Papers

Quantum batteries store energy in nonequilibrium quantum states. The ergotropy sets the work extractable by an arbitrary global unitary, but a device usually exchanges energy through a single local port. It remains unclear how much stored work such a port can deliver, and whether topological edge modes help. We study a dimerized spin chain of Su-Schrieffer-Heeger (SSH) type. Excited two-level ancillas charge it at one end by repeated collisions, and an initially empty two-level load discharges it at the same site. An exact free-fermion mapping yields a bound on the load work for all number-conserving quadratic extraction protocols. The load gains extractable work only if a single-particle mode of the battery holds more than half an excitation. Resonant charging fills the topological edge mode above this threshold. A single boundary pulse then delivers 98.6% of the maximal load work allowed by the bound for representative parameters. Under the same six-collision controls, the trivial-phase load remains passive. Within the protocol that selects the load pulse for maximum delivered work, cycle power peaks at a finite collision number while delivered work continues to increase over the scanned range. Over the tested range, the advantage survives hopping disorder, which preserves chiral symmetry. On-site disorder breaks this symmetry and leaves a finite fraction of loads passive. Topology thus supplies a localized access channel, while spectral matching and symmetry decide whether stored energy becomes usable local work.

Membership is the root privacy primitive in machine learning: to date, no hardware-based out-of-distribution detection on black-box models has been demonstrated against constant-time, static neural networks with masked confidence. This paper performs the first cycle-level examination of how large language models and vision transformers interact with various modern microarchitecture components, including integrated accelerators, as LLMs scale in size and answers the question of whether the data that a model was trained on affects its execution footprint even without any input-dependent branch, dynamic optimization, or early exit and in constant-time models. The results confirm that the answer is yes and identify which modern hardware components, such as TLBs or on-core accelerators, reveal or amplify that effect. The results also answer whether the signal is informative enough to reliably classify the in-/vs/out-of-distribution property of membership. To understand why, we perform a systematic root cause analysis and find that the transformer's tokenization steps, which happen during training, alter the locality of the accesses the model makes to fetch the vocabulary token later during inference and, as a result, change the page table access patterns and TLB in a previously unknown data-dependent way, causing microarchitectural state to vary significantly based on whether or not the input was in the distribution of the transformer training data. Building on the above observation, we introduce TranScope: the first microarchitecture tool for detecting membership information with low cost, no need for a surrogate model, and significantly higher robustness, e.g., 0.6 AUC for PETAL (best previously reported) vs 0.9 AUC (ours). This reintroduces hardware as both an opportunity, e.g., a tool for checking copyright violation for the first time, and a new channel for inferring membership (MIA).

Bidirectional video diffusion models denoise entire videos in parallel, yet when trained on effectively unlimited in-distribution data from procedural generators, continue to violate physical laws and simple symbolic rules. We introduce Serial-to-Parallel Diffusion (S2PD), which performs autoregressive diffusion at high noise before switching to parallel diffusion at low noise. The autoregressive phase provides the serial computation needed to coordinate interdependent events and produce valid state transitions while the parallel phase jointly refines the entire video and reduces sampling time relative to fully serial generation. We implement S2PD with two architectures: a pixel-space diffusion transformer trained from scratch and a pretrained video model adapted through LoRA fine-tuning with causal attention. Across games, physical simulations, and real video, S2PD follows rules more reliably than matched bidirectional baselines and generates videos with greater temporal stability and sampling efficiency than other serial methods.

Worst-group accuracy (WGA) evaluates a trained predictor but does not characterize how its frozen backbone behaves when a new head is learned. We introduce BiasFlow, a hook-based toolkit for monitoring class-attribute centroid alignment (IBMI), within-class centroid separation (W-IBMI), and feature-projection sensitivity. IBMI is confounded by class-attribute correlation and is not a measure of causal feature reliance. We pair these diagnostics with BiasFlow Regularization (BFR), a supervised, composable class-conditional centroid-alignment penalty. W-IBMI verifies the quantity BFR optimizes; it is scale dependent and does not independently establish attribute removal. Across the reported small-scale benchmarks, adding BFR improves or preserves mean WGA, with gains up to +26.0 pp on UrbanCars. The principal independent stress test freezes CelebA-Std backbones and trains fresh heads on biased data: BFR+GroupDRO improves WGA from 40.7% to 64.1%, while Male probe accuracy decreases from 92.5% to 72.2%. Attribute information remains recoverable, and cross-task results are mixed. A controlled synthetic-watermark ImageNet experiment additionally improves watermark-shift accuracy by +23.0 pp under matched training. These results support evaluating centroid geometry and resistance to biased head retraining alongside WGA, within the tested protocols.

Braiding and fusing anyons are fundamental to topological quantum computation, yet existing circuit protocols often employ ancillas and measurements. We determine the minimal circuit resources for performing such anyon operations in quantum double models of generic finite groups. Our approach braids anyons through a sequential unitary circuit without ancillas. Intuitively, this is obtained by deforming the underlying graph on which the anyons live. More generally, we introduce graph rewiring, a diagrammatic calculus that assigns a geometric interpretation to individual two-qudit gates and tracks their action on stabilizers. For fusion, we prove a general no-go theorem excluding purely unitary protocols for fusing unknown anyon types without introducing ancillas. Remarkably, we show such "input-agnostic" fusion is enabled by simply using a qudit reset. Hence, contrary to folklore, ancilla resources are not needed to braid or fuse non-Abelian anyons (non-Abelions). This work has implications in experimental efforts in error correction: our fusion protocol highlights reset as an underutilized primitive for error correction without the spacetime-overhead of ancillas and measurements.

Multimodal Diffusion Transformers (MM-DiTs) jointly process visual and textual representations throughout generation. These models repeatedly update the text tokens through multimodal attention, forming dynamic contextual tokens whose function is not well understood. In this work, we introduce a framework for reading this contextual space through natural-language interrogation. We train a lightweight bottleneck network that maps intermediate contextual tokens into the input space of a frozen Large Language Model (LLM), allowing the LLM to answer questions about the emerging image directly from these hidden representations. Our reader reveals that contextual tokens encode a rich, global representation of the emerging scene: generation-specific semantics, including attributes left underspecified by the prompt, are accessible surprisingly early in denoising, while increasingly fine-grained details become readable over time. Remarkably, this information remains decodable even when the MM-DiT receives an empty prompt, showing that contextual tokens accumulate substantial image-specific information from the evolving visual representation itself. We further find that generations with more readable contextual representations tend to receive higher human-preference scores. Building on these observations, we introduce Contextual Alignment, a training technique that explicitly reinforces the visual-semantic information encoded in the contextual tokens, improving generation quality and distributional coverage. Together, our results establish contextual tokens as both an interpretable view into the internal dynamics of MM-DiTs and an effective target for improving generative models.

LLM agents that orchestrate frozen vision-language-action (VLA) policies improve across episodes through text memory, which records what the agent did but not how the task is done. A demonstration video shows it, but fits poorly into an agent's context. The full video slows every turn, fixed keyframes lose the contact detail that decides whether a grasp holds, and what the agent needs shifts from the task's structure while planning to the frames around each contact. We introduce Recursive Video In-Context Learning (RV-ICL), a training-free method that turns a demonstration into a hierarchy the agent navigates rather than a prompt it receives. The hierarchy is built from the sub-events of the demonstration, such as grasps and releases. Its levels grow finer, from keyframes of the whole task to phases, moments and short clips, and are exposed through read-only tools. The agent reads the coarse levels before planning. During execution it re-enters the hierarchy whenever a step needs more detail and loads only the clip of its current sub-goal. One demonstration per task is enough. Built on RPent, RV-ICL raises success from 92.6% to 96.5% on LIBERO-PRO and from 86.7% to 95.8% on LIBERO-Plus.

The measures of distance frequently used in quantum information, such as trace distance or fidelity-based purified distance, are the smallest extensions of the corresponding distance measures in classical information theory. However, this way of extending classical distances is not unique, and a whole spectrum of other extensions can be defined. Here we study the question of how quantum information processing tasks are affected by a change of the error metrics to the maximal extensions of the classical distance measures, imposing more demanding error requirements on operational tasks. We find that the performance of some tasks, such as channel coding or quantum resource distillation, is not affected whatsoever by this change. However, for other tasks such as channel simulation and quantum resource dilution, we show that the rates are now given not by functions of the standard (Umegaki) quantum relative entropy, but ones based on the Belavkin-Staszewski relative entropy. Our results connect the Belavkin-Staszewski relative entropy with operational tasks and reveal that known reversibility results in quantum information, including the reverse Shannon theorem and the asymptotic reversibility of quantum resource transformations, no longer hold under maximal distance measures.

We test Refracted Gravity (RG), a phenomenological modified gravity model that modifies the Poisson equation by introducing an effective gravitational permittivity $\epsilon(\rho_{\rm bar})$ which depends on the baryonic matter density $\rho_{\rm bar}$ and mimics the behaviour of dark matter in cosmic structures. We perform a kinematical analysis of member galaxies for a sample of 74 galaxy clusters, compiled from the Cluster Infall Regions in the Sloan Digital Sky Survey (CIRS). We stack the clusters in three bins of increasing velocity dispersion to mitigate the effects of systematic errors induced by deviations from dynamical equilibrium and from the assumption of spherical symmetry. By means of the MG-MAMPOSSt code, we reconstruct the gravitational potential in RG and infer the model parameters jointly with the gas and stellar mass-profile parameters, adopting broad, physically informed priors on the baryonic components. The estimates of the main parameter in Refracted Gravity, the vacuum permittivity $\epsilon_0$, range from $0.04$ to $0.37$, across the three main stacks, depending on the model adopted for the hot gas component. The constraints are mutually compatible within the quoted uncertainties and overlap with previous cluster-scale estimates. The agreement becomes less sensitive to gas parametrisation when selecting the less-disturbed clusters for building the stacks. These results characterize the sensitivity of cluster-scale RG constraints to baryonic modelling and sample selection

We investigate converse bounds on the average decoding error probability in finite-blocklength classical-quantum channel coding. We first present a lower bound for multiple quantum hypothesis testing in terms of pairwise trace distances and derive a corresponding fidelity bound. We then obtain a spectrum-based converse that depends only on the a priori probabilities and spectra of the states. We show that the converse bound remains tight for the quantum depolarizing channel under suitable conditions. For codes with product-state outputs, this converse takes an explicit form involving products of output-state eigenvalues. We apply it to binary codes over the quantum amplitude damping channel using the input states $\vert+\rangle$ and $\vert-\rangle$. For this setting, we also discuss a normal approximation to the spectrum-based converse in the large blocklength regime. In all numerical examples considered, the spectrum-based converse is tighter than the other converse bounds at low noise levels.

Programmable metasurfaces—arrays of elements whose electromagnetic response can be tuned electronically—can process signals as a wave propagates through or reflects off them, a paradigm known as wave-domain analog computing. The design strategy is typically indirect: the surface is fitted to a prescribed linear operator, such as a discrete Fourier transform, and the result is read from output power measurements. Using direction-of-arrival (DoA) estimation as a canonical benchmark, we study what this indirection costs across analog computing topologies, namely reflective reconfigurable intelligent surfaces (RIS), beyond-diagonal RIS (BD-RIS), and stacked intelligent metasurfaces (SIM). We first derive Cramér–Rao bounds for the complete physical estimation chain, including one that accounts for output detectors measuring only intensity. We then formulate a goal-oriented alternative where the surfaces are trained directly through the deployed readout, with no operator target. We observe that goal-oriented training lets sparsely connected or shallower surfaces reach the task performance of their more complex operator-fitted counterparts—despite nearly identical Fisher information—and that the advantage persists under coarsely quantized (1–3 bit) control. From 16 to 64 elements, the depth saving persists, while the connectivity a beyond-diagonal RIS needs grows with the aperture. Under tuner and per-layer loss, the goal-oriented designs keep their advantage.

In a recent breakthrough, Fauzan proved the irrationality of $\zeta(5)$. In this paper, we combine Fauzan's determinant method with classical simultaneous approximations of zeta values studied by Zudilin and Dauguet–Zudilin to prove the linear independence of $1$, $\zeta(3)$, and $\zeta(4)$ over $\mathbb{Q}$. As a corollary, $\zeta(3)/\pi^4$ is irrational.

Accurate T-staging is crucial for guiding personalized treatment strategies for laryngopharyngeal cancer. However, current clinical practice relies on invasive biopsy procedures, whereas CT-based staging remains challenging due to the complex patterns of tumor invasion. Recent computer-aided approaches face two key challenges: 1) Structural relationship modeling: existing methods underrepresent anatomically structured patterns of tumor invasion, as they either process whole CT volumes without tumor-specific anatomical constraints or rely on labor-intensive tumor segmentation. 2) Fine-grained cross-modal alignment: while radiology reports contain organ-specific invasion details, current methods that apply global feature fusion struggle to accurately align individual anatomical structures with their corresponding textual descriptions. To address these issues, we propose an anatomy-aware multimodal framework that integrates organ-level CT context and radiology reports into a unified representation for laryngopharyngeal T-staging. The framework first constructs an Anatomy-Structured Organ Graph (AOG) that captures invasion patterns between primary sites and surrounding organs, then performs Organ-Anchored Cross-Modal Alignment (OCA) so that each organ node aggregates textual evidence from the radiology report, and finally refines this graph representation by injecting organ-specific invasion cues extracted from the report via Report-Enhanced Graph-Refinement (REG), yielding a multimodal organ graph that combines spatial and textual evidence. Extensive experiments demonstrate that the proposed framework achieves superior performance in T-staging of laryngopharyngeal cancer.

An important question in quantum information theory asks whether every gate in the Clifford hierarchy is semi-Clifford or not, as such gates admit resource-efficient implementations via gate teleportation. Extending recent results for prime dimensions~\cite{silva_clifford_2025}, we prove that every hierarchy gate on a single qudit of dimension \(d\) is semi-Clifford, if and only if \(d\) is square-free. In composite dimensions, \(\mathbb{Z}_d^2\) is a symplectic module rather than a vector space, motivating a distinction between four types of gates. Semi-Clifford gates ($\mathcal{SC}$) are the ones that become diagonal under left and right multiplication by Clifford gates, while a broader class $\mathcal{N}$ admits a permutation diagonal form, under left and right multiplication by Cliffords. Lagrangian semi-Clifford gates ($\mathcal{LSC}$) conjugate some maximal abelian Pauli subgroup to another, whereas generalized semi-Clifford gates ($\mathcal{GSC}$) map some maximal abelian Pauli subalgebra to another under conjugation. In square-free dimensions, we have \(\mathcal{SC}=\mathcal{LSC}\) and \(\mathcal{N}=\mathcal{GSC}\), but when \(d\) is not square-free, these equivalences can fail, as Lagrangian submodules can become non-free. We also show that every third-level gate of the one-qudit Clifford hierarchy is a generalized semi-Clifford in arbitrary dimensions.

Quantum query complexity enjoys a variety of pleasing joint computation properties: for example, a composition theorem asserting $Q(f\circ g)=\Theta(Q(f)Q(g))$ for all Boolean functions $f$ and $g$; a direct sum theorem asserting that computing $k$ copies of a function (or search problem) costs $\Omega(k)$ times as much as the cost of computing one copy; and a direct product theorem asserting that for Boolean functions, even succeeding at the direct sum problem with exponentially small probability still requires $\Omega(k)$ times the cost of computing one copy to bounded error. However, all of these results are strictly for worst-case quantum query complexity. For example, if we have a fixed distribution $\mu$ over inputs, the direct sum theorem says nothing about the quantum query complexity of computing $k$ copies of $f$ when the input comes from the product distribution $\mu^k$ instead of being worst-case. (Note that while a standard Yao-type minimax theorem guarantees a hard distribution for the direct sum problem, there's no guarantee that this hard distribution is a product distribution.) A similar problem occurs for the direct product theorem and the composition theorem: none of these results respect distributions. In this work, we give distributional joint computation lower bounds for the composition, direct sum, and direct product problems. Along the way, we introduce some new tools for handling quantum query lower bounds, including (a) a new “multiplicative” variant of the $\gamma_2$ norm (which we use in place of the multiplicative adversary method for proving the direct product theorem), and (b) a new “Shaltiel-free” measure of quantum query complexity, which we show characterizes the composition behavior of distributional quantum query complexity and satisfies pleasing properties.

Some diffusion posterior samplers construct Gaussian-tilted intermediate distributions along the reverse process. We observe that these targets can be pulled back to clean-space posteriors with weaker conditioning, with samples transported analytically to the corresponding noisy-space target through a Gaussian bridge. For the sequential Monte Carlo (SMC) sampler MCGDiff, the effective observation variance of this pulled-back problem is up to twice the diffusion-noise variance. We exploit this structure to initialize MCGDiff directly at an intermediate time: an approximate solver samples the softened clean-space posterior, the Gaussian bridge maps these samples to the tilted target, and only the remaining SMC suffix is run. This trades asymptotic consistency for finite-particle performance. With moment-matching posterior sampling (MMPS) as the solver, the hybrid improves sliced Wasserstein distance by roughly $2\times$ at matched particle count on a structured Gaussian-mixture inverse problem, and by more than an order of magnitude when the posterior-relevant mode is rare under the prior. A prior-initialization control, which retains the bridge but drops the clean-space conditioning, shows that on MCGDiff's standard Gaussian-mixture benchmark most of the improvement is insensitive to the conditioning. Conditioning the initialization gives a further consistent gain on the structured problem, and becomes decisive on a rare-mode problem, where resampling cannot repopulate a mode absent from the initial population.

Every recurrence of a looped language model adds cost in training, decoding, prefill, and reinforcement learning (RL). The closer recurrent states get to fixed points, the less the path to them matters. This enables truncated backpropagation in training; terminal key-value (KV) sharing for decoding with almost no loss in accuracy; a distilled student that prefills up to 1.79x faster; and RL updates that compute gradients from saved rollout states, 2x faster than backpropagating through the replayed trajectory. We therefore improve the two components of training that shape these fixed points: the depth prior and input injection. Fixed-depth training breaks KV sharing, and Huginn's broad depth prior supports sharing but dilutes supervision at the target depth more than sharing requires; we learn the prior from prediction feedback, with an entropy term that keeps it broad. Existing injection schemes let the state's component along the input amplify or cancel the injection; we remove this component with orthogonal injection. From 100M to 1.6B parameters, the learned prior and orthogonal injection lower perplexity at every scale relative to Huginn's prior and existing injection schemes, respectively. At 1.6B, the learned prior with a 3x smaller KV cache matches the downstream average of fixed-depth training with the full cache.

We determine the two-way quantum, entanglement-distribution, private, and secret-key capacities of the pure-loss bosonic channel under an unconditional mean transmitted-photon-number constraint. For transmissivity $\eta$ and mean photon number $N$, all four capacities coincide and are given by $g(N)-g((1-\eta)N)$, where $g$ is the entropy of a thermal mode. This resolves the longstanding energy-constrained capacity problem by establishing the optimality of the reverse-coherent-information rate introduced in 2009. Our central tool is sector teleportation simulation, which we develop specifically to transfer the input-energy constraint to the shared entanglement resource, thereby preserving the constraint throughout converse arguments for general adaptive protocols. For every $N>0$, we further show that the corresponding strong-converse thresholds coincide with the unconstrained value $-\log_2(1-\eta)$. For parallel protocols with a hard total-photon-number cutoff, the same energy-constrained capacity formula instead satisfies an exponential strong converse. As an extension, we determine the energy-constrained two-way capacities of the quantum-limited amplifier, showing that classical assistance does not improve its unassisted quantum and private rates. Together, these results establish a unified finite-energy framework for adaptive bosonic communication and reveal how energy constraints fundamentally reshape both attainable rates and converse bounds.

Sliders provide an intuitive interface for continuous image editing. In current generative approaches, however, the slider is simply a rescaling of the method's strength parameter, such as an adapter coefficient, a prompt weight, or an interpolation factor. This strength relates poorly to perceptual change. The image can partially revert as the slider moves, long stretches of the range produce no visible difference, and short intervals transform the image abruptly. Remapping the strength could fix this uneven pace, but only if the trajectory is monotone, which current methods do not enforce. We therefore distinguish the slider from the strength, and require perceptual distance from the input to grow linearly with the slider value. We introduce UniSlider, a lightweight LoRA trained on a few-step editing backbone so that its strength approximates this ideal slider. Few-step sampling lets us impose this objective in pixel space without intermediate ground truth, and the backbone's output is preserved at full strength. However, a low-rank adapter cannot make the strength fully uniform. Our slider is thus an inference-time remapping of the strength, obtained by adaptive sampling. Since training optmizes to make the trajectory monotone, this remapping closes the remaining gap without extra training or parameters. On a new benchmark of 300 continuous edits evaluating uniformity, monotonicity, edit fidelity, and identity preservation, UniSlider outperforms all prior methods and is preferred in a user study.

Memory has become integral to the LLM agent ecosystem, supporting information retention and reuse across interactions. However, most existing agent memory systems construct memory in a query-agnostic manner, which can incur unnecessary preprocessing cost and discard details that later prove essential. Recent studies have begun shifting memory processing toward runtime adaptation, but typically specialize in particular operations or fixed processing schemes, leaving flexible control over performance, cost, and latency largely underexplored. To address this challenge, we present MemPilot, a flexible framework that orchestrates on-demand memory curation under different performance–cost–latency preferences. Specifically, we optimize a multi-step LLM policy via reinforcement learning to iteratively choose between retrieving from query-agnostic memory and delegating query-specific curation of raw multimodal history to heterogeneous LLMs and VLMs. The policy jointly controls evidence amount, curation instructions, model selection, and visual access, enabling fine-grained allocation of runtime computation. To optimize this policy under competing objectives, we adapt objective-wise advantage decoupling by separately estimating each objective's advantage before aggregation. Moreover, we introduce prefix-based marginal utility estimation for fine-grained credit assignment across multi-step rollouts. Experiments on five multimodal agent-memory benchmarks demonstrate favorable performance–cost–latency trade-offs across optimization preferences, with preference sweeps yielding broader frontiers than existing trade-off-aware baselines.

Open-source web agents are now strong enough to execute realistic browser tasks, but training them with reinforcement learning still depends on weak supervision: binary task success is too sparse for credit assignment, while frontier-language-model judges are too expensive to call at every step and cannot be assumed available at deployment. We introduce CLIFT, a training and test-time scaling method built around conformal self-verification. During training, the agent answers natural-language verification questions about its own rollouts; a Compositional Conformal Certifier keeps only question signals whose URL-conditional evidence agrees with a training-time judge, assigns signed trust weights through polarity-aware lift, and blends the resulting verifier score into per-step rewards in a way that never subtracts from the judge baseline. At test time, the same certified bank is frozen and reused as structured evidence for Conformal Trajectory Selection (CTS): the agent samples a greedy rollout and one or more diverse retries, the self-verifier summarises each URL trace, and a conservative majority-vote rule chooses whether to swap away from the current incumbent without calling any external judge. This single mechanism supports three settings. On WebArena Infinity, CLIFT achieves state-of-the-art performance among open-source web agents. On VisualWebArena, a bank trained with the open model transfers to GPT-5.5 at test time and reaches state-of-the-art performance under the canonical harness. On Online Mind2Web, without training an agent on the benchmark, translating the certified question bank improves a live-web agent in zero-shot evaluation. Together these results position conformal self-verification as a way to turn costly judge feedback into a reusable training signal and a judge-free test-time scaling signal.

Magic state cultivation is a promising method for preparing high-fidelity non-Clifford resource states - a central component of fault-tolerant quantum computing - with low spacetime overhead. In benchmarking cultivation, existing resource estimates assume access to noiseless transversal measurement of the output magic state to allow for postselection during the final escape stage of the cultivation process. However, such measurements would destroy the magic state and are therefore unavailable during computation. We develop an open-boundary postselection method, Caliper, that uses only nondestructive mid-circuit syndrome information and substantially outperforms existing methods operating under these constraints. In some regimes, our postselection method uniquely recovers the resource efficiency of prior estimates. However, in other cases, we observe a tradeoff. At spacetime costs reported in previous works, the achievable logical error rates can be orders of magnitude higher, and recovering the reported logical error rate requires substantially larger codes and longer escape stages. Our results highlight how practical considerations involving mid-circuit postselection can impact the design and performance of magic state cultivation protocols.

Compared to self-employment or performance-based pay, fixed wages reduce earnings uncertainty but are prone to market distortions through moral hazard and adverse selection. Using a model of wage contracts under asymmetric information, I show that these distortions map to marginal treatment response functions in a marginal treatment effects (MTE) framework. I apply this framework to a field experiment in which data-entry workers choose between a randomized fixed wage and a piece rate. I find evidence of both moral hazard and adverse selection: Fixed wages reduce worker productivity by an estimated 8.74 percent relative to the mean. Meanwhile, a 10 percent increase in the wage offer attracts a marginal worker whose productivity is higher by 1.90 percent of the mean. Using semi-parametric MTE estimation, I calculate the welfare loss associated with asymmetric information and the marginal values of public funds (MVPFs) for policies aimed at recovering this loss. I find that a tax on piece rates can mitigate adverse selection into fixed wages, generating tax revenue at social costs as low as $0.88 per dollar.

We identify a barrier that helps explain why proving stronger quantum state-preparation lower bounds has been so difficult. In particular, we establish a quantum analogue of the Razborov-Rudich natural proofs barrier for state-preparation lower bounds. We call a property of quantum states natural if it holds for a sufficiently large fraction of Haar-random states and can be efficiently tested when given all of the state's amplitudes. Under a standard cryptographic assumption, we show that no natural property can prove superpolynomial state-preparation lower bounds even against a fixed level of the Magic Hierarchy. We show that several existing state-preparation lower-bound techniques are natural in our sense, including arguments based on approximate degree, not being a unique ground state of a local Hamiltonian, and mutual information.

Scientific figures often encode quantitative results that are not readily available in machine-readable form, making accurate plot digitization important for verifying and reusing published findings. Yet it remains unclear how accurately current models recover plotted values from real scientific figures, as existing benchmarks rely largely on synthetic charts or cover only a limited range of chart types. We introduce PlotGround, an automated pipeline for building plot digitization benchmarks from real scientific figures and their author-released source data. PlotGround maps figures to source tables, identifies reconstructable panels, and generates quantitative questions with source-grounded reference values. We use PlotGround to construct PlotGround-1k, a human-verified benchmark of 1,119 questions from 1,066 bioRxiv preprints. Across sixteen multimodal models, the best reaches 87.5% accuracy at a $\pm 5\%$ relative-error tolerance. Tightening the tolerance to $\pm 2\%$ lowers every model's accuracy by 11-24 percentage points, revealing a gap between approximate visual reading and precise quantitative recovery. PlotGround's paired figure-source structure lets us compare how accurately the same values are recovered from figures and from source tables. Providing source tables instead of figures raises a coding agent's accuracy from 90.0% to 97.4% while cutting cost by 72%.

We introduce TasteVal, a benchmark to evaluate the experimental research taste of frontier models. We define research taste as the ability to pick interesting problems to solve, design experiments, and interpret experimental results. TasteVal measures the experimental component of research taste; given a fixed research problem, we measure how well a model iteratively designs experiments and draws conclusions from their outcomes. We operationalize experimental research taste as compute efficiency; a Researcher who reaches the same score as an expert human using half the serial experimental compute has twice the experimental taste. Experimental taste thus acts as a multiplier on experimental compute, making it a key input to forecasts of AI progress. TasteVal consists of 8 novel, challenging, open-ended tasks representative of frontier AI R&D. To isolate taste from coding ability, the model under evaluation acts as a Researcher that iteratively designs experiments while a fixed Coder agent implements them and reports their results. The Researcher executes until either the 40 H100 hour or 120 wall-clock hour budgets are exhausted. We recruit 24 human experts, at least 2 per task, and take the best expert attempt per task as the expert baseline. We evaluate 20 models released between 2023 and 2026. The best-performing model, Opus 5.5, exceeds our expert baseline, with a compute multiplier of 2.3x (95% CI 1.15-4.37), at roughly 1/30 of our baseliners' average per-run cost. On TasteVal, the compute multiplier of frontier models has doubled approximately every 3.0 months since December 2025 (95% CI 1.7-5.0), up from every 14 months between 2023 and December 2025. Measured by final normalized performance, frontier models show no trend break, doubling every 14.6 months. To keep TasteVal uncontaminated, we do not release the tasks.

Large longitudinal cohorts often contain wrist accelerometry without optical heart-rate sensing, motivating recovery of cardiac information from motion signals already collected during sleep. We present SeqSmoother, a transformer-based temporal corrector for sleep heart rate (HR) estimation from wrist accelerometry. SeqSmoother combines spectral descriptors with an intermediate Nightbeat-derived frequency anchor and a physics-motivated sub-harmonic feature designed to identify harmonic frequency lock-on. All inference-time features are derived from wrist accelerometry, while ECG is used only to construct reference HR labels and training-label quality weights. We evaluate SeqSmoother using 13 participant-disjoint held-out folds and compare it with the official Nightbeat implementation under a matched 60-s window and 15-s step protocol. Across all out-of-fold predictions, SeqSmoother achieved a participant-macro MAE of 1.60 bpm. On Nightbeat-retained matched intervals, Nightbeat achieved lower absolute error than SeqSmoother (0.615 versus 1.091 bpm), while SeqSmoother provided estimates over a larger portion of the eligible recording; Nightbeat produced final estimates for 72.85% of the SeqSmoother-eligible out-of-fold grid. Separately, the proposed sub-harmonic ratio achieved an AUROC of 0.972 for identifying reference-defined harmonic lock-on candidates. These findings reveal an accuracy-availability trade-off between learned temporal modeling and quality-gated signal processing while providing empirical support for a physics-informed approach to identifying frequency-tracking failures in accelerometer-based sleep HR estimation.

We study mistake bounds for differentially private online learning and online prediction under oblivious realisable adversaries. Online learning requires the learner to release a hypothesis at each time step whereas in online prediction, the learner only needs to make predictions without releasing a hypothesis. Using a novel lower bound for private online learning and an upper bound for private prediction, we show that the sample complexity of these two problems are separated by a factor that grows with the time horizon for every class of finite Littlestone dimension $d$. First, we prove that every $(\epsilon,\delta)$-private online learner has a deterministic realisable stream of length $T$ on which the mistake bound is at least $\mathbb E [M_T]= O\left(\frac d\epsilon \log^{2/3} T\right)$. In particular, this is the first non-trivial lower in the range $1/T<\delta<1/\log T)$ left open in earlier works[SR22,DSS24,LWY24]. Second, we prove that for every class of of Littlestone dimension $d$, there exists an $(\epsilon,\delta)$-jointly private predictor with at most $2^{2^{cd^2}}\epsilon^{-2}\log^2(2/(\epsilon\delta))$ expected mistakes, independently of $T$, for some absolute constant $c>0$. Thus, for every fixed class of finite Littlestone dimension when $\delta=\Theta(1/\log T)$, private learning requires $O\left(\log^{2/3} T\right)$ expected mistakes, whereas private prediction admits $O(\log\log T^2)$.

In this paper, we study the weak limit of the invariant measures attached to an analytic family $\{f_t\}_{t\in\mathbb{D}^*}$ of regular polynomial automorphisms over $\mathbb{C}^N$ parametrized by the unit punctured disk which may degenerate at the origin, by using the techniques called hybrid spaces developed by Boucksom–Favre–Jonsson. For each $t$, the invariant measure $\mu_t$ is constructed by Sibony, and we show that the weak limit of $\{\mu_t\}_t$ over the hybrid space is given by the invariant measure of the regular polynomial automorphism $f_{\mathbb{C}((t))}^{\operatorname{an}}$ over $\mathbb{C}((t))$ induced from the original family. This is a generalization of the author's previous result for degenerating families of dynamics of Hénon mappings.

Efficient fault-tolerant quantum computing architectures benefit from low-weight logical operators. However, reported canonical bases for high-rate, high-distance codes can be several times heavier than the code distance, and fundamental obstructions can prevent minimum-weight logical operators from forming a complete canonical basis. To address this issue, we develop design principles for group-valued quantum error-correcting codes with low-weight full regular canonical logical bases, generated by translating a set of seed operators. We introduce frame width $f$, the smallest weight cutoff that permits a complete canonical logical basis, as an additional code parameter alongside distance $d$ and check weight $w$. Cataloguing these parameters together with explicit logical bases can inform resource estimates and the compilation of fault-tolerant logical operations. First, we identify a structural obstruction: for fully populated binary monomial CSS checks over a group of odd order, full regularity forces distance two, while for groups of power-of-two order, full row rank of both check matrices after replacing every group element by $1$ guarantees full regularity. Second, we demonstrate our principles through explicit high-rate lifted-product (LP) and pair-partition (PP) codes: an $[[1088,128,22]]$ LP code with $f=22$ and an $[[1024,256,24]]$ PP code with $f \leq 27$, both at $w=10$. At $w=11$, our PP constructions attain either $d=25$ with $f \leq 27$ or $d=24$ with the proved optimum $f=25$, demonstrating both a tradeoff among these parameters and a strict separation between distance and frame width. Finally, we propose a method combining symmetry reduction and stabilizer-based pruning for exhaustive distance certification, and use it to efficiently establish exact distances for LP examples with more than a thousand qubits.