Papers

We report fine-grained, deterministic instability in Dynamic Tensor Rematerialization (DTR), an online eviction policy for memory-constrained DNN training, measured on the reference DTR simulator (simrd) using public execution traces. On an LSTM trace, memory budgets differing by 0.10% of unconstrained peak memory select fast and slow execution regimes whose overheads differ by as much as 7.3x; the slow regime is driven by broadly repeated re-eviction of the same storages (evictions per storage rise from 1.33 to 8.27 while the set of distinct evicted storages is essentially unchanged: 5,233 vs 5,236, with the two sets overlapping at Jaccard 0.999). On a ResNet-32 trace, a fine budget sweep reveals a deterministic feasibility inversion: the run is feasible at ratio 0.101, infeasible (OOM) across 0.102-0.106, and feasible again from 0.107. We trace the immediate cause of the OOM to a fully pinned recursive rematerialization frontier that exceeds the budget after every evictable tensor has been evicted. Ablations using the DTR authors' own variants implicate the joint size-staleness scoring term in the observed LSTM instability. We argue these are at least two distinct budget-sensitive pathologies rather than one mechanism, and we separate what is demonstrated from what remains hypothesised. All results concern the reference simulator; reproduction in a production runtime is future work. Code, instrumentation, and raw results accompany this preprint.

We study the right derivative of the zeta-star correspondence between binary expansions and multiple zeta-star values of infinite depth. At a finite-index point, this derivative is a normalized difference of two finite-depth multiple zeta-star values. We use this identity to express its local behavior in terms of lower-truncated star sums and their boundary coefficients. The homogeneous contribution at summation variable three contracts with ratio $2/3$; the full coefficient also receives a positive forcing term. A weighted sum of the remaining coefficients contracts by at most $1/2$ at each binary step. On the open parameter interval, we prove that the derivative is right-continuous everywhere and continuous precisely at the non-dyadic points. Its left jump at a dyadic point is given by an explicit positive multiple series, with divergence exactly at $1/2-2^{-M}$, $M\ge2$. A refinement of the index gives an infinite upper right Dini derivative at every interior point. We also prove that the Hausdorff, packing, and modified upper box dimensions of the graph are all $\log_2(8/3)$. The Hausdorff lower bound uses the dimension of the Bernoulli convolution with parameter $2/3$, together with a separate measure-transfer argument. Finally, the pointwise Hölder exponent at a non-dyadic point is $\log_2(3/2)$ divided by its dyadic approximation exponent, and the corresponding Hausdorff spectrum is linear.

3D Gaussian Splatting reconstructs a scene as a collection of Gaussian primitives from a set of posed photographs called the capture. The number of primitives used to represent the scene affects reconstruction quality, storage, and rendering cost. How to select this number automatically across capture scales remains unresolved: configurations effective on standard benchmarks can leave larger captures with too few Gaussians to reconstruct fine details. We observe that the surface to represent, given by the capture's extent and resolution, is known before training, whereas its content complexity becomes apparent during training, through the reconstruction quality on the training views. We introduce TangoGS, which combines capture-derived model sizing with training-based adaptation: the capture determines the scale of the model, and training feedback determines its final size within that scale. Before training, TangoGS derives a learning allowance for model growth from the capture's total pixels after discounting views that re-observe the same scene points. During training, reconstruction quality guides how many Gaussians to add and remove. On 13 standard benchmark scenes, TangoGS matches the mean PSNR of the best-performing evaluated baseline, LeGS, with $48\%$ fewer Gaussians. On eight large captures, the same configuration automatically scales to larger models when necessary, achieving the highest mean PSNR among evaluated methods: $0.54$ dB above the runner-up with $2.3\times$ as many Gaussians. Together, capture-derived learning allowances and training-quality guided density control enable a state-of-the-art quality–size compromise across scene scales without retuning.

Novel view synthesis (NVS) models can produce realistic new views of the same scene from different viewpoints. However, these generated views are not always geometrically consistent with one another. Multi-view (MV) consistency has shown promise as a tool for evaluating these NVS models. Its potential for multimedia forensics, however, remains largely unexplored, particularly for localizing geometric inconsistencies across wide-baseline image pairs. To enable research in this direction, we introduce DeformView, a wide-baseline MV dataset with pixel-level annotations of geometric inconsistencies. Using DeformView, we evaluate state-of-the-art MV consistency-scoring methods and show that approaches developed for NVS evaluation transfer poorly to the forensic task of geometric inconsistency localization. To address this limitation, we propose DEFECt3R, a lightweight learning-based classifier that uses cross-view feature relationships to localize geometric inconsistencies at the pixel level. By learning from explicit supervision, including hard negatives from geometrically consistent yet deformed views, DEFECt3R improves localization performance and substantially reduces false positives compared to existing consistency-scoring methods. Ablation experiments further show that both feature representations and correspondence quality contribute to localization performance. Overall, our findings demonstrate that MV geometric consistency is a promising yet underexplored signal for multimedia forensics and establish a benchmark and baseline for geometric inconsistency localization in wide-baseline MV image pairs. Code and dataset are available at https://github.com/IDLabMedia/DeformView-DEFECt3R

The semiring framework and its extensions form the basis of a rich collection of theoretical results and implementations for provenance tracking of database queries. Many real-world queries use aggregation and conditions on the aggregate values. Support for such queries has been proposed by introducing semimodule elements as aggregate values and formal comparisons between aggregate values as tuple annotations, which takes the approach outside the standard semiring framework. In this work, we show how to introduce a semantics for the provenance of such queries in arbitrary commutative semirings with monus (or m-semirings), without the need for additional operators. This semantics is shown to agree with the standard provenance of the aggregation-free self-join rewriting of HAVING COUNT(*) queries in semirings that are absorptive and where times distributes over monus. We derive algorithms for this semantics and implement them within the ProvSQL system, with viable performance on a real-world dataset for probabilistic query evaluation.

Individuals turn to large language models (LLMs) for guidance across a wide range of economic tasks, from comparing loan options and planning savings to deciding what raise to ask for or how much to charge for their services. LLMs are known to reproduce social biases, and biased economic guidance may influence what users believe they are worth, what they ask for, and what they ultimately accept. This risk is especially salient in India, where economic outcomes are shaped by demographic categories such as caste and urban-rural location. Existing LLM bias benchmarks, however, are largely designed around Western demographic categories and therefore miss key axes of economic disparity in the Indian context. We introduce RupeeBias, a benchmark for auditing demographic bias in LLM-generated economic guidance across Indian economic settings. RupeeBias consists of 39,150 prompts spanning four use cases: salary estimation, salary increment estimation, counter-offer recommendation, and service pricing recommendation. The benchmark follows a single-attribute counterfactual design, holding the description of the user's qualifications, experience, or service offering fixed while varying one demographic identifier at a time. RupeeBias covers 87 India-specific demographic identifiers across six axes: caste, religion, regional identity, gender, disability, and urban-rural location, with all prompts constructed in both English and Hinglish. We evaluate nine LLMs on RupeeBias and find systematic demographic disparities across all six axes. For otherwise identical prompts that differ only in demographic identifier, LLM-generated economic outputs differ by 20.2% on average. We publicly release RupeeBias to support future research on demographic bias in LLM-generated economic guidance across India-specific demographic and economic contexts.

Mean motion resonances (MMRs) are traditionally thought to form during the gas-rich phase of protoplanetary disks via disk-driven migration. Whether MMRs can form solely through collisional damping after disk dispersal has not been systematically investigated. To study MMR formation in the post-gas-disk era, we perform a large suite of N-body simulations, each containing 10,000 planetesimals around an M-dwarf star and evolving over 10 Myr. We compare two groups of simulations that differ only in their initial mass distributions. Each group consists of 10 runs with statistically identical initial planetesimal orbital distributions, generated using different random seeds. One group adopts a uniform planetesimal mass distribution, whereas the other features a bimodal distribution with massive embryos embedded in a swarm of smaller planetesimals. At 1 Myr, 40% of systems (8/20) exhibit near-resonant period ratios, but this fraction drops to 25% (5/20) by 10 Myr as higher-order commensurabilities dissolve. Resonant-angle tracking shows that all persistent systems have circulating resonant angles at 10 Myr, although one system undergoes a transient libration phase between about 2 and 7 Myr before returning to circulation. These results demonstrate that collisional merging can drive MMR formation even without a gas disk, although in this scenario near-resonant configurations are strongly favored over true resonances. Furthermore, this mechanism produces similar MMR fractions in both groups, suggesting that MMR formation is insensitive to the initial mass distribution, although the latter may modulate the formation pathway to a limited extent.

Let $P$ and $Q$ be proper upward monotone unary generalized quantifiers on a fixed domain $D$. We study the validity, over all structures $M$ with domain $D$, of \[ M \models Px\,Qy\,R(x,y) \rightarrow Qy\,Px\,R(x,y), \] called scope dominance. The iteration of two quantifiers lies between their branching product, which is generated by rectangles, and the dual of the branching product of their duals. We give criteria for equality with these bounds, obtain a new proof of the countable characterization of scope dominance, and characterize the case in which the inner quantifier is "at least $\kappa$ many." Under the Generalized Continuum Hypothesis, Goldberg's theorem shows that, for ultrafilters, scope dominance is equivalent to equality between iteration and the branching product, whereas an elementary counterexample on a domain of size $\aleph_2$ shows that this equivalence fails for arbitrary upward monotone quantifiers. For the case where both quantifiers are filters, the question becomes set-theoretic: the existence of a counterexample is equiconsistent with the existence of a measurable cardinal.

In this paper, we consider the local minimizers to a class of non-autonomous functional with double-phase Orlicz growth: \begin{equation*} u \in W^{1,1}(\Omega; \mathbb{R}^N) \mapsto \int_\Omega \Big( {\varphi_1}\big(|Du|_{\mathbb{A}^u}\big) + a(x)\,{\varphi_2} \big(|Du|_{\mathbb{A}^u}\big) \Big)\, dx, \end{equation*} where $|Du|_{\mathbb{A}^u} := \big\langle \mathbb{A}(x,u)\,Du, Du \big\rangle^{\frac{1}{2}}$ with $\mathbb{A}(x,u) = \big\{A^{\alpha \beta}_{ij}(x, u)\big\}_{i,j = 1\cdots N}^{\alpha,\beta = 1\cdots n}$ as the uniformly elliptic bounded symmetric tensor field, and $\varphi_k(\cdot)$ for $k=1,2$ are two different $N$-functions. We prove the partial regularity of their local minimizers if the continuity modulus of $a(\cdot)$ and $\varphi_k(\cdot)$ are fit for the gap conditions, and $\mathbb{A}(x,u)$ meets an optimal regularity in $(x,u)$. This is an extension from the single Orlicz growth functional to the double-phase one based on an essential improvement of several important inequalities.

We show experimentally obtained frequency- and power-dependent optical response of a room-temperature cesium vapor cell used as a Rydberg microwave receiver based on electromagnetically induced transparency. An applied microwave field couples the ${70S_{1/2}}$ Rydberg state to the nearby ${70P_{1/2}}$, ${70P_{3/2}}$, ${69P_{1/2}}$, and ${69P_{3/2}}$ states, modifying the probe-transmission spectrum. Probe transmission is recorded over a 800 MHz range of coupling laser detunings. These EIT spectra are measured for RF frequencies from 9 to 13.5 GHz and over a 40 dB range of RF power. We analyze the RF induced splittings in the recorded EIT spectra. We observe both limiting cases, where the EIT peak separation is either dominated by the RF detuning or by the RF power. At higher powers we are also able to observe complex features due to coupling of multiple nearby Rydberg states. We further analyze the probe-transmission signal at zero coupling-laser detuning. Sigmoidal fits to the power-response curves are used to determine the midpoint power and the 10-90% transition interval as functions of RF frequency.

Computing non-dilute diffusion coefficients from Onsager transport coefficients of a multicomponent alloy requires the thermodynamic factor, which measures the curvature of the free energy with respect to composition. As diffusion in metals is mediated by vacancies, the curvature that matters is not that of the vacancy-free alloy, which is straightforward to compute, but that of an alloy carrying a dilute concentration of vacancies. The resulting matrix is nearly singular, and composition fluctuations from which it is obtained converge slowly in Monte Carlo simulations. Here we develop two routes that avoid sampling vacancy composition fluctuations. The first constructs the thermodynamic factor from two quantities, the free-energy curvature of the vacancy-free alloy and the vacancy concentration. The second truncates the semi-grand canonical partition function at a single vacancy and recovers the full thermodynamic factor of the vacancy-containing alloy. The same separation shows that the largest eigenvalue of the diffusion matrix is the vacancy tracer diffusion coefficient for any number of components and any degree of non-ideality, while the remaining eigenvalues scale with the vacancy concentration. Across binary through quinary alloys of the Hf–Mo–Nb–Ti–Zr system, both routes reproduce eigenvalues and eigenvectors of diffusion matrices obtained from fully converged Monte Carlo simulations. The single-vacancy expansion reaches this accuracy with roughly one tenth of the sampling effort. A scaling relation extends both routes to local vacancy chemical potentials away from equilibrium, without additional simulations. These results allow thermodynamic descriptions from atomistic models or CALPHAD assessments to be used directly in mesoscale simulations of mass transport.

We investigate chiral transfer and dissipative generation of even-parity Bell states in a two-qubit system with coherent two-photon driving and engineered two-photon loss. We find that adiabatic encirclement of a second-order exceptional point induces direction-dependent transfer between the Bell states $|\Phi^+\rangle$ and $|\Phi^-\rangle$, governed by a time-integrated low-loss branch-selection mechanism. We develop a hybrid-Liouvillian description with a control parameter $q$ that interpolates between conditional no-jump dynamics and unconditional Lindblad evolution, and use it to assess how exceptional-point-induced chirality survives in the presence of quantum jumps. When the same parameter loop is initialized in the separable state $|00\rangle$, it directly generates strong even-parity entanglement. Together, these results extend dissipative Bell-state control beyond the single-excitation manifold. They further demonstrate that exceptional-point-based protocols can unify chiral state transfer and entanglement generation within a single engineered two-photon platform.

Scalar travel-time relations can identify propagation geometry without velocity inputs, ray labels or equation supervision during learning. Physical interpretation is applied after learning through the high-frequency isotropic eikonal relation (Aki and Richards, 2002). Here we show that local value differences constrain derivatives through sampling geometry and smoothness. Controlled neural fields recover propagation directions with $5.34\pm0.07^{\circ}$ median angular error, and local P- and S-wave velocities in a prospectively locked, test-sealed realization. Classical representations also recover geometry. Scalar-matched perturbations separate value accuracy from derivative reliability; regularized inversion can recover structure when direct gradients fail. A field trained on 2.90 million Chinese arrivals retains continental crust–mantle boundary structure: correlation with a withheld reference is $\rho=0.910$ raw and 0.592 after quadratic detrending (spatial-shift $p=0.020$). Direct catalogue values yield $\rho=0.909$, and thickness contrasts are compressed. Manual-only California arrivals support broad velocity organization with limited lateral fidelity. Synthetic quantum fields containing phase information support probability-flow and trajectory readouts without supervising those quantities. These results establish conditional identification across representations and physical settings: informative scalar relations can constrain geometry, while known physical relations supply its interpretation. They do not imply that accurate scalar fitting alone guarantees reliable physical derivatives.

In critical limits of statistical models, there exist non-local correlators that are conformally covariant without belonging to a conformal field theory. To describe such correlators, we define conformal correlator systems, whose axioms are weaker than those of CFT. As a result, $N$-point correlators are determined by $4$-point correlators, instead of $3$-point correlators in CFT. In two dimensions, conformal spins may take arbitrary complex values, leading to correlators with nontrivial monodromies. We show that sphere $4$-point correlators and torus $1$-point correlators obey nontrivial monodromy constraints. We construct correlators with abelian monodromies in free bosonic theories and in critical loop models.

The return migration of skilled workers provides origin countries many important benefits. The likelihood of return migration is thought to depend on relative economic opportunities, but quantitative evidence of its relative importance remains limited. Here we measure return migration rates using the public profiles of software developers on GitHub, geocoded across ten waves between 2012 and 2026, in a panel containing more than 270,000 international movers. Within five years of departure, 8.8% of movers have returned. Origin-country rates range from under 1% to 16%. Hazard models with corridor fixed effects show that migrants whose home countries experience greater catch-up growth return at significantly higher rates. Immutable features of country pairs such as cultural or geographic distance do not explain the estimate. We forecast future return rates using IMF growth projections, predicting, for example, that India's return rate will rise 12% by 2035 and China's 7%. The difference is driven mostly by India's faster projected growth at home. As developing economies close income gaps with the destinations their emigrants have settled in, they recover more of the workers they lost.

Artificial neural networks (ANNs) usually represent neural transmission with fixed trainable weights during a training batch, which omits short-term changes in synaptic efficacy. In addition, the discrete time-step simulation requires additional temporal processing that many conventional ANN architectures do not use. To overcome these challenges, we propose Purin, a biology-inspired and ANN-compatible mechanism, that introduces synaptic efficacy modulation into conventional convolutional neural networks. Purin uses a time-interval-based abstraction for neural activities, which allows Purin to introduce short- and long-term synaptic efficacy changes without using discrete time-steps. Purin introduces a bounded factor to represent temporary synaptic efficacy changes, together with two weight matrices that represent input-side and output-side efficacy. The weight matrices are updated by backpropagation and interpreted as the long-term synaptic efficacy changes. Experimental results show that after removing the confounding factors in the AlexNet, VGG11, and GoogLeNet architectures, Purin improves the classification accuracies in all three models across the evaluated datasets.

Problem. Ultra-high-resolution (UHR) remote sensing with vague user intents has two bottlenecks: visual tokens are expensive, and tool calling must be format-reliable (pretrained models emit zero tool calls zero-shot). Method. WeaveAgent, a two-stage tool-routing agent, decouples routing from visual perception. Stage A is routing-first: emission is trained, not elicited. Stage B executes conditionally: intrinsic queries enter visual answering (full-scene thumbnail; a WeaveEarth-style evidence board as an optional fixed-budget, approx. 5k-token compression interface); extrinsic queries execute tool call on original full-resolution imagery, answering from tool observations in a second, observation-masked round. Training: alignment SFT, then GRPO under reward R_WA2. Results. Alignment SFT lifts extrinsic routing from 0% to 80.75% (323/400); GRPO suppresses 9 intrinsic mis-emissions while tool selection is unchanged. The trained 2B system does not beat the zero-shot 8B baseline overall (0.263 vs. 0.250), a diagnostic contribution. Oracle attribution separates two repair ingredients: loading the observation into context lifts extrinsic answer accuracy from 0.025 to 0.425 under marker-free cross-mode returns, and the two-turn SFT stage adds a further +9.3 points to 0.518 at a small routing cost. A +/- image ablation shows emission suppression is visually grounded, and a query-register matrix shows LLM-rewritten queries cost trained checkpoints 2-11 points. Scope. All training and evaluation use the 5,000 / 3,273 / 1,000-record VagueUHR corpus (600 intrinsic + 400 tool-requiring; the base seeds synthesis and is not used for optimization). Single-pass evidence construction runs at 7.31 s per image on an RTX 4090. Code, data, and evaluation protocols will be released.

We prove that the Pólya-Szegő inequality for the $L^p$ norm of the Riesz fractional $s$-gradient fails for every $p\in[1,2)$ and $s\in(0,1)$, thus providing a negative answer to a question posed by Nguyen and Squassina in 2017.

Bringing an aerial robotics method from simulation to flight should not require rebuilding a mature autopilot or adding a companion computer. Cybflight is an open-source embedded Rust research autopilot whose typed, replaceable interfaces connect hardware access, perception, state estimation, trajectory planning, and control. This modular development and compile-time optimization workflow is demonstrated with replaceable Rust implementations of model predictive contour-tracking control (MPCTC) and incremental nonlinear dynamic inversion (INDI) running on one STM32H743 without a companion computer. Using this configuration, the vehicle reaches 12.38 m/s during indoor flight, while an outdoor flight using global navigation satellite system (GNSS) position updates reaches 31.4 m/s. These flights show that running demanding estimation and nonlinear control entirely on a flight-controller microcontroller need not come at the expense of a modular autopilot structure.

In this paper, we study the elliptical shape of the numerical range of tridiagonal matrices of orders $3,4$ and $5$ with zero entries on the main diagonal, a positive constant value $a$ along the super-diagonal, and sub-diagonal entries are \begin{align*} \begin{cases} r, sr^2, r^3, sr^4,\ldots, r^{n-1} & \mbox{ if $n$ is even, and } r, sr^2, r^3, sr^4,\ldots, sr^{n-1} & \mbox{ if $n$ is odd}, \end{cases} \end{align*} where $n$ denotes the order of the matrix, $s>0$ and $r$ is a real number. We further investigate the flat portions on the boundary of the numerical range of these $n\times n$ matrices in the special case $r=-1$. Some results given in \cite{chien2011numerical} are obtained as a particular case of our results.

Convergence in parameter estimation classically requires “persistency of excitation," which is a non-degeneracy condition on a trajectory-dependent signal. To the best of our knowledge, this paper develops the first such excitation theory for a nonlinear PDE. In the context of identifying the unknown viscosity from spectral observations in the two-dimensional Navier–Stokes equations for incompressible fluids, our excitation condition is computable, verifiable a priori, and sharp with respect to scaling. Our approach employs a data assimilation methodology to account for an unknown initial state and select candidate viscosities by minimizing the loss between the observed low-modes of the fluid velocity and low-modes of a nudging-based filter assimilating these observations. The main novelty of our framework is to distinguish a new functional, $\mathsf{W}$, representing the work done by the filter's associated sensitivity variable on the enstrophy, around which our entire analysis is centered. We show that $\mathsf{W}$ is explicitly comparable to the observability Gramian of the associated Gauss–Newton iteration, and subsequently establish the following dichotomy at every critical point of the observational loss: either $\mathsf{W}$ exceeds a minimal excitation threshold, in which case the candidate viscosity obeys an explicit error estimate that depends inversely on $\mathsf{W}$ and the observational density $N$, but directly on the error between initial conditions, or else the observations are quantitatively insensitive to parameter updates in a way that is detectable to the user. Our approach is energy-based, and therefore expected to be adaptable to many other nonlinear dissipative systems.

Structural learning and physical-property inference can draw on different sources of knowledge. We learn morphology from 33,140 natural, satellite, and texture images, without geological training models, then infer Earth properties from physical observations. Identical frozen weights map 128 coordinates into each $256\times256$ compressional-velocity, shear-velocity, or density field. Fixed adapters assign physical meaning; observation-specific likelihoods determine new posteriors without retraining. Across 256 unseen Earth models, image coordinates reduce projection error fourfold relative to a dimension-matched cosine basis. Matched inversions favor image coordinates for the tested curved-velocity targets and a smooth radial basis for the tested faulted targets. A simplified 2.5-dimensional application with 915 measured observations along one 120.9-km southern-California profile constrains a velocity–density contrast. Grouped arrival-time and gravity errors fall by 43% and 44% relative to the pretrained image prior, although the matched smooth radial basis predicts better. Direct layered-Rayleigh updating reduces blind phase-velocity error by 28% at five sites, whereas P-and-gravity-only sharing worsens it, distinguishing reusable morphology from beneficial cross-property coupling. These results establish a practical separation: images supply structural alternatives, and heterogeneous physical observations determine their experiment-specific posterior probabilities.

Build pipelines are integral to modern software development, yet their energy footprint remains largely invisible to practitioners. Existing CI energy tools either rely on model-based estimation (due to hardware access restrictions in cloud runners) or report only total pipeline energy without decomposing it into meaningful phases. Joule-Profiler is an open-source command-line tool for Linux that measures hardware energy consumption via Intel RAPL (CPU), NVML (NVIDIA GPU), and attributes it to user-defined program phases by monitoring standard output. In this tool paper, we demonstrate Joule-Profiler in the context of Maven build pipelines using Google Gson as a case study. By applying a token pattern-matching Maven plugin invocations, we analyze builds across the 5 most recent Gson releases into per-phase energy profiles, comparing cold builds (empty local repository) and warm builds (cached dependencies).

This study derives charged black string solutions in the presence of a Letelier–Alencar cloud of strings and quintessence within Einstein–Maxwell–$\Lambda$ theory. We study the curvature scalars in order to investigate the asymptotical behavior of these solutions. We calculate the relevant conserved and thermodynamic quantities, confirming their compliance with the first law of thermodynamics. System stability is rigorously analyzed through heat capacity and Gibbs free energy, identifying regimes of local and global equilibrium. Additionally, by interpreting the cosmological constant as thermodynamic pressure within an extended phase space, we establish the validity of the extended first law and derive the corresponding Smarr relation for these solutions. We also investigate the null geodesic structure, deriving the equation for the photon cylinder radius, which generalizes the known result for charged black strings by incorporating the effects of the Letelier–Alencar cloud of strings and quintessence. The numerical analysis of this equation determines the existence and location of photon cylinders as functions of the model parameters.

We examine the relation between star formation and dense gas, traced by HCN(1-0), using a sample of 343 Galactic clumps spanning the full range of evolutionary stages representative of high-mass and cluster formation, including quiescent, protostellar, young stellar object (YSO), and HII regions. Earlier Galactic studies, based primarily on luminous HII regions, reported a linear correlation between infrared (IR) and HCN luminosities, consistent with extragalactic surveys, and interpreted it as a universal star formation law. Our larger evolutionary sample yields a significantly steeper slope of $1.80\pm0.08$, inconsistent with extragalactic trends. However, we find that each of the star-forming evolutionary stages individually has a shallower slope ($\sim$1.2) than that of the full sample, with quiescent clumps showing no correlation. We estimate the Milky Way's total IR and HCN luminosities from dense clumps by integrating the bolometric luminosities and clump masses of Hi-GAL sources. Both fall significantly below extragalactic values, demonstrating that clumps alone cannot account for the IR-HCN relation. Only $\sim$5-25 per cent of Galactic HCN luminosity arises from dense clumps, implying that most extragalactic HCN traces gas that is not involved in star formation. Additionally, dense clumps contribute only $\sim$2-3 per cent of the Milky Way's IR luminosity, which is dominated instead by extended HII regions and PDRs tracing longer star formation timescales. As a result, the observed linear relation in nearby galaxies reflects a nearly constant star formation efficiency averaged over large areas and long timescales, rather than a direct link between SFR and dense gas mass.

Contact-rich manipulation requires robots to balance accurate motion tracking with compliant interaction, yet most visual-action policies leave compliance fixed at the controller level. We present Imp-ACT, a methodologically grounded and practical approach to incorporating direction-dependent Cartesian stiffness modulation directly into demonstration collection, without manual stiffness selection or offline target reconstruction. During teleoperation, a self-tuning impedance controller adapts stiffness along the instantaneous direction of motion while maintaining compliance in orthogonal directions. The adapted stiffness is applied and recorded alongside visual observations and motion commands, capturing motion and compliance under the same dynamics. We implement this pipeline using Action Chunking with Transformer (ACT) to predict end-effector pose, gripper action, and motion-direction stiffness from visual, proprioceptive, and wrench observations. The performance of Imp-ACT is evaluated on wiping and plug insertion using both success rate and quantitative measures of contact behavior. Compared with fixed low- and high-stiffness baselines, Imp-ACT achieves comparable or higher success while maintaining low interaction forces. In wiping, it reduces contact-force vibration by approximately $29\times$ relative to the compliant baseline and $180\times$ relative to the stiff baseline. In plug insertion, it reduces forces orthogonal to the insertion direction by $43\%$ relative to the better fixed-stiffness baseline. These results highlight the benefit of maintaining sufficient stiffness along the direction needed for task execution while preserving compliance in other directions to limit contact forces and accommodate environmental constraints.

Full-duplex speech language models continuously accumulate acoustic key-value (KV) states, making long-running interactions memory-intensive. During listening, the model can finish processing an audio unit before the next arrives; we term the remaining interval listening-time slack. We propose acoustic-to-text KV compression, which introduces a transcription side channel to convert incoming speech into compact textual memory within this interval. When the cache exceeds a target budget during inference, older acoustic states are evicted while transcripts and recent acoustic context remain. We train the side channel with LoRA using cross-entropy on transcription segments. To preserve listening and speaking behavior, we apply knowledge distillation to the original model's token-level output distributions at native prediction positions. On ten-minute LongSpeech sessions, our MiniCPM-o 4.5 implementation reduces peak streaming KV-cache size by 64.6% compared with the same model without eviction. The proposed method also improves transcription, temporal question answering, and summarization over the baseline. Full-Duplex-Bench evaluations further show comparable pause-handling, turn-taking, and interruption performance.

We study interactive multiprover games and many-round protocols in which the communication between the verifier and the provers is quantum. Parallel repetition is known to suppress soundness error of classical two-prover games arbitrarily close to zero, as shown by Raz (STOC '95), even with quantum strategies as shown by Yuen (ICALP '16), and Bavarian, Vidick and Yuen (STOC '17). Yet, we show that quantum communication can have the opposite, unexpected effect. Specifically, we exhibit a one-round quantum game and two-round $\mathsf{QMIP}$ protocols whose local (unentangled) value is strictly less than one for a single instance, yet equals one under $n$-fold parallel repetition for every $n \geq 2$. The key is that parallel repetition not only imposes additional winning conditions but also supplies additional quantum resources as entanglement in the exchanged states can be exploited jointly across copies. Our multiround examples build on superactivation of zero-error capacities of quantum channels. To establish this connection, we introduce quantum games and interactive protocols which capture the one-shot zero-error classical and quantum capacities, both with and without entanglement assistance. Our constructions use quantum-state verification and a teleportation-based reduction from two rounds to one. In contrast, when shared entanglement is allowed, we show that parallel repetition cannot increase the entangled value when at most $3$ messages are exchanged, consistent with classical results of Bellare, Impagliazzo and Naor (FOCS '97), and that of single-prover systems by Kitaev and Watrous (STOC '00). Combining this monotonicity with our quantum game-channel correspondence, we show that entanglement-assisted zero-error classical and quantum capacities cannot be superactivated.

Pocket-conditioned molecular diffusion updates ambient atom coordinates, but many lead-optimization objectives are expressed on quotient features such as distances, contacts, and anchored substructures. We introduce budgeted quotient-residual guidance (QRG), an inference-time correction that makes these quotient objectives active without retraining the molecular generator. QRG lifts quotient covectors to metric-horizontal ambient directions and delivers them through a trust budget set by the frozen sampler's own step norm: quotient geometry chooses the direction, while sampler motion bounds the scale. We derive the horizontal lift, closed-form sampler-budget update, KL/kinetic interpretation around a frozen reverse step, equivariance conditions, and a product-budget split for budget-capped section and residual controls. Controlled quotient tasks confirm that sampler-relative delivery activates signals that raw local quotient gradients leave dormant. On frozen TargetDiff backbones, official seed-0 CBGBench ligand-generation/editing sweeps show practical quality-runtime gains: Local-QRG improves validity from 0.815 to 0.864 on fragment growing, 0.664 to 0.707 on scaffold hopping, and 0.681 to 0.712 on linker design, while PredNext-QRG improves fragment/scaffold and remains near-neutral on linker. Novelty remains 1.000 and diversity is preserved in the matched multi-seed molecular slice, giving task-dependent improvements without sampler retraining or backbone modification. Overall, QRG provides a lightweight route to quotient-aware inference for frozen molecular samplers with explicit runtime accounting.

Different measurements of the same environment can preserve the average quantum channel yet change the feedback strength needed to prepare a state ensemble. We establish this distinction for a multiqubit register under complete Pauli monitoring, targeting the uniform ensemble of real-amplitude pure states. With unit detection efficiency, public records, and bounded Hamiltonian feedback, fixed transverse noise requires a minimum peak strength proportional to $\epsilon^{-1}$ at distributional error $\epsilon$. Tangent noise instead admits a construction requiring only $O[\sqrt{\log(1/\epsilon)}]$. At fixed preparation time, we determine the optimum over all allowed feedback policies and derive a geometric lower bound beyond radial symmetry. The separation identifies the measurement realization as part of the control-resource specification for monitored ensemble preparation.