Papers

State-of-the-art visual locomotion controllers are increasingly capable at handling complex visual environments, making evaluating their real-world performance before deployment increasingly difficult. This work intends to narrow this train/evaluation gap by developing a collection of hyper-photo-realistic, closed-loop evaluation environments - The Neverwhere Benchmark Suite - comprised of over sixty 3D Gaussian Splatting reconstructions of urban indoor and outdoor scenes. Our goal is to encourage large-scale and reproducible robot evaluation by making it easier to create and integrate Gaussian splats-based reconstructions into simulated continuous testing setups. We also underscore the potential pitfalls of relying exclusively on 3D Gaussian-generated data for training, by providing policy checkpoints trained over multiple Neverwhere scenes and their performance when evaluated in novel scenes. Our analysis illustrates the necessity of sourcing diverse data to ensure performance. Code and data are available on the project page: https://ziyc.github.io/neverwhere-bench/.

Classification and regression trees are typically constructed using a greedy splitting rule that maximizes the immediate reduction in prediction error at each node. Although this strategy is computationally efficient, it can miss splits that yield small short-term gains but create substantial downstream improvements after further partitioning. We propose a look-ahead tree-building method that evaluates each candidate split by the prediction error reduction achieved after growing a conventional CART subtree below that split. Because the full look-ahead procedure can be computationally expensive, we also describe a smart look-ahead algorithm that learns downstream split values using node-level features. The proposed framework preserves the interpretability of recursive partitioning while improving split selection in hierarchical or interaction-driven settings. We conduct a simulation study comparing conventional, full look-ahead, and smart look-ahead methods under several settings and apply the proposed methods to analyze two real data examples demonstrating the merit of the new methods.

Collecting high-quality robot data for contact-rich manipulation tasks is essential for enabling robots to acquire real-world skills. However, existing data collection solutions often lack the capability to obtain stable and high-frequency tactile feedback, limiting their effectiveness in contact-rich manipulation scenarios. In this work, we propose a versatile teleoperation system with tactile-driven assistance to enable high-frequency and stable contact-rich manipulation. The proposed XRoboToolKit-T teleoperation system incorporates a tactile-informed force control architecture, designed to ensure both stable and precise force control in contact-rich manipulation during teleoperation. The stabilizer haptic module rapidly analyzes the normal force distribution and infers pseudo shear force, enabling real-time tactile-based assistance during manipulation. The refiner haptic module integrates a vision-language-action model to predict and refine manipulation actions based on tactile sensing data and task descriptions. We apply the proposed teleoperation system to challenging contact-rich manipulation tasks, including grasping a deformable rubber pipette for liquid transfer and inserting a medical syringe into a vascular training pad, to demonstrate the effectiveness of tactile-informed force control. Furthermore, the system achieves higher data collection efficiency and improved manipulation stability compared to state-of-the-art teleoperation without tactile assistance.

Simulations based on large language models (LLMs) have proven to be powerful for understanding human behavior, making them valuable additions to the social scientific toolkit. However, LLMs are ultimately black boxes based on deep neural networks which limits their value for social science. This is because of a lack of (i) interpretability: i.e. the ability to assign clear mechanisms driving observed behavior; and a lack of (ii) steerability: i.e. the ability to mute or amplify specific theoretically meaningful mechanisms of action to drive specific model behavior. Here, we demonstrate how the black box could be opened up to further enrich LLM-based simulations. Specifically, we compare three types of methods: (1) prompt-based manipulation, (2) SAE-derived feature steering, and (3) probe-based direction steering and examine their utility for LLM-based social scientific simulations. We do so by interpreting and steering two foundational components of human behaviors, namely preferences (risk attitudes, altruism) and capabilities (divergent creativity, product innovation), operationalized using four classic economic and creative tasks implemented as natural-language interactions. Overall, our results show that SAE- and probe-based techniques often outperform basic prompt-based methods for steering LLM agents, although this advantage depends on the specific prompting strategy involved. Together, SAEs and probes constitute an effective pipeline for social scientists seeking to interpret and steer agents in social simulations: SAEs decompose agents' internal representations into human-readable features, after which probes can reliably shift agents' behaviors in specified directions. We discuss implications of these methods for future work using LLM agents for social scientific simulations.

Vulnerability classification based on root cause weaknesses is essential for numerous cybersecurity activities, where the Common Weakness Enumeration (CWE) serves as a public repository of such flaws. However, its overlapping entries create a non-orthogonal structure. The result is the same vulnerability being mapped to multiple weaknesses, complicating Root Cause Analysis (RCA) and triage. To address this, NIST Special Publication 800-231 introduces the Bugs Framework (BF), which organizes vulnerabilities into <cause, operation, consequence> triples and links such triples into a causal chain, so that a vulnerability carries its root cause and its sink together instead of a single terminal label. To date, however, BF has been specified but not evaluated regarding its performance against the challenges to automated classification. The evidence required for adoption has not been investigated empirically. We evaluate BF as a classification target and a complement to CWE using a systematically screened corpus of automated Common Vulnerabilities and Exposures (CVEs) linked to CWE research. We assess the reproducibility of CVE-to-BF classification through two evaluations. The first is qualitative: an anonymized inter-rater study in which 2 subject-matter experts (SMEs) independently mapped 13 CVEs onto the four BF axes. Annotators showed strong agreement on the cause and operation axes, while the attribute axis indicated fair agreement. We also tested our automated framework across two large language model (LLM) deployments under different budgets for reproducibility analysis. Despite limitations, such as evidence availability and the absence of retrievable fix commits for closed-source software, our findings support the claim that BF is a more structured and automation-friendly framework than CWE. Our exploration reveals specific gaps in BF, including under-specified guidance on attributes.

Conversations with large language models (LLMs) can substantially shift beliefs and attitudes, raising concerns about manipulation using AI persuasion. Here we test whether a light-touch AI literacy intervention - a brief warning that LLMs can be prompted to persuade and may present information selectively - helps protect users. Across two experiments (total N = 3,208 Americans) in which participants conversed with an LLM instructed to shift their views about different political topics, the presence of a warning reduced belief change by roughly one-half (-48.1%, 95% CI [-59.5%, -36.8%]) relative to the control. Importantly, the warning did not significantly reduce trust in generative AI more broadly. Light-touch literacy interventions can help protect users against AI political persuasion.

The hippocampus, a key brain region for learning and memory, exhibits rich structural diversity, sparse communication, and robust dynamics with incredible energy efficiency. It offers promising insights for novel computing capabilities, particularly when co-designed with emerging hardware technologies. In this work, we draw inspiration from the rodent CA3 hippocampal subregion to develop the first spiking neural network with neuronal diversity and biologically-realistic resting state dynamics demonstrated on memristor hardware. We propose a network downscaling methodology utilizing a 4-prong objective function and demonstrate a small-scale CA3-inspired network with 179 Izhikevich-modeled neurons, 3 neuronal types and 17,996 synapses with similar resting-state dynamics as the orders-of-magnitude larger full-scale network. The small-scale network is mapped to an FPGA/memristor platform using a greedy algorithm and 18,316 memristors. Benefiting from memristor noise, the hardware implementation shows continuous periodic behavior, outperforming simulated hardware. This work showcases the potential of biologically-realistic algorithms on emerging hardware for neuromorphic computing.

This paper presents a Belarusian-specific data-cleaning pipeline and fine-tuning for English-Belarusian machine translation. Our cleaning pipeline distinguishes itself from others by employing a correction tool that addresses the issue of the two orthographies of the Belarusian language, noise in the training data, interference from other languages and other misspelling issues common in Belarusian on the internet. A matched ablation on unfiltered training data shows substantial benefits from filtering for all fine-tuned models, with the LLM-based models gaining roughly twice as much from filtering as the dedicated encoder-decoder MT system, supporting the view that for Belarusian MT one of the primary bottlenecks is data quality.

Binary recompilation is a technique for operating directly on executable code. It promises to automate two important tasks: retrofitting security mitigations onto legacy binaries, and migrating binaries across instruction set architectures (ISAs). Yet today, there is no fully automated system that can reliably lift arbitrary binary executables to a compiler intermediate representation (IR) such as LLVM IR, or that can fully statically and reliably translate non-trivial binary executables from one ISA to another. The main underlying problem is that recovering a program's control flow graph (CFG) statically is impossible in general: computed branches can jump to targets that cannot be determined without actually running the program. Existing systems resort to runtime fallback mechanisms, requiring a significant portion of the binary translation machinery to accompany the translated program on the target machine. This article presents a fully static, whole-program binary lifting system requiring no runtime translation support on the target. Rather than attempting to distinguish code from data, we treat every byte offset as a potential branch target and lift the entire binary in a brute-force manner, constructing a superset CFG that conservatively contains all feasible control flows. Statically unresolvable computed branches are thereby reduced to lookups in a dispatch table that points to the corresponding translated control flow path. We have implemented this approach as a prototype binary recompiler from x86-64 binaries to LLVM IR, requiring no code/data heuristics. We validate it with a fully static cross-compilation to AArch64, achieved by reusing existing LLVM backends with no modification.

Voice-assistant interruptions tend to be intrusive because existing systems fail to consider the affective state, cognitive load and situational context of the user when deciding when and how to interrupt.Voice-assistant interruptions tend to be intrusive, since existing systems do not consider the affective state, cognitive load or situational context of the user when determining when and how to interrupt. In this paper, EmpathicVA, a closed-loop framework integrating physiological sensing, vocal-affect analysis, contextual modeling and reinforcementlearning interruption policy, is introduced. A hierarchical fusion model involves integrating HRA, EDA, respiration, acousticprosodic features, linguistic embeddings, and contextual cues and computing the probabilities of five affective states. A Double Deep Q-Network selects immediate response, brief or extended delay, empathetic response, or silent mode based on these probabilities, context and interaction history. The multimodal model obtained an accuracy of 92.3% and an F1-score of 0.922 at the macro level on a held-out test set, outperforming the highest accuracy unimodal model by 6.0 percentage points. Comparing the six-week within-subject field study with 48 participants with a baseline and context-only assistants, there was a corresponding increase in satisfaction, trust, and appropriateness of timing, as well as a large reduction in interruption-related stress episodes. The results suggest that affect-aware timing and restraint are both important in voice interaction in addition to the response wording.

Pedestrian-count forecasting supports pedestrian-oriented Intelligent Transportation Systems (ITS), including crowd monitoring, pedestrian-traffic staffing and routing, and proactive risk mitigation during surges. Recent time-series foundation models (FMs) report strong zero-shot accuracy on heterogeneous forecasting benchmarks, but it remains unclear whether these gains transfer reliably to pedestrian sensing deployments. We benchmark seven univariate forecasting approaches spanning four paradigms: Seasonal Naive, gradient-boosted trees (LightGBM, CatBoost), deep learning models (N-HiTS, PatchTST), and two pretrained FMs (TimesFM, Chronos-2). Experiments cover two complementary regimes: (i) a five-day special event dataset SAIL2025 at 3-minute resolution with limited in-domain history; and (ii) Melbourne pedestrian sensors as a multi-year hourly dataset (2010–2017) with strong seasonality. We compare the MAE and RMSE results per sensor across datasets and multiple forecast horizons. Results show three consistent findings. First, with limited historical data, Seasonal Naive remains a strong baseline for long-horizon forecasting on high-volume sensors, while trained models can degrade when the next day differs substantially from prior days. Second, boosted trees can be competitive on lower-volume sensors but exhibit higher sensitivity on high-volume sensors under event-driven shift. Third, FMs excel in the seasonal and data-rich regime under long-context configuration. The findings highlight the importance of choosing pedestrian forecasting models based on both the underlying data conditions and the forecasting horizon.

Laboratory automation can improve experimental throughput, accessibility, and reproducibility, but many robotic laboratory systems remain difficult to reconfigure. This challenge is especially pronounced in free-space optics, where experiments are built from heterogeneous components, require precise alignment, and are frequently rearranged as experimental goals change. In this work, we present the Programmable Infrastructure for Cloud Optics (PICO), a robotic cloud-laboratory architecture designed to make reconfigurable optical experiments easier to program, operate, and reproduce. PICO provides a common domain-specific abstraction and software layer through which experimental configurations and actions can be controlled across different user interfaces. This enables the same physical laboratory to support remote interactive use, scripted experiments, autonomous routines, and features such as version control. We implement PICO on a robotic free-space optics platform and demonstrate it through an experimental case study.

Whitney's theorem allows isomorphism testing for connected simple graphs, apart from $K_3$ and $K_{1,3}$, to be formulated as distinguishing their line graphs. However, the relation between fixed-dimensional Weisfeiler–Leman (WL) expressivity on line graphs and on their roots remains unresolved. We study this relation through Implicit Line-Graph WL (ILG-$k$-WL), which is exactly $k$-WL on $L(G)$, executed over the edges of $G$ with line-graph relations derived from endpoint incidence and without explicitly constructing $L(G)$. On the Whitney-general class, the relation between root-domain and line-graph WL depends on $k$. For $k=1,2$, ILG-$k$-WL adds no distinguishing power beyond root-domain $1$-WL and misses some pairs that $1$-WL separates. For $k=3$, we prove the backward containment $L(G)\equiv_{3\text{-WL}}L(H)\Rightarrow G\equiv_{3\text{-WL}}H$. Strongly regular witness pairs, including the Shrikhande/rook pair, show that ILG-$3$-WL is strictly more expressive than $3$-WL. The backward containment also extends to disconnected graphs with no isolated vertices when every connected component is Whitney-general. Deterministic ILG-$3$-WL separates all three substructure-counting witness pairs, all $105$ pairs in SR25, and $359$ of $400$ BREC pairs. An untrained dense ILG-$3$-GNN gives the same pairwise verdicts on these evaluations.

Chain-of-thought reasoning has revolutionized natural language processing by enabling large language models (LLMs) to decompose problems into intermediate steps before answering. Yet confining reasoning to the textual domain presents limitations for tasks requiring direct manipulation of visual representations. Recent efforts augment multimodal LLMs with external visual expert tools such as depth estimation or object detection modules, but these remain fundamentally limited by their reliance on narrow, rigid operations that cannot flexibly generate or transform visual content. We propose ReImaGin, which leverages image generation models as a flexible visual reasoning mechanism for multimodal LLMs: unlike fixed-function tools, they accept natural language commands and can perform open-ended visual operations, like removing an occlusion or generating a floorplan from multiple disjoint views of a room. Across six diverse visual reasoning tasks including multi-view spatial reasoning and collision prediction, ReImaGin consistently outperforms both text-only reasoning and specialist vision-tool baselines, with gains of up to 25%, demonstrating the advantage of flexible, generative visual reasoning.

On a delivery platform, personalized store ranking greatly influences what users find and order. Unlike digital-only domains, candidate stores are local and bound by real-time availability and delivery operations. One central modeling tension is between surfacing new stores for trial and preserving ranking quality for sessions with reorder intent. We present Universal Venue Ranker (UVR), a production system deployed at Wolt that pairs a bidirectional transformer encoder for sequential user modeling with a GBDT ranker integrating contextual, user, and store features. Trained across all stores and domains of a country while enforcing local delivery constraints at inference, UVR replaces four previously separate ranking models (three for restaurants, one for retail) with a single unified system. Label smoothing and trial-biased sample weighting steer the model toward new stores, lifting offline trial MRR by +12% to +30% over production while regressing reorder MRR in five of six countries. These regressions leave Global CVR, our core online metric, which blends trial and reorder sessions, statistically unchanged. We validate UVR in three consecutive A/B tests, the first two across Wolt's largest operating markets and the third spanning all operating countries and both domains. UVR V1 delivers +5.5% Merchant Trial Rate and +0.16% Global CVR over the previous production ranker; V2 adds a further +0.45% Merchant Trial Rate on top; and V3, our cross-domain unification of the restaurant and retail rankers, adds a further +1.31% Retail Merchant Trial Rate, together accounting for substantial incremental gross order value and a materially simplified serving stack.

Machine learning-based partial differential equations (PDEs) solvers have attracted significant attention in recent years. Most progress in this area has been driven by deep neural networks such as physics-informed neural networks (PINNs) and kernel method (such as physics-informed Gaussian Processes). We introduce a physics-informed random feature method for countering part of the spectral bias which PINN-based solvers are facing for a certain class of PDEs. Random feature method was originally proposed to approximate large-scale kernel machines and can be viewed as a specialized randomized neural network. Compared to other state-of-the-art PINN-based solvers which require a large number of collocation points, our proposed method reduces the computational complexity. In this paper, we develop a rigorous approximation error analysis and derive high-probability error bounds on the $H^1$ norm. We provide extensive numerical tests for verifying our theoretical guarantees on error decay rates, as well as several comparison tests to showcase our claimed capability for combating spectral bias in these deep learning based methods.

Humanoid robots often execute motion commands through whole-body controllers (WBCs) that track targets while maintaining balance and stability. However, most WBCs are blind to scene geometry, which can lead to collisions from imperfect target motions that are geometrically unsafe due to perception, planning, or teleoperation errors. We propose RECAL, a Robot–Environment Cross-Attention Layer that wraps a blind WBC to trade off target tracking against collision avoidance using external scene geometry. RECAL supports collision-aware tracking of floating-base and end-effector commands, including collision avoidance for held objects. It represents the robot, held objects, and environment as point clouds, using cross-attention between robot/object points and the environment to produce geometry-aware control features. In simulation, RECAL improves collision avoidance while preserving target-tracking performance across frozen-arm and adaptive-arm locomotion, object-carrying, and standing-manipulation scenarios relative to alternative geometry-aware WBC architectures. We further demonstrate the controller on a real Digit V3 humanoid robot.

Current deep optics simulations struggle to balance high fidelity with computational efficiency, particularly for complex polarized and diffractive systems. To address this, we introduce a fully differentiable Shooting and Bouncing Ray (SBR) algorithm that simultaneously models geometric and physical optics fields. Unlike standard wave optics methods, our approach enables accurate, gradient-based optimization of polarized compound refractive-diffractive systems without the computational cost of full-wave simulations. We validate our method against analytical Fraunhofer diffraction patterns and state-of-the-art Fourier optics and wave optics methods. We showcase the practical utility of our differentiable engine through one application: designing a wide field-of-view achromatic lens with multiple Diffractive Optical Elements (DOEs).

Goldwasser et al. showed that undetectable backdoors can be planted in machine learning models trained with the Random Fourier Features (RFF) algorithm, under a hardness assumption tied to the Continuous Learning With Errors (CLWE) problem. Under standard cryptographic assumptions, even a full white-box audit of a model's weights cannot detect this class of backdoor. The construction is stated in terms of cryptographic reductions and probabilistic lemmas, without a reference implementation, and relies on secondary machinery such as the Sparse Gaussian Pancakes distribution and a homogeneous CLWE conditional density. Its realizability in ordinary numerical code is not obvious from the paper alone. This paper implements the white-box CLWE-RFF backdoor construction end to end using only numpy and scipy, to test whether this threat is realizable with commodity scientific-computing tools or requires specialized cryptographic infrastructure. We give two samplers for the core $GP_d(b_k)$ distribution. The first is a rejection-sampling proxy. The second is an exact closed-form sampler derived from the homogeneous CLWE density and verified against its own analytic form. Using this implementation, we run statistical indistinguishability tests, covering both weight-space and functional black-box comparisons. We find no evidence of detectable difference between backdoored and clean models across a range of sparsity ratios $\rho = d_{\text{sparse}}/D$. We report which parts of the construction were straightforward to realize, which required derivation not spelled out in the paper. We also highlight which parts we did not attempt to reproduce, including the underlying lattice hardness reduction. We see this work as a contribution to understanding the practical realizability of the Goldwasser white-box CLWE core, not as a new theoretical result.

Electrified chemical industries with operational flexibility can reduce operating costs by shifting production and distribution decisions in response to time-varying electricity prices. However, chemical plants operate within process networks where coordinated demand response can exploit flexibility across multiple stakeholders. Centralized coordination requires access to stakeholders' local scheduling models and proprietary operational data, often incompatible with data-privacy requirements. Distributed optimization with an independent central coordinator (ICC) avoids direct model sharing, but iterative exchange of coupling variables can still reveal private model parameters. We propose a privacy-preserving distributed coordination framework for coordinated demand response in industrial networks. The framework integrates secure aggregation with an ICC-based alternating direction method of multipliers (ADMM) algorithm, so plant-level messages are numerically masked and become useful to the ICC only after aggregation. We test the framework on a multi-plant industrial gas network in which three air-separation units jointly schedule production and shipments to shared customer regions. To support stable participation, we incorporate a two-phase revenue-sharing mechanism that reallocates savings so every plant improves relative to its decentralized status quo. In a 31-day rolling-horizon simulation with synthetic data representing heterogeneous electricity prices and demand, the coordinated policy reduces total network cost by 19.77% relative to decentralized operation and achieves a full-month cost within 3.08% of a centralized social-welfare-maximization benchmark. We further quantify a conservative worst-case collusion mode, showing how unmasked iterates and auxiliary information can expose private objective parameters.

We prove that, for every constant $\varepsilon>0$ and every function $f:\Sigma^n\to\{0, 1\}$, there is a coalition of $O(n/\sqrt{\log n})$ coordinates and a target output $b\in\{0, 1\}$ such that, after the remaining coordinates are sampled uniformly and independently, the coalition can choose its values to make the output equal to $b$ with probability at least $1-\varepsilon$. The bound is independent of the alphabet size and also holds for monotone Boolean functions on $[0,1]^n$, resolving a conjecture of Friedgut (Combinatorics, Probability and Computing, 2004). Unlike the Boolean cube setting, where Kahn, Kalai, and Linial (FOCS, 1988) give a coalition bound of $O(n/\log n)$, no sublinear bound independent of the alphabet size was previously known. In collective coin flipping, our result gives the first sublinear bound on the number of bad players needed to force a fixed output with probability at least $1-\varepsilon$ in any one-round protocol with independent uniform messages, regardless of the message length. A key ingredient in our proof is an encoding that lets us relate the influence of a function on a product space to the $p$-biased influence of the encoded function. We then rely on a structure theorem of Hatami (Annals of Mathematics, 2012) for functions with small $p$-biased influence to bias the encoded function.

Negation is typically modeled through its linguistic realization, although spoken interaction is accompanied by tightly coordinated nonverbal behavior. We ask whether contexts centered on spoken negation cues contain measurable multimodal behavioral information: whether they can be distinguished from matched control contexts without lexical or acoustic input, where this information occurs in time, which modalities carry it, and whether it extends to the dialogue partner. We study 27 human-human interviews conducted in virtual reality, comprising temporally aligned gaze, facial, head, body, hand, and finger behavior and 964 annotated negation cues. Treating classification as a predictive probe, we compare 20 time-series models while excluding lexical and acoustic information, and then systematically vary temporal context, interactional source, modality availability, and event timing. Across grouped 10-fold cross-validation, the strongest probes reach up to .75 mean held-out AUROC from speaker-side behavior. Temporal analyses show that predictive information is concentrated around cue onset but remains detectable over a broader surrounding interval, while dialogue-partner behavior carries weaker predictive information with a comparatively diffuse temporal profile. Ablation and timing perturbations further show that facial features produce the largest modality-ablation effect and that the trained probe is sensitive to the temporal organization of the observed events.

Silicon sampling uses large language models (LLMs) to simulate survey respondents. Whether it recovers cross-national variation, and why, remains unresolved. This study evaluates it against European Social Survey Round 11 (30 countries, 42 items) with two open-weight LLMs under first- and third-person prompts, plus backstory and response-format experiments. Aggregate recovery is moderate and uneven across items. Adding the country name to a three-variable demographic backstory raises the median per-item correlation between simulated and observed country means from -0.03 to 0.52, and the richer profiles tested add no consistent gain. The respondent's country label acts as a country-level assumption that respondent detail does not revise. Naming the response-scale endpoints in words stops the model from ranking countries backwards, so the answer format sets the direction of the ranking. Individual-level recovery remains negligible in every condition and does not track aggregate recovery across countries. An average of neighboring countries, using no LLM, recovers country levels more accurately than every model condition and ranks them about as well. Silicon sampling can thus support exploratory country-ranking comparison after item-level validation and with the response format reported. It does not support individual or distributional inference.

We introduce ParsHate, a manually annotated dataset of 10,000 Persian tweets spanning 2013-2022, representing the first decade-long benchmark for hate speech detection in Persian. The dataset contains 31% hateful content and supports both hate detection and multi-label fine-grained target identification across seven structured target categories. ParsHate also distinguishes explicit and implicit hate, marks explicit and implicit targets, and provides span-level rationales. Data collection combines random and score-stratified temporal sampling to reduce keyword-driven bias while preserving natural label distributions. Applying SOTA models for Persian hate-speech detection on ParsHate shows moderate performance (79% F1), especially with samples from earlier years, and low performance with target identification (25.5% macro-F1). This emphasizes the diverse sampling of hate speech in ParsHate and its challenging nature that requires more advanced methods for better performance. Dataset is made publicly available.

The Skolem Problem asks whether a given linear recurrence sequence (LRS) has a zero term, and is equivalent to proving an effective version of the Skolem-Mahler-Lech theorem, which states that a non-degenerate LRS has finitely many zeros. Decidability of the Skolem Problem however, has remained open for many decades. Bilu et al. (2022) showed that the Skolem Problem for simple LRS is decidable subject to the weak $p$-adic Schanuel Conjecture and the Skolem Conjecture (also known as the exponential local-global principle), the latter of which states that an LRS has an integer zero if and only if it has a zero modulo every integer $m$. This paper works towards understanding the Skolem Conjecture. We provide an example showing that a natural strengthening of the Skolem Conjecture, where we restrict $m$ to be a prime power, is false. This failure is accounted for by the relationship between $p$-adic zeros of LRS (which arise naturally in the proof of the Skolem–Mahler–Lech theorem and have been studied algorithmically by Bacik et al. (2026)) and twisted rational zeros (introduced by Bilu et al. (2025)). By studying this relationship further, we are able to prove the main result of this paper: a local-global principle for simultaneous zeros of two coprime LRS, subject to the $p$-adic Schanuel Conjecture. A fundamental step in proving this result is characterising when twisted rational zeros are (or are not) $p$-adic zeros for infinitely many primes $p$, which we do unconditionally. This also answers two open questions of Bilu et al. (2025). Finally, we conjecture that the existence of twisted rational zeros is the only way that the aforementioned strengthened Skolem Conjecture may fail, which is supported by a heuristic argument.

Weight-only post-training quantization is the cheapest way to shrink a retrieval embedder, and the received advice for applying it – protect the embedding table, allocate bits by module sensitivity, prefer a ranking-aware objective over weight reconstruction – was carried into LLM quantization largely intact. We test that advice on retrieval embedders directly, quantizing five checkpoints from four architecture families across a grid of bit widths and group sizes, and isolating the embedding, attention and feed-forward blocks at each width. Every heuristic fails to transfer as stated. The embedding table never emerges as the dominant isolated protection priority in any family, despite being the largest tensor in several of them. Module sensitivity does not survive as a transferable ordering: at INT4/g16 the spread between modules is too small to allocate against, at INT3 the ordering becomes family-dependent and joint damage stops being the sum of its parts, and at INT2 comparable reconstruction error accompanies retention ranging from 1.3 to 65.9 percent of full precision. A cheap reconstruction proxy is useful for screening uniform bit widths but substantially less reliable for choosing which tensors to protect; its apparent strength across the whole grid is a range-extension artifact. A distilled 109M student at INT3 holds 78.04 NDCG@10 in 68.4 MB and dominates the extreme-PTQ arm of its own 0.6B teacher, 297.9 MB at 64.46, on both size and quality – but only inside the task it was distilled for. Sizes are byte counts of files that exist rather than arithmetic estimates, and the measurement repository carries the byte provenance for every one of them.

Employers increasingly delegate initial screening to automated systems, which in many cases reject an application before any human reads it. Acceptance of such systems plausibly depends both on how well they perform and on the procedure that produces the decision. Prior studies rarely vary procedure and performance independently, leaving it unclear whether applicants value procedure for its own sake or for the errors it corrects. In a preregistered paired-profile conjoint experiment, 1,919 United States job seekers made eight choices between systems with independently randomized levels of decision authority, error rate, explanation, opt-out, appeal, and independent bias audit. The value of the appeal, the opt-out, and the bias audit did not rise as wrongful rejections became more common, each staying within a preregistered equivalence bound. Human involvement carried more weight than any procedural feature, moving stated choice about as much as cutting wrongful rejections from 30% to 10%. These patterns constrain a simple error-correction account and are consistent with applicants valuing procedure partly for its own sake, so that improving a system's performance does not substitute for a right applicants can invoke.

Modern GPUs require not only SIMT-style parallelism but also software-managed concurrency between compute and data movement to reach maximum performance. Performance engineers must reason about subdividing work into the hierarchy of computation resources (threads, warps, warpgroups, blocks, clusters), and, in many cases, also must use asynchronous tensor core and memcpy instructions on different levels of the memory hierarchy (registers, tensor core accumulators, shared memory, global memory). Unlike CPUs, where out-of-order execution is managed by hardware and hidden from programmers, GPUs expose explicit instruction reordering to software through these asynchronous instructions. Well-established GPU programming languages generally offer either direct low-level control without safety guarantees (e.g., CUDA C++ inline assembly or intrinsics) or easier-to-analyze, high-level abstractions (e.g., Triton's tile-based model) that hide asynchronous instructions in the compiler backend, which may prevent performance engineers from maximizing performance by tuning critical details. We propose Exo-GPU, an imperative, low-level language that creates minimal abstraction over CUDA. Our key idea is to treat parallelism and synchronization as mere annotations on sequential code rather than as fundamental control flow primitives, enabling verification that these constructs do not alter the program semantics. The benefit is twofold: programmers can reason about code without hidden control flow or mutation, while allowing the Exo-GPU compiler to verify sequential-parallel equivalence–guaranteeing that parallel execution is functionally equivalent to its sequential interpretation. We used Exo-GPU to author GEMM kernels for the H100 GPU, using wgmma, TMA, and split-k. Our kernels achieved over 80% of theoretical peak on large problem sizes, in some cases outperforming the vendor-provided CUBLAS library.

Interocular differences in visual perception can induce a variety of effects when fused by the brain. For example, prior works have found that carefully crafted binocular differences in local detail can improve contrast. It has also been found that when the frequency content of two stimuli are slightly different, blur suppression leads to a fused percept that is typically dominated by the sharper image. In this paper, we develop a psychophysical framework to measure the perception of natural image stimuli with interocular frequency differences across the visual field. To this end, we study the effect of dichoptic foveation, which we define as the application of blur to one eye and a simultaneous sharpening filter to the other. Stimuli were viewed in a virtual reality (VR) head-mounted display (HMD) and placed at different retinal eccentricities. Study data were scaled to a perceptual just objectionable difference (JOD) scale, and a 4D model was fit to it; our results suggest that interocular frequency differences can be well described by a simple computational model. We applied the model in a realistic VR scenario with free exploration of 360{\deg} videos to improve a base foveated rendering system by enhancing high frequency information dichoptically.

A language model's representation geometry is not predetermined; it evolves as the model runs. A faithful account of that geometry must capture that dynamic process, and so cannot be based solely on model-independent statistics such as co-occurrence. Here we introduce a mean-field analysis of attention. The average attention from one token to another defines a kernel that carries representations layer to layer and can be iterated through the network to model how the geometry is transformed. We condition this average two ways. Conditioned on a whole corpus, the kernel predicts the average-case evolution of representation geometry. Conditioned instead on a single context, it predicts the expected geometry for that context. A head's departure from that prediction, its mean-field deviation, isolates the context-specific computation that the mean field misses. Under the corpus-conditional reading, the kernel yields an open-loop model: from the input embeddings and the frozen weights alone, we can iterate the kernel and the model's own MLPs over token representations, never consulting a measured deviation at any layer. The resulting prediction is highly accurate. In early training the model and its corpus mean field are indistinguishable. Replace every attention head with its mean field, and the substitution leaves the loss on real text unchanged. Around the onset of induction, the two diverge, and the gap widens as representations become contextualized. Under the context-conditional reading, deviation from the mean field is a task-agnostic measure of context-specific computation. The residual decomposes additively into unusual attention routing and contextualization of the transported values. Across controlled induction and few-shot settings, greater deviation tracks greater reliance on in-context information.