arXiv:2607.15750v2 Announce Type: replace
Abstract: Fully homomorphic encryption (FHE) schemes such as RNS-CKKS enable privacy-preserving services through direct computation on encrypted data. While recent FHE compilers optimize FHE programs, they operate at the coarse-grained ciphertext level, where each ciphertext operation comprises a sequence of polynomial operations. At this granularity, the compilers miss polynomial-level optimization opportunities across ciphertext operations. This work presents Recifhe, a new multi-level compiler that supports both ciphertext-level and polynomial-level optimization. At the ciphertext level, Recifhe transforms a non-FHE input program into an FHE program by inserting ciphertext management operations and applies global optimizations. At the polynomial level, Recifhe eliminates redundant polynomial computations across ciphertext operations. Recifhe achieves a 1.25x speedup over ciphertext-level-only optimization.
Science Journals
arXiv:2607.15879v2 Announce Type: replace
Abstract: Much empirical legal research depends on translating unstructured text into structured variables. In corporate governance research as elsewhere, this translation has traditionally relied on human coding of documents such as charters and bylaws, a process that is costly, difficult to scale, and often opaque. This paper introduces DECODEM, a set of benchmark datasets for evaluating the automated extraction of corporate governance variables from organizational documents. The benchmarks pair randomly sampled corporate charters and bylaws with high-quality human annotations covering a range of governance provisions commonly studied in empirical work.
Using these datasets, the paper evaluates several large-language-model extraction pipelines that vary in prompt design, task decomposition, and document handling. The underlying task consists of a set of document-level binary classification problems, one for each governance variable. The results show that automated extraction is feasible at a high level of accuracy for many provisions, with median performance near the upper bound across approaches. At the same time, performance varies systematically across variables, with a small number of provisions accounting for most of the remaining errors. More elaborate prompting strategies and cascading pipelines do not consistently improve performance for frontier models, but substantially narrow the gap between frontier and efficiency-oriented models in some settings, suggesting that pipeline design can partly substitute for model capability.
By providing a standardized benchmark and a systematic evaluation of extraction methods, the paper demonstrates that current frontier models can extract legally meaningful information from complex corporate documents with high accuracy and suggests an important future role for automated feature extraction in constructing corporate governance datasets.
arXiv:2607.15893v2 Announce Type: replace
Abstract: While the internal mechanisms of autoregressive (AR) transformers have been studied extensively, much less is known about diffusion language models (DLMs), an emerging alternative that generates text by iterative denoising. In this work, we study how DLMs implement induction, a mechanism behind in-context learning in which the model finds a repeated context and copies the token that followed it. Our analysis compares attention-only AR models and absorbing-mask DLMs with matched architectures. We find that DLMs learn a bidirectional induction circuit, where previous-token and next-token heads write local context into the residual stream and later induction heads use it to find and copy the answer from the matching source position. The circuit is direction-symmetric, working whether the source appears in the past or in the future. When only left context is visible, matching what an AR model sees, the DLM does not outperform its AR counterpart in induction capabilities. However, we observe it has stronger induction when both sides of the masked token are visible, pointing to bidirectional context access rather than a stronger one-sided mechanism. Beyond induction, we provide causal evidence that DLMs compute the global fraction of masked tokens and use it as an implicit timestep, even though they are given no explicit timestep embedding.
arXiv:2507.18780v2 Announce Type: replace
Abstract: We consider data-driven reduced-order models of partial differential equations with shift equivariance. Shift-equivariant systems typically admit traveling solutions, and the main idea of our approach is to represent the solution in a traveling reference frame, in which it can be described by a relatively small number of basis functions. Existing methods for operator inference allow one to approximate a reduced-order model directly from data, without knowledge of the full-order dynamics. Our method adds additional terms to ensure that the reduced-order model not only approximates the spatially frozen profile of the solution, but also estimates the traveling speed as a function of that profile. We validate our approach using the Kuramoto-Sivashinsky equation, a one-dimensional partial differential equation that exhibits traveling solutions and spatiotemporal chaos. Results indicate that our method robustly captures traveling solutions, and exhibits improved numerical stability over the standard operator inference approach.
arXiv:2511.20956v2 Announce Type: replace
Abstract: Breast ultrasound (BUS) reporting relies on clinically meaningful lesion descriptors, including BI-RADS category, lesion shape, margin, echogenicity, posterior features, pathology, and histology. However, many public BUS datasets provide structured annotations and lesion masks without paired radiologist-written reports, limiting the development of vision--language models for BUS report generation. We propose BUSTR, a descriptor-aware vision--language framework that uses structured lesion information to enable report generation under limited report supervision. BUSTR first constructs descriptor-derived reports from available annotations and radiomics features extracted from lesion masks. It then trains a multi-head Swin Transformer encoder with multitask supervision to learn descriptor-aware visual representations across datasets with partially overlapping annotation sets. The projected visual tokens condition a frozen LLaMA-based language model, and training is guided by a dual-level objective combining token-level cross-entropy with representation-level cosine alignment. At inference, BUSTR generates reports from BUS images without access to structured descriptors, lesion masks, or radiomics features. We evaluate BUSTR on the public BrEaST and BUS-BRA datasets using natural language generation and clinical efficacy metrics. BUSTR improves report similarity and descriptor recovery compared with representative report-generation baselines, with notable gains for lesion shape, margin, posterior features, and pathology, as well as improved BI-RADS sensitivity and F1-score on BrEaST. These results suggest that structured BUS descriptors, lesion masks, and radiomics features can provide useful supervision for descriptor-aware BUS report generation when paired radiologist-written reports are unavailable.
arXiv:2601.08662v3 Announce Type: replace
Abstract: This tutorial is designed to make reinforcement learning (RL) more accessible to undergraduate students by offering clear, example-driven explanations. It focuses on bridging the gap between RL theory and practical coding applications, addressing common challenges that students face when transitioning from conceptual understanding to implementation. Through hands-on examples and approachable explanations, the tutorial aims to equip students with the foundational skills needed to confidently apply RL techniques in real-world scenarios.
arXiv:2607.16015v2 Announce Type: replace
Abstract: 6D pose estimation remains a key challenge in robotics and computer vision, particularly in industrial environments. The deployment of currently available data-driven methods is often limited by resource-intensive data pipelines, reliance on textured 3D models, and sensitivity to geometric deviations caused by damages or assembly defects. We present PIXIE, a zero-shot framework that estimates the 6D pose of an object from an RGB image using only an untextured 3D model. Synthetic depth and normal maps are rendered from sampled reference viewpoints and matched to the query image via a pretrained cross-modality feature matcher. Matched keypoints are back-projected to obtain 2D--3D correspondences for PnP-based pose estimation. Relying exclusively on geometry makes the method inherently robust to lighting and texture variation, while correspondence filtering handles geometric deviations between the model and physical object. We evaluate on widely-used public benchmarks, reporting state-of-the-art results on texture-less objects without object-specific training, and introduce a novel dataset with assembly defects, texture variations, and occlusion to demonstrate real-world applicability.
arXiv:2110.13814v4 Announce Type: replace-cross
Abstract: We study actual bidding behavior when a new auction format gets introduced into the marketplace. More specifically, we investigate this question using a novel dataset on internet display advertising auctions that exploits a staggered adoption by different publishers (sellers) of first-price auctions (FPAs), instead of the traditional second-price auctions (SPAs). We analyze the auction format change using difference-in-differences regressions and a synthetic difference-in-differences estimator, which better handles pre-trends. The results show that revenue per sold impression (price) jumps considerably for treated publishers relative to control publishers, with increases ranging from 25% to 70% of the pre-treatment price level of the treated group. Moreover, for later auction format changes, the increase in price levels under FPAs relative to those under SPAs tends to dissipate over time, reminiscent of the revenue equivalence theorem, although the extent of this reversion depends on the specification. We view these results as suggestive of initially insufficient bid shading following the format change, as opposed to an immediate transition to a new Bayesian Nash equilibrium, with prices tending to decline in several specifications in a manner consistent with gradual adjustment in bidding behavior as bidders learn to shade their bids. Our work constitutes one of the first field studies on bidders'responses to auction format changes, providing an important complement to theoretical model predictions. As such, it provides valuable information to auction designers when considering the implementation of different formats.
arXiv:2209.15130v3 Announce Type: replace-cross
Abstract: We study a general matrix optimization problem with a fixed-rank positive semidefinite (PSD) constraint. We perform the Burer-Monteiro factorization and consider a particular Riemannian quotient geometry in a search space that has a total space equipped with the Euclidean metric. When the original objective f satisfies standard restricted strong convexity and smoothness properties, we characterize the global landscape of the factorized objective under the Riemannian quotient geometry. We show the entire search space can be divided into three regions: (R1) the region near the target parameter of interest, where the factorized objective is geodesically strongly convex and smooth; (R2) the region containing neighborhoods of all strict saddle points; (R3) the remaining regions, where the factorized objective has a large gradient. To our best knowledge, this is the first global landscape analysis of the Burer-Monteiro factorized objective under the Riemannian quotient geometry. Our results provide a fully geometric explanation for the superior performance of vanilla gradient descent under the Burer-Monteiro factorization. When f satisfies a weaker restricted strict convexity property, we show there exists a neighborhood near local minimizers such that the factorized objective is geodesically convex. To prove our results, we provide a comprehensive landscape analysis of a matrix factorization problem with a least squares objective, which serves as a critical bridge. Our conclusions are also based on a result of independent interest stating that the geodesic ball centered at Y with a radius 1/3 of the least singular value of Y is a geodesically convex set under the Riemannian quotient geometry, which as a corollary, also implies a quantitative bound of the convexity radius in the Bures-Wasserstein space. The convexity radius obtained is sharp up to constants.
arXiv:2404.14732v3 Announce Type: replace-cross
Abstract: I propose a general quantitative framework to evaluate the quality, track the historical development, and guide future optimization of photovoltaic (PV) absorbers at any development level, both lab-made and computer-simulated. The framework is centered around a PV figure of merit designed to include efficiency limitations that are not captured by classic detailed balance methods derived from the Shockley-Queisser limit. A more stringent set of figure-of-merit-driven efficiency limits are calculated for 28 experimentally synthesized PV absorbers and 10 PV computationally modeled absorbers. Among early-stage absorbers, this analysis reveals very large differences in their likelihood of achieving high PV efficiencies in the future. Since the proposed figure of merit is instantly evaluated from a single equation, it can be a suitable objective function for closed-loop research on PV materials in autonomous labs, while also providing a quantitative bridge between computationally determined material properties and PV efficiency.
arXiv:2601.03043v4 Announce Type: replace
Abstract: Large language models (LLMs) demonstrate strong capabilities across a wide range of complex tasks and are increasingly deployed at scale, placing significant demands on inference efficiency. Prior work typically decomposes inference into prefill and decode stages, with the decode stage dominating total latency. To reduce time and memory complexity in the decode stage, a line of work introduces sparse-attention algorithms. In this paper, we show, both empirically and theoretically, that sparse attention can paradoxically increase end-to-end complexity: information loss often induces significantly longer sequences, a phenomenon we term ``Less is Less'' (Lil). To mitigate the Lil problem, we propose an early-stopping algorithm that detects the threshold where information loss exceeds information gain during sparse decoding. Our early-stopping algorithm reduces token consumption by up to 90% with a marginal accuracy degradation of less than 2% across reasoning-intensive benchmarks.
arXiv:2601.09761v3 Announce Type: replace
Abstract: It is shown that equations describing the Galilean electromagnetism in the presence of sources hold invariant under the l-conformal Galilei group for an arbitrary (half)integer parameter l. The group contains transformations which link an inertial frame of reference to those moving with constant accelerations of order up to 2l-1, thus pointing at potential dynamical instability.
arXiv:2603.07530v2 Announce Type: replace
Abstract: In-context imitation learning enables robots to adapt to new tasks from a small number of demonstrations without additional training. However, existing approaches typically condition only on state-action trajectories and lack explicit representations of task intent. This limitation hinders performance in complex and ambiguous task settings where the same actions may be consistent with different objectives. To address this, we present In-Context Imitation Learning with Visual Reasoning (ICLR), a novel framework that augments demonstration prompts with structured visual reasoning traces representing anticipated future robot trajectories in image space. ICLR also jointly learns to generate reasoning traces and low-level actions within a unified autoregressive transformer, enabling the model to mimic not only action prediction but also the reasoning process that leads to those actions. We extensively evaluate ICLR in both simulation and real-world manipulation tasks and demonstrate consistent improvements in success rates and generalization to unseen tasks and novel object configurations compared to other in-context imitation learning methods. These results suggest that incorporating embodied visual reasoning represents a promising direction for enhancing the robustness and generalization of robotic in-context learning systems.
arXiv:2406.10930v2 Announce Type: replace-cross
Abstract: In our earlier paper, "2 CSPs all are approximable within a constant differential factor" (ISCO 2018, LNCS 10856), we introduced a family of combinatorial designs called 'alphabet reduction pairs of arrays' (ARPAs). These designs are parameterized by three integers $q,p,k$, with $p\leq q$ and $k\leq p$: $q$ is the size of the alphabet from which the arrays draw their entries; $p$ is the maximum number of distinct symbols allowed in a row of the second array; $k$ is the largest integer for which the two arrays coincide -- up to row permutations -- on any $k$-element subset of their columns. The first array must contain at least one occurrence of the word $0\ 1 \cdots\ q-1$ as a row. The idea is to cover as many occurrences of this word as possible using as few words as possible, each containing at most $p$ distinct symbols.
ARPAs are related to the approximability of constraint satisfaction problems with bounded constraint arity ($k$-CSPs). In this context, we are particularly interested in ARPAs that maximize the frequency of the word $0\ 1 \cdots\ q-1$. We call such ARPAs 'optimal' and study them in this paper. To this end, we introduce a simpler family of combinatorial designs called 'Cover pairs of arrays' (CPAs), which can be viewed as partially defined ARPAs with Boolean entries. We prove that ARPAs and CPAs are equivalent with respect to maximizing the frequency of their target word. As a corollary of our proof, computing the frequency of the target word in optimal ARPAs reduces to solving a linear program in $q + p + 1$ continuous variables and $k + 1$ constraints. We also prove the optimality of previously known ARPAs for $p=k$ and provide optimal ARPAs for $k=1$ and $k=2$.
arXiv:2410.16005v4 Announce Type: replace-cross
Abstract: Metastatic prostate cancer is one of the leading causes of cancer-related morbidity and mortality worldwide. It is characterized by a high mortality rate and a poor prognosis. In this work, we explore how a clinical oncologist can apply a Stackelberg game-theoretic framework to prolong metastatic prostate cancer survival, or even make it chronic in duration. We utilize a Bayesian optimization approach to identify the optimal adaptive chemotherapeutic treatment policy for a single drug (Abiraterone) to maximize the time before the patient begins to show symptoms. We show that, with precise adaptive optimization of drug delivery, it is possible to significantly prolong the cancer suppression period, potentially converting metastatic prostate cancer from a terminal disease to a chronic disease for most patients, as supported by clinical and analytical evidence. We suggest that clinicians might explore the possibility of implementing a high-level tight control (HLTC) treatment, in which the trigger signals (i.e. biomarker levels) for drug administration and cessation are both high and close together, typically yield the best outcomes, as demonstrated through both computation and theoretical analysis. This simple insight could serve as a valuable guide for improving current adaptive chemotherapy treatments in other hormone-sensitive cancers.
arXiv:2412.18613v2 Announce Type: replace-cross
Abstract: Illusions are entertaining, but they are also a useful diagnostic tool in cognitive science, philosophy, and neuroscience. A typical illusion shows a gap between how something `really is' and how something `appears to be', and this gap helps us understand the mental processing that led to how something appears to be. Illusions are also useful for investigating artificial systems, and much research has examined whether computational models of perception fall prey to the same illusions as people. Here, I invert the standard use of perceptual illusions to examine basic processing errors in current vision language models. I present these models with illusory-illusions, neighbors of common illusions that should not elicit processing errors. These include such things as perfectly reasonable ducks, crooked lines that truly are crooked, circles that seem to have different sizes because they are, in fact, of different sizes, and so on. I show that many current vision language systems mistakenly see these illusion-illusions as illusions. I suggest that such failures are part of broader failures already discussed in the literature.
arXiv:2501.16801v2 Announce Type: replace-cross
Abstract: We present an effective theory for describing electron dynamics driven by an optical external field in a Schr\"{o}dinger's cat state. We show that the reduced electron density matrix evolves as an average over trajectories $\{\rho_\alpha\}$ weighted by the Sudarshan--Glauber $P$ distribution $P(\alpha)$ in the weak light--matter coupling regime. Each trajectory obeys an equation of motion, $\mathrm{i} \partial_t\rho_\alpha=\mathcal{H}_{\alpha} \rho_\alpha-\rho_\alpha\mathcal{H}_{\alpha}$, where an effective Hamiltonian $\mathcal{H}_{\alpha}$ becomes non-Hermitian due to quantum interference of light. The optical quantum interference is transferred to electrons through the asymmetric action between the ket and bra state vectors in $\rho_{\alpha}$. This non-Hermitian dynamics differs from the conventional one observed in open quantum systems, described by $\mathrm{i} \partial_t\rho=\mathcal{H}\rho-\rho \mathcal{H}^\dagger$, which has complex conjugation in the second term. We confirm that the reduced, trajectory-resolved effective theory agrees with full electron-photon simulations for the few-electron Dicke model, thereby validating the interferential non-Hermitian description in the weak-coupling regime.
arXiv:2501.17936v2 Announce Type: replace-cross
Abstract: Rare events of large-scale spatially-correlated exponential random fields are studied. The influence of spatial correlations on clustering and non-sphericity is investigated. The size of the performed simulations permits to study beyond-$7.5$-sigma events (one in $10^{13}$). As an application, this allows to resolve individual Hubble patches which fulfil the condition for primordial black hole formation. It is argued that their mass spectrum is drastically altered due to co-collapse of clustered overdensities as well as the mutual threshold-lowering through the latter. Furthermore, the corresponding non-sphericities may imply possibly large changes in the initial black hole spin distribution.
arXiv:2603.28679v2 Announce Type: replace
Abstract: Introductory artificial intelligence (AI) courses present significant learning challenges due to abstract concepts, mathematical complexity, and students' diverse technical backgrounds. This paper presents an experience report examining the redesign of in-class instructional time in a university-level Introduction to Artificial Intelligence course, inspired by CS Unplugged approaches. We redesigned the summer offering, integrating embodied, unplugged simulations, collaborative programming labs, and structured reflection to provide students with a first-person perspective on AI decision-making. We maintained identical assignments, exams, and assessments as the traditional lecture-based offering. We found that students in the redesigned course reported higher attendance, stronger agreement that assessments measured their understanding, and greater overall course effectiveness, despite no significant differences in self-reported learning. Post-course interviews indicate that unplugged simulations and collaboration fostered a safe, supportive learning environment that increased engagement and confidence with AI concepts. These results highlight the importance of in-class instructional design in improving students' learning experiences without compromising rigor.
arXiv:2604.06820v2 Announce Type: replace
Abstract: LLMs make it increasingly easy to generate deceptive content at scale, creating a need for scalable misinformation risk evaluation based on whether readers find such content credible and are willing to share it. A natural approach is to ask an LLM these questions directly and treat the returned scores as predictions of the corresponding human ratings. Implicit in this practice is the assumption that asking about a reader response produces the score that best predicts it. We test this assumption using matched credibility and willingness-to-share ratings for 290 deceptive articles from 317 participants and eight LLM evaluators. Unexpectedly, the assumption holds for credibility but fails for sharing. For every evaluator, credibility scores track human sharing at least as closely as sharing scores, while sharing scores offer no detectable benefit beyond credibility when predicting responses to unseen scenarios. This pattern persists when the questions are asked separately or in reversed order. Our results suggest that directly asking for the target response may not always yield the most effective score for predicting it. Comparing direct scores with indirect paths through related judgments may reveal a more effective predictive route. Deciding what to ask may be as important as refining how to ask it.
arXiv:2604.08523v2 Announce Type: replace
Abstract: AI agents may be able to assist with emails and documents, but can they reliably complete everyday online workflows on real websites? Everyday online tasks offer a realistic yet unsolved testbed for evaluating the next generation of AI agents. To this end, we introduce ClawBench, an evaluation framework comprising 153 everyday online tasks that people need to accomplish regularly in their lives and work, spanning 144 platforms across 15 categories, from completing purchases and booking appointments to submitting job applications. These tasks require capabilities beyond existing benchmarks, such as obtaining relevant information from user-provided documents, navigating multi-step workflows across diverse platforms, and write-heavy operations like filling in many detailed forms correctly. Unlike existing benchmarks that evaluate agents in offline sandboxes with static pages, ClawBench operates on production websites, preserving the full complexity, dynamic nature, and interaction challenges of real-world web environments. An interception layer captures and blocks the final submission request, ensuring safe evaluation without real-world side effects. Our evaluations of 8 frontier models show that both proprietary and open-source models complete only a small portion of these tasks. For example, Claude Sonnet 4.6 achieves only 33.3%, which exposes gaps in current AI agents. Progress on ClawBench brings us closer to AI agents that can function as general-purpose assistants.
arXiv:2502.12984v5 Announce Type: replace-cross
Abstract: In this paper, we propose a general approach for approximate simulation and analysis of delay differential equations (DDEs) with distributed time delays based on methods for ordinary differential equations (ODEs). The key innovation is that we 1) propose an Erlang mixture approximation of the kernel in the DDEs and 2) use the linear chain trick to transform the resulting approximate DDEs to ODEs. We refer to this as the Erlang ODE approximation of the DDEs, and we prove that the Erlang mixture approximation converges for continuous and bounded kernels if the number of terms increases sufficiently fast. Furthermore, we show that if the kernel is also exponentially bounded, the Erlang ODE approximation can be used to assess the stability of the steady states of the original DDEs and that the solution to the ODE approximation converges. Additionally, we propose an approach based on bisection and least-squares estimation for determining optimal parameter values in the approximation. Finally, we present numerical examples that demonstrate the accuracy and convergence rates of the approximations and the efficacy of the proposed approach for bifurcation analysis and Monte Carlo simulation. The numerical examples involve a modified logistic equation, chemotherapy-induced myelosuppression, and a point reactor kinetics model of a molten salt nuclear fission reactor.
arXiv:2505.10811v2 Announce Type: replace-cross
Abstract: We consider the $\Set{H}_2$-optimal feedback control problem, for the case in which the plant is passive with bounded $\Set{L}_2$ gain, and the feedback law is constrained to be output-strictly passive. We show that this problem distills to a convex, infinite-dimensional optimal control problem, in which the optimization domain is the Youla parameter for the closed-loop system. We devise truncated, finite-dimensional optimizations to find sub-optimal controllers, and lower bounds on the optimal objective. Furthermore we show that both these optimizations converge to the optimal objective of the original infinite-dimensional problem as their respective domains are increased. The idea is demonstrated on a simple vibration suppression example.
arXiv:2505.23460v5 Announce Type: replace-cross
Abstract: Chirality-induced spin selectivity (CISS) refers to phenomena where molecular chirality governs spin polarization. While symmetry simply requires chiral molecules to support spin-vector correlations, we show that CISS is fundamentally a conditioned measurement of these correlations. We illustrate this principle for spin-resolved one-photon ionization of a randomly oriented ensemble of chiral molecules. We introduce and quantify the phenomenon of enantio-sensitive locking of the photoelectron current to its spin, thereby providing a complete description of spin-conditioned photoelectron currents in one-photon ionization.
arXiv:2506.00452v5 Announce Type: replace-cross
Abstract: In orthogonal frequency division multiplexing (OFDM), accurate channel estimation is crucial. Classical signal processing-based approaches, such as linear minimum mean-squared error (LMMSE) estimation, often require second-order statistics that are difficult to obtain in practice. Recent deep neural network (DNN)-based methods have been introduced to address this, but they often suffer from high inference complexity. This paper proposes an Attention-aided MMSE (A-MMSE), a model-based DNN framework that learns the linear MMSE filter via the Attention Transformer. Once trained, the A-MMSE performs channel estimation through a single linear operation, eliminating nonlinear activations during inference and thus reducing computational complexity. To improve the learning efficiency of the A-MMSE, we develop a two-stage Attention encoder that captures the frequency and temporal correlation structure of OFDM channels. We also introduce a rank-adaptive extension that adjusts the filter rank at deployment time, enabling efficient operation under resource-constrained receivers. Numerical simulations show that A-MMSE consistently outperforms baseline methods across a wide range of signal-to-noise ratio (SNR) conditions. In particular, the A-MMSE and its rank-adaptive extension provide an improved performance-complexity trade-off.