputs from acceptance of valid outputs and rejection of invalid outputs, while making each metric's evidence scope explicit. Across four SpecGen configurations, restricting the generalized tree edit distance (GTED) comparison, a reference similarity measure, to $32$ jointly measurable VERINA tasks changes the VERINA configuration's position from second to fourth in mean similarity, showing the importance of measurement coverage. In an authored control, a specification achieves $100\%$ positive test recall and negative test rejection while accepting $0\%$ of required inputs. This demonstrates that perfect postcondition scores can miss an unusable input contract, motivating separate feedback on input coverage and output constraints.

We study the $(n-2)$-plane Kakeya problem in $\mathbb{R}^n$ ($n\geq4$) subject to convex Wolff axioms. In other words, we study finite families of $\delta$-neighborhoods of $(n-2)$-plane segments ($\delta$-flakes) in $\mathbb{R}^n$, with the property that not too many of them can be contained inside a common convex set. Our main result is that the union of $\delta$-flakes from such a set must have almost maximal volume. This can be viewed as a higher-dimensional generalization of Wang and Zahl's work for $\delta$-tubes in $\mathbb{R}^3$. The latter implies the Kakeya set conjecture in $\mathbb{R}^3$. As a consequence, we resolve Rogers' $2$-plane Kakeya problem in $\mathbb{R}^4$ subject to certain curvature conditions on two-dimensional families of direction subspaces in the Grassmannian $\mathrm{Gr}(4,2)$.

The ribosome flow model (RFM) is a nonlinear compartmental model for unidirectional transport along an mRNA chain. We study a generalized RFM in which the classical linear free-space factor is replaced by a monotone nonlinear steric-hindrance response. After recalling the global equilibrium and periodic entrainment properties needed for the analysis, we derive an exact weighted edge-dissipation identity for the gain of entrainment (GOE). This yields a global no-GOE criterion formulated through the curvature of the nonlinear hindrance law. In particular, the quadratic hindrance model admits no positive GOE in any dimension and for arbitrary periodic forcing amplitudes, with equality only when all rates share a common multiplicative modulation. In one dimension, we obtain a sharp convexity-based classification and construct positive-GOE examples when the relevant convexity fails. In higher dimensions, the second variation is governed by a symmetric Jacobi matrix whose diagonal terms encode local curvature and whose off-diagonal entries represent nearest-neighbor transport coupling. This leads to an exact Riccati criterion, a dimension-uniform condition for heterogeneous chains, and a closed-form critical dimension in the homogeneous case. We also exhibit a nonlinear hindrance for which low-dimensional chains have no positive second-order GOE, while a higher-dimensional chain admits a realizable positive GOE. These results reveal a dimension-induced gain mechanism that is absent from the classical RFM.

We show that the radial sector of a Carroll-Weyl gauged null string in Poincaré $\mathrm{AdS}_D$ is locally trivial in BRST cohomology. In a quotient-adapted BFV basis, the radial coordinate, its conjugate momentum, and the associated ghosts form a contractible quartet. An explicit contracting homotopy extends this reduction to the full algebra of local spatial jets, leaving precisely the $(D-1)$-dimensional ILST BRST complex. The same quartet carries zero net Virasoro central charge, so that the anomaly condition of the parent theory, $D=27$, descends to the ILST condition $d=26$.

We propose Tetris3D, a generative framework for single-image 3D scene reconstruction that recovers objects which are physically and geometrically coherent as a scene. Existing methods often generate objects independently or couple them implicitly, providing limited guidance for ensuring fine-grained spatial compatibility between neighboring objects that interact with one another. To address this, we explicitly condition the generation of each object on the geometry of surrounding objects and their physical relationships, guiding its shape and pose to remain geometrically and physically plausible within the scene. Moreover, we introduce ComOb, a physics simulation-based dataset of 1.2M scenes featuring physical interactions across diverse object categories, with per-object meshes and pairwise physical relation annotations. Comprehensive experiments on synthetic and realworld scenes show that Tetris3D recovers coherent object shapes and poses even when interacting regions are occluded, and achieves state-of-the-art performance in both generation quality and physical stability.

As we move through the world and carry out everyday tasks, we encounter objects that may become relevant only later. We are capable of recalling where we left something or what was inside a container, even without knowing we would need it later. Here, we study how an embodied assistant can build a similar memory from egocentric videos, by observing a person's day-to-day activities. We present Ledger, a persistent 3D object memory that combines object locations, their histories, and contextual descriptions. It associates observations across the recording and retains objects after they leave the view, including those the person never touches. It clusters each object's observations by resting locations and records a move only after repeated evidence, reducing the effect of localization noise. Short descriptions preserve details such as an object's contents or supporting surface. It saves these records to later answer spatial questions without having to access the original images or video. Our memory raises HD-EPIC accuracy from 29.7% to 42.6%, UCS-Bench accuracy from 33.8% to 38.5% and localizes Ego4D objects with a 0.99 m median error on returned predictions. Our analyses identify complementary roles for temporal persistence, contextual descriptions, and retrieval. Our study on 100 stitched streams of multiple scenes each further exposes failures in both retrieval and construction. Per-scene construction partially recovers the performance lost across scene changes compared to that of single scene streams.

For every positive integer $n$, we prove that there is a constant $C_n$, depending only on $n$, such that $Vol(-K_X) \leq C_n \epsilon^{-(2^n-n-1)}$ whenever $X$ admits an $\epsilon$-lc log Fano boundary. The exponent is optimal in every dimension at least two, even for toric Fano varieties of Picard number one. In particular, the optimal exponent for fourfolds is eleven. We also prove an anticanonical interpolation theorem with optimal exponent $2^n-1$. The proof combines signed discrepancy estimates under projection, finite morphisms to projective space, and the canonical bundle formula. A reduction to a base bounded independently of $\epsilon$, followed by a volume estimate along a flag, yields the sharper volume exponent.

Modern language models undergo reinforcement learning with verifiable rewards (RLVR) on top of already-trained checkpoints. A key promise of RLVR is the discovery of new reasoning strategies. In principle, a model can sample novel ideas absent from its prior training data. In practice, however, augmenting RLVR with strong novelty incentives has seen limited success and can degrade model quality. Because verifiable rewards supervise only a narrow slice of the model's knowledge and behavior, such degradations are difficult to recover from. Instead, we decouple exploration from optimization in a framework we call Exploration-Distillation (ExpDis). We train one or more explorer policies with a novelty bonus in the reward, filter their trajectories for correctness and quality, and distill them into a separate student policy. The student policy is then trained without a novelty bonus. We repeat the above procedure for several rounds, alternating between exploration and optimization. This decoupling allows us to aggressively scale exploration without degrading the student policy. Across seven mathematical reasoning benchmarks and two model families, ExpDis outperforms DAPO at the same wall-clock budget. Moreover, we observe improved pass@$k$ scaling, indicating that ExpDis produces models that generate more diverse correct solutions.

We introduce a flexible mixture model for the joint source-frame component-mass distribution of compact binaries, inferred simultaneously with cosmology, together with a framework that identifies structures in the reconstructed density, quantifying their posterior stability, local prominence, and cosmological relevance. Applied to GWTC-5.0, the workflow clarifies several familiar and emergent structures. The $\sim35~M_\odot$ feature is classified as a clear near-equal-mass peak despite an ambiguous interpretation in the primary-mass projection. A weaker unequal-mass concentration near $(21,12)~M_\odot$ is recovered frequently but with lower prominence. A tentative joint transition near $60~M_\odot$ appears at mildly unequal masses, with a more robust manifestation in the primary-mass projection. For cosmology, the two-dimensional mass model constraint on $H_0$ is $33\%$ narrower than a model that flexibly reconstructs only the primary mass, and only $\sim~10\%$ wider than analyses based on strongly parameterized one-dimensional spectral structure. At the feature level, we show that the $\sim10~M_\odot$ and $\sim35~M_\odot$ concentrations depend on $H_0$ mainly through a common scale which captures the entire $\sim~(1+z)$ cosmological rescaling of both masses; the intermediate unequal-mass feature instead retains residual dependence in the orthogonal mass-ratio direction. More generally, we argue that population prominence, formal inclusion in an astrophysical feature catalogue, and cosmological relevance need not coincide. These results suggest that a genuinely two-dimensional view may be required to fully exploit spectral-siren information. Our workflow provides a reproducible way to detect population structure and identify which aspects of its geometry are associated with spectral-siren constraints. The feature finder is directly applicable to other population reconstructions.

End-to-end robot policies trained through imitation learning remain constrained by limited data diversity, making reliable zero-shot deployment in real-world settings challenging. Shared-autonomy methods enable human correction through teleoperation, but specialized hardware and operator training hinder deployment at scale. Other approaches incorporate human guidance as additional policy inputs, often requiring architectural changes and dedicated training for steerability, which limits their applicability across policies. We present RoboPrompt, a general-purpose, lightweight robot policy steering system that enables users to guide policy behavior through intuitive, sparse inputs, including drawn traces, target points, and coarse directional instructions. RoboPrompt decouples human-intention translation from the underlying policy: a reusable module converts human guidance into action drafts, which are refined through the diffusion or flow-matching dynamics of the base policy. By controlling action generation in noise space, RoboPrompt balances human intent with the policy prior without modifying the base policy architecture or fine-tuning it for steerability. Experiments demonstrate effective steering across Diffusion Policy, $\pi_{0.5}$, and FastWAM. We further use steered rollouts for online policy improvement through DAgger. After 2-3 rounds of iteration, average success rates increase by 15.5% for $\pi_{0.5}$ across three tasks and by 21.3% across three policies(Diffusion Policy, $\pi_{0.5}$, FastWAM) on the Insert Bread task, while average human intervention counts decrease by 44.0% (2.86 to 1.60) and 81.9% (2.60 to 0.47), respectively.

Conditional memory architectures such as DeepSeek Engram use input n-grams to look up learned embeddings, expanding the capacity of large language models (LLMs) with limited additional computation. Beyond model scaling, this architecture has demonstrated the potential to decouple factual knowledge storage from general-purpose computation, offering a promising route to updating factual knowledge while keeping the Transformer backbone fixed. Realizing this potential is challenging because different expressions of a fact may activate different n-gram embeddings, while updating shared embeddings can unintentionally change the model's predictions about other facts. We propose EngramEdit for decoupled knowledge updates through conditional memory. EngramEdit first computes target memory representations that make the model predict the updated fact across multiple expressions. It then jointly updates the shared n-gram embeddings to match these targets across expressions and edits, penalizing updates to frequently reused embeddings more strongly to preserve unrelated knowledge. Experiments show that EngramEdit enables independent factual knowledge updates through conditional memory, achieving near-perfect editing success. Revised knowledge is usable across unseen expressions and in multi-hop reasoning, with nearly three times the strongest baseline's accuracy under chain-of-thought (CoT) prompting. Unrelated knowledge and general capabilities are largely preserved even as factual updates accumulate. These findings show that EngramEdit turns conditional memory into an editable knowledge interface, extending its role beyond model scaling to support decoupled knowledge updates.

In this paper, we consider the problem of predictive rerouting of connected and automated vehicles (CAVs) in mixed traffic with electric vehicle charging demand. We provide a framework that combines a diffusion convolutional recurrent neural network (DCRNN) with a routing policy that accounts for congestion, CO2 emissions, route length, and charging demand. The DCRNN uses historical network observations to forecast traffic conditions and charging demand. These forecasts are then used to evaluate feasible alternative routes for eligible CAVs. A route change is accepted when the alternative preserves connectivity to the original destination and improves the prescribed route cost. We evaluate the proposed framework in SUMO under controlled traffic disruptions at five CAV penetration levels, ranging from 5% to 45%. We compare its performance with K-shortest-path routing, predictive-density routing, V2X proactive routing, and a reference scenario without rerouting. In the considered scenarios, the proposed framework reduces the average travel-time index by approximately 1.8% relative to the reference scenario and achieves the lowest average travel-time index, highest average speed, and lowest aggregate CO2 emissions among the active routing methods. Across all five penetration levels, it accepts 41 route changes, compared with 146 for K-shortest-path routing and 153 for V2X proactive routing. The reference scenario retains lower aggregate emissions and distance traveled, illustrating the tradeoff between congestion reduction and the additional travel associated with rerouting.

Quantum error-correcting circuits are affected by multiple physical noise mechanisms, whose contributions to logical failure must be understood to evaluate code performance and guide improvements in hardware. The individual error budget contributions for an error type $i$ can be characterized by its logical sensitivity $\nu_i = \frac{\partial p_L}{\partial p_i}$, which measures the response of the logical error rate $p_L$ to each physical noise parameter $p_i$. Normally, $\nu_i$ is measured using linear fits such as finite differences, where $p_L$ is measured at two or more values of $p_i$ to calculate partial derivatives. In this paper, we develop a differentiable estimator to obtain all components of $\nabla_{\mathbf p}p_L$ simultaneously from a single Monte Carlo data set at one noise configuration, using information about the underlying fault configurations. In surface code simulations with circuit-level noise, the estimator agrees with conventional finite differences while requiring one to two orders of magnitude fewer shots to achieve the same variance. We apply this technique to quantify error budgets, resolve sensitivities at the individual qubit level, and infer effective code distance. Finally, we incorporate the sensitivities into Newton root finding to locate and trace threshold contours in multidimensional noise models. These results provide an efficient method for extracting and applying logical sensitivity information from standard quantum error correction simulations.

We constrain oscillatory features in the primordial power spectrum using the public DESI DR1 galaxy and quasar clustering data. Our analysis is based on a custom pipeline for the redshift-space power spectrum and bispectrum of the six main DESI DR1 tracers, modeled with the Lagrangian effective field theory,which provides a principled resummation of the non-linear damping of the features. We consider oscillations that are linear and logarithmic in wavenumber, and, for the first time, consistently include the galaxy bispectrum in the search. We also study the dependence of the results on the background cosmology, which is important given the tension between the DESI BAO and Planck data within the baseline cosmological model. Marginalizing over $\Lambda$CDM parameters, we bound the amplitude of linear and logarithmic oscillations to $A_{\rm lin}<0.033$ and $A_{\rm log}<0.035$ respectively (at 95$\%~$CL) across the frequency ranges $\omega_{\rm lin}^{\rm phys}\in(20,800)~{\rm Mpc}$ and $\omega_{\rm log}\in(2.5,80)$. The constraints reach percent-level precision away from the frequencies degenerate with the baryon acoustic oscillations (BAO). Fixing to the Planck 2018 cosmology, the bounds tighten to $A_{\rm lin}<0.024$ and $A_{\rm log}<0.031$, but in this case we also find a $\simeq 3\sigma$ global preference for a logarithmic feature with $\omega_{\rm log}\simeq 15$, dominated by the LRG2 sample, whose local wavelength matches the apparent BAO scale at wavenumbers best constrained by DESI. Since the preference for this feature disappears once the cosmological parameters are varied, we expect that it is a manifestation of the known distance-scale tension between the DESI and Planck data rather than a primordial signal. Our analysis thus emphasizes the importance of cosmology marginalization in searches for primordial features.

We establish a charge-regular formulation of separated Kerr–Newman perturbations and an explicit map from master solutions to physical radiative fields. We derive regular angular and radial first-order systems and their second-order reductions, with finite-charge deformations of the spin $\pm1$ and $\pm2$ Teukolsky equations. The field and boundary normalizations recover the standard Kerr conventions while accounting for the nonuniform zero-charge limit at infinity. For physical Einstein–Maxwell perturbations with the usual peeling behavior, we obtain explicit outgoing gravitational and electromagnetic amplitudes and energy fluxes at nonzero real frequency. The resulting finite-charge radiation-composition ratios depend only on the regular angular functions, independently of the radial excitation.

Real-time robot control demands enough visual history to infer motion and task progress, but processing that history can delay action. We present Long-WAM, a model-system framework for scaling the context of causal world-action models under real-time control constraints. Our central finding is that access to history is not the same as using it: longer histories pay off far more when the video foundation is pretrained autoregressively (AR). We first learn causal prediction from robot and egocentric videos without action labels, then preserve this history-to-future structure during world-action adaptation. On RoboCasa GR-1, increasing context from 0.0 to 19.2 seconds raises success from 63.3% to 78.7%, whereas a bidirectionally pretrained initialization shows no net gain; robot-domain AR pretraining further raises peak success on GR-1 and LIBERO-Long. Long-WAM also achieves the best results among compared methods on LIBERO-Long, RoboTwin 2.0, and DOMINO. Streaming observation encoding, asynchronous execution, and hardware-specific acceleration enable deployment on RTX 5090, DGX Spark, and Jetson AGX Thor without dropping future prediction; on RTX 5090, each action chunk, including future-video latent prediction, takes 107.4 ms. Real-time deployment on Unitree G1 and YAM supports dynamic and long-horizon manipulation, including 95% success on dynamic cup stacking, where Pi0.5 and Fast-WAM succeed in none of 20 trials. As a memory-informed executor, Long-WAM also complements higher-level planning in composite tasks.

Heavy-tailed noise has been widely observed in modern machine learning, motivating the use of methods like gradient clipping and normalization. While these methods are well understood in centralized settings, much less is known in decentralized ones, where applying a nonlinearity to local gradients affects both optimization and consensus. Recent works on decentralized non-convex optimization have studied both clipping and normalization under heavy-tailed noise, with clipping yielding suboptimal rates and normalization needing local momentum or mini-batches to converge. This raises the question: can a baseline decentralized method using a nonlinearity achieve optimal convergence rates under heavy-tailed noise? We answer affirmatively with clipped decentralized SGD ($\mathtt{DSGD}$). For smooth non-convex costs under bounded $p$-th moment noise, $p \in (1,2]$, we show that clipped $\mathtt{DSGD}$ achieves order-optimal rates both with high probability and in expectation. Moreover, we establish a linear speed-up in the number of agents, which, to our knowledge, has not been shown for decentralized methods with clipping. The key technical ingredient is a sharp analysis of the consensus gap that exploits the structure of clipping, relegating network effects to higher-order terms. Our results highlight an important distinction between clipping and normalization in decentralized settings: while normalized $\mathtt{DSGD}$ can fail to converge, clipping retains magnitude information, enabling $\mathtt{DSGD}$ to be convergent and order-optimal. Numerical experiments validate our theory.

Vision-language-action models (VLAs) are strikingly sensitive to instruction phrasing and do not inherit the language robustness of the vision-language models they are built on. A one-word edit can move success by tens of points: $\pi_{0.5}$ turns on a LIBERO stove 100% of the time for "switch on the stove" and 2% for "switch on the hot plate", and a $\pi_0$ checkpoint finetuned with rephrase augmentation still shows swings of up to 61 points. We characterize this sensitivity with statistically tested single-edit swings and an oracle phrase search, which shows that phrasing alone nearly closes the 21-point gap between in-distribution and out-of-distribution tasks. We then reduce it without modifying the policy. Because the sensitivity is systematic, it can be expressed as explicit rules: we score many phrasings of a few training tasks, have a large language model distill the evidence into ten to twenty rephrasing rules, and at deployment rewrite each incoming instruction once under these rules. The rules improve the frozen $\pi_0$ by 16 to 27% relative on twelve held-out tasks across adversarial, VLM-generated, and human-generated phrasings, with gains concentrated on out-of-distribution tasks. The pipeline replicates on $\pi_{0.5}$ and LIBERO, lifting in-finetune success from 93.6% to 97.8%. The method requires no retraining and no per-step verification, and applies zero-shot to unseen tasks and instructions. Project website: https://sttawm.github.io/rephrase-before-you-act