Computers and Society
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
All of arXiv
Monday 9 November
How does a mathematical community discuss AI on a federated social platform? We study 24,920 public posts from January–September 2026 by 272 accounts in a fixed roster of 1,002 Mastodon accounts whose profiles indicate mathematical research or study. LLM-assisted extraction identifies 1,191 AI-for-mathematics posts by 101 accounts. Their share of all posts rises from 1.1% in January to 11.5% in September, with contributing accounts increasing from 10 to 59. We represent discussion through linked target–aspect–stance triples, separating what is discussed from the property assessed and the author's evaluation. From July to September, capability and usefulness decline as a share of related discussion, while social and institutional concerns increase; changing contributors account for much of the latter shift. Among 38 accounts evaluating both technical quality or utility and communication, credit, or verification procedures, the procedural negative-component fraction is 19.4 percentage points higher on average (95% account-bootstrap interval: 3.8–35.1). Matching targets within posts gives a similar direction. Repeated extraction varies substantially at the level of individual triples, although replacing the repeated sample preserves the aggregate contrast. Around OpenAI's October 6 mathematics release, related-post share rises from 9.6% to 13.6%, and contributing accounts increase from 13 to 28. Post-release discussion includes appreciation of the results' potential mathematical value and applications alongside criticism of promotion centered on result counts, limited methodological disclosure, and verification arrangements.
Artificial Intelligence
Correct Answers from Insufficient Histories: A Controlled Study of Agent Memory Evaluation
An answer can match the reference for a complete agent history even when the reader's visible history does not determine it. We study this distinction using executable device logs. Our protocol separates reference agreement, certified evidence sufficiency, and context-appropriate behavior. Insufficiency requires two legal histories yielding byte-identical visible contexts but different answers; public chronology and query cutoffs certify sufficient prefixes. On 80 confirmation histories, Qwen3-32B matches 87 hidden references among 203 insufficient raw-prefix instances, with only seven appropriate abstentions. A fixed-input DeepSeek-V4-Flash replication matches 85 and abstains on 21 of the same 203 instances. Each reader makes 1,600 requests. Compact CSV's estimated appropriateness effects remain inconclusive: +3.16 percentage points, clustered 95% interval [0.00,6.57], for Qwen and +2.46, [-1.84,6.94], for DeepSeek. A separate held-out judge study achieves 99.38% correctness but retains support errors and truncations; a deterministic baseline solves all constructed cases. These results support an evaluation diagnosis across two hosted reader configurations, not gains for a general memory system or natural-language evidence assessment.
Read the full abstract
Machine Learning
MOOD: Benchmarking Post-Hoc OOD Detection for Materials Property Prediction
Out-of-distribution (OOD) detection is essential for deploying machine learning models in high stakes scientific applications, yet methods that flag OOD inputs without retraining the model have been validated almost exclusively on image classification. We introduce MOOD (Materials Out-of-Distribution Detection), a benchmark for evaluating whether these post-hoc OOD detectors transfer to graph neural networks (GNNs) on materials property prediction. MOOD evaluates 15 post-hoc detectors and 2 supervised diagnostic probes across 4 GNN architectures, 5 MatBench tasks, and 15 physically motivated compositional and structural shifts, yielding 264 evaluation settings and over 13,000 AUROC measurements across three encoder training seeds. Our evaluation reveals three key findings. First, current post-hoc detectors perform near chance; all 15 detectors achieve an AUROC around 0.5. Second, detector failures decompose into two regimes: method bottlenecks and encoder bottlenecks. Method bottlenecks arise when supervised probes can separate in-distribution (ID) from OOD samples but unsupervised detectors fail, meaning that the embeddings contain OOD signal that current detectors do not extract. Encoder bottlenecks arise when even supervised probes fail to separate ID from OOD samples, indicating that the relevant shift is not accessible in the frozen GNN representation. Third, when OOD prediction error exceeds ID error, the AUROC of distance-based and density-based detectors correlates with the size of that gap. Collectively, these findings indicate that reliable OOD detection in scientific domains requires representation learning strategies that explicitly preserve OOD relevant signal, particularly for tasks where current GNN architectures encode no extractable signal at all.
Read the full abstract
Computation and Language
Improving Confidence Estimation in Language Models with Linear Representations of Ambiguity
As language models (LMs) are increasingly deployed in downstream applications, accurately estimating their likelihood of success on a given example is critical. However, confidence estimates from LMs are poor predictors of their likelihood of success. Accurate confidence estimation is linked in part to the detection of aleatoric uncertainty; we hypothesize that LMs possess representations of at least some aspects of this uncertainty in the form of representations of the inherent ambiguity of the input. Using difference-in-means on minimal pairs of ambiguous and unambiguous sentences, we derive ambiguity representations in LM activations, and find that steering with them selectively (albeit weakly) modulates LMs' judgments of the input's ambiguity. We then propose DART, a fine-tuning method that encourages updates to lie within the span of the ambiguity directions, and penalizes changes to the LM activations outside of this subspace. DART greatly improves calibration, encodes these improvements into the model parameters, and maintains model accuracy on multiple-choice and generative QA tasks.
Read the full abstract
Differential Geometry
On evolution of mean curvature and minimal areas in general codimension for inverse problems
Many seemingly standard formulas for the evolution of the mean curvature vector in higher codimension assume only normal variations, which are ill-suited for analyzing the evolution of submanifolds- with-boundary whose boundaries may not be moving orthogonally at each time. Such situations arise when one considers inverse problems for minimal areas, for example, where such boundaries are confined to a larger ambient boundary. In this note, we provide coordinate-free identities for use in such problems, including higher order derivatives of mean curvature. We translate these identities to some of the existing inverse problem literature in codimension one. We also hope readers may find the computational framework helpful in other situations arising in geometric analysis.
Read the full abstract
Computer Vision
MDA framework: multi-architecture domain adaptation for unsupervised person re-identification
Unsupervised person re-identification not only faces the challenge of domain gaps, but existing methods also rarely support multiple network architectures. To tackle both issues, this paper proposes a multi-architecture auto-adaptive domain adaptation method: Mamba augmented multi-architecture domain adaptation framework. Specifically, this paper constructs the multi-architecture domain adaptation capability at two levels. At the first level, we propose a composite operator(MDA operator) that encapsulates Mamba as a component for enhancing domain adaptation. At the second level, based on this operator, we propose a Mamba-enhanced multi-architecture Domain Adaptation framework(MDA framework). Comprehensive experiments on person re-identification benchmarks validate the consistency and flexibility of the proposed framework. The MDA framework supports versatile deployment strategies without being confined to a single integration point. Under a single deployment configuration, it surpasses several SOTA methods. Moreover, it demonstrates substantial performance improvements across both CNN and Transformer architectures. When embedded into a CNN backbone, the framework achieves a 10.2% mAP improvement on the adaptation task from CUHK03 to Market-1501. Integrated with a Transformer backbone, it yields a 19.8% mAP improvement on the same task.
Read the full abstract
Cosmology and Nongalactic Astrophysics
Vacuum energy interacting with dark matter and dark radiation: alleviating the $H_0$-$S_8$ seesaw
Extensions of $\Lambda$CDM that raise $H_0$ with additional energy before recombination, such as dark radiation and early dark energy (EDE), also raise $S_8$. We study whether a late-time energy exchange between the vacuum and the dark sector can offset this increase. The coupling rate scales as $\bar m H^{\alpha}$ and is negligible before recombination for $\alpha \le -2$; the energy is transferred along the four-velocity of geodesic cold dark matter, so the dark matter produced by the vacuum carries no density perturbation and slows the growth of structure. We add the exchange to dark radiation ($\bar m\tilde\Lambda$CDM) and to a one-parameter EDE model ($\bar m$EDE), and constrain the models with Planck 2018, DESI DR2 BAO and Pantheon+ data (CBS), and with the SH0ES calibration added (CBSH) to test the exchange when $H_0$ is held at a high value. For CBS the posteriors favour an energy flow from the vacuum into dark matter at 1.8-2.7$\sigma$, and the exchange lowers $S_8$ by 0.02-0.08 but does not raise $H_0$: $\bar m\tilde\Lambda$CDM gives $H_0 = 68.83 \pm 0.91$ km s$^{-1}$ Mpc$^{-1}$, 3.1$\sigma$ below SH0ES. For CBSH the exchange removes the increase of $S_8$ caused by dark radiation, giving $S_8 = 0.800 \pm 0.016$ at $H_0 = 71.34 \pm 0.70$ km s$^{-1}$ Mpc$^{-1}$, and about a third of that caused by EDE for $\alpha = -2$, with the coupling at 1.1-2.0$\sigma$. The reduction of $S_8$ is concentrated at $z \lesssim 0.3$: at the redshifts probed by KiDS-Legacy the lensing amplitude decreases by only 1-2%, the shear $\chi^2$ of the dark-radiation models changes by less than 3, and the exchange reduces but does not remove the excess shear $\chi^2$ of EDE. A late-time exchange can therefore offset part of the $S_8$ increase of early-time solutions, but it does not ease the Hubble tension, and current data do not require it.
Read the full abstract
Sound
Reverberant Speech Conditioning for Material-Aware Acoustic Parameter Mapping
Dense maps of room acoustic parameters (T30, EDT, C50, DRR) describe how a room sounds at every position. Scene-wide estimation predicts such maps from a floor plan but requires a reference impulse response, which is hard to obtain. Our system instead receives a floor plan and easily recorded reverberant speech from one talker-microphone pair. We compare four speech-derived conditioning inputs: MFCCs, an embedding from the blind estimator BERP, BERP's scalar reverberation-time estimate, and an impulse response reconstructed by Rec-RIR. Our main benchmark is the MRAS dataset, where material assignments vary at fixed geometry. The reconstructed impulse response performs best among the four inputs, reducing T30 error from 31.5% with the floor plan alone to 19.9%, 83% of the reduction achieved by the reference response. We find that adding speech mainly sets the overall level of the map, with little improvement in predicting how the acoustics vary across the room.
Read the full abstract
Computer Vision
Answer or Defer? Testing VLMs' Stated Confidence on Brain MRI
Vision-language models (VLMs) could support medical-image interpretation, but their use requires deciding which interpretations to accept and which images to refer to a clinician. We test whether thresholding stated confidence can retain useful coverage while limiting errors among accepted answers, and how this depends on the population. Eight open-weight VLMs, including two reasoning models, assess 3,888 brain MRI slices from post-treatment glioma patients (BraTS 2024) and healthy volunteers (IXI) for tumor or gross abnormality. BraTS labels are mask-derived; IXI slices are treated as negatives. On BraTS, no threshold selects a nonempty set with at most 5% empirical error for any model, even when chosen retrospectively. This also holds for four models evaluated on a large-lesion, distant-negative subset. At a 10% error budget, Gemma-4-12B retains 24.4% of slices and every other model at most 2.5%. To investigate this limited selection ability, we separate confidence differences between predicted answers from differences within each answer. For four models with available outcome-level summaries, pooled mean confidence differs more between answers than between correct and incorrect responses within each answer. A covariance decomposition explains how this pattern can produce opposite confidence-correctness associations across populations. At a 20% error budget, accepting every "tumor" answer covers more slices than any confidence threshold for two models. These findings caution against using unvalidated stated confidence as a standalone deferral signal and support population-specific validation before deployment. Code is available at https://github.com/amir-sbg/VLM-Answer-or-Defer.
Read the full abstract
Computational Complexity
Simulating Nondeterministic Turing Machines in Subexponential Deterministic Time
We show that every multitape nondeterministic Turing machine can be simulated deterministically in subexponential time. Specifically, we show that the recent algorithms for $3$-$\mathsf{SAT}$ presented by OpenAI and Anthropic imply $\mathsf{NTIME}[t]\subseteq\mathsf{DTIME}[2^{O(t \log \log t/\sqrt{\log t})}]$, and that a stronger consequence of their proof yields \[\mathsf{NTIME}[t] \subseteq\mathsf{DTIME}[2^{O(t /\sqrt{\log t})}]. \] More generally, we show that any subexponential time algorithm for solving circuit-CSPs transfers to a deterministic simulation of $\mathsf{NTIME}$ with the same savings. Our results follow from new generalizations of Cook–Levin, which may be of independent interest.
Read the full abstract
Computer Vision
End-to-End Joint Training of Super-Resolution and Compression Artifact Removal for Industrial Machine Vision
In industrial machine vision, images captured by edge devices are often downsampled and compressed to reduce bandwidth, introducing resolution loss and compression artifacts that severely degrade the performance of downstream tasks, such as defect detection. Existing quality enhancement methods typically apply super-resolution (SR) or compression artifact removal (CAR) independently or in a sequential manner and are largely optimized for human visual fidelity rather than machine perception. This paper proposes an end-to-end quality enhancement framework that integrates SR and CAR with a unified training strategy that optimizes both human-perception and machine-perception objectives. We investigate fixed-weight and uncertainty-based dynamic-weight loss functions to balance reconstruction quality and machine vision task performance. Extensive experiments on PCB Defect, DeepPCB, and NEU-DET datasets show that the proposed jointly trained model generally outperforms standalone and separately trained SR-CAR pipelines. In particular, under severe JPEG compression, our method improves defect detection performance from 0.461 to 0.887 mAP@50 on PCB Defect over sequential SR-CAR baselines, with only a slight reduction in PSNR. These results demonstrate that joint optimization is critical for compression-aware industrial machine vision systems.
Read the full abstract
Human-Computer Interaction
ROS2MR: Connecting ROS 2 Motion Planning to Mixed-Reality Trajectory Visualization
Mixed reality (MR) allows users to inspect planned robot motion within their physical surroundings, but displaying trajectories from robotics software requires converting their joint configurations and timing into a format the MR application can play back. This paper presents ROS2MR, an extension of the MR-Robotics LAB platform that converts YAML-serialized ROS 2 joint trajectories, including MoveIt motion plans, into the platform's existing JSON playback format. The converter validates the input, maps arm joints by name, and resamples the trajectory at a nominal rate of 30 Hz, using cubic Hermite interpolation when joint velocities are supplied and linear interpolation otherwise, while binary gripper commands are sampled separately using a zero-order hold. A Python export helper combines sequences of motion plans and gripper states into a single trajectory and checks positional agreement at segment boundaries. The MR application was also adapted for Apple Vision Pro in addition to Meta Quest 3. A technical evaluation showed that the converter preserved the normalized duration and endpoint configurations for dense, sparse, and irregularly timed inputs. With supplied velocities, the position RMSE was approximately 2.91x10^-5 degrees for all sampling patterns; without them, sparse input increased it to 0.85 degrees, and close positional agreement did not guarantee accurate velocity profiles. A pick-and-place sequence retained its open–closed–open gripper order, although off-grid transitions were shifted to the output sampling grid. The resulting file-based pathway enables inspection of ROS-generated trajectories in MR without a live ROS connection.
Read the full abstract
Algebraic Geometry
Zero-cycles on quotients of multiple double covers and their products
We prove the 0-cycle form of the generalized Bloch conjecture for smooth projective models of subgroup quotients of multiple double covers of projective space, with smooth branch hypersurfaces and reduced simple normal crossings branch divisor. Our integral character criterion yields, in every dimension at least four, simply connected varieties of general type with 0-cycle Chow group \(\mathbb{Z}\), canonical volume four, and no birational product decomposition. We also determine optimal integral supports for componentwise product quotients and character decompositions of degree zero Chow groups. Within this covering class, we classify the minimum volume models and compute their pluricanonical map degrees.
Read the full abstract
Algebraic Topology
Multiplicativity of the Serre Spectral Sequence in Homotopy Type Theory
Working in homotopy type theory, we define relative homotopy groups and give three equivalent definitions of the product on these groups. Using this product, we establish the multiplicativity of the cohomological Serre spectral sequence. We show a connection between the homotopy product and the cup-product in cohomology, which leads to the classically familiar description of the starting page of the spectral sequence as a tensor product of graded rings. As applications, we compute cohomology rings of several classifying spaces and loop spaces of spheres.
Read the full abstract
Machine Learning
KKT-Perturbation-Aware Communication for Fully Distributed Online Convex Optimization with Nonseparable Coupling
Nonseparable costs and coupled constraints in distributed online convex optimization often arise when system performance and operational constraints depend on the joint configuration of all agents and cannot be decomposed into individual contributions. Such nonseparability leads to fully distributed decision-making with challenging partial local information, where each agent evaluates its local primal-dual update using estimated global information via a local belief of the joint action and the shared Lagrange multiplier to be exchanged with neighbors. While the belief consensus established by prior works ensures no-regret learning, exchanging beliefs comprising full-dimensional primal and dual states at every round imposes a communication overhead that scales with both the joint decision dimension and the network size. Existing event-triggered schemes typically activate communication according to state variation or consensus error. Under nonseparable coupling, however, errors of equal magnitude can produce markedly different perturbations after passing through the KKT operator. We therefore propose a randomized KKT-perturbation-aware communication mechanism, where the triggering of each agent's primal and dual communication channels depends on their impact on the KKT operator. The proposed mechanism transfers the adaptive-sampling principle of coordinate descent from allocating computation across decision blocks to allocating communication across information channels: channels with larger estimated perturbations receive greater priority for refresh. Under standard regularity assumptions, the resulting algorithm achieves $O({T}^{2/3})$ expected static regret and $O({T}^{2/3})$ expected cumulative constraint violation. Numerical experiments demonstrate improved learning and constraint-enforcement performance per transmitted bit over belief consensus baselines.
Read the full abstract
Disordered Systems and Neural Networks
How Path Integral Diffusion Decides a Finite Spin Glass
When an AI diffusion-type generator samples a frustrated spin glass, which macroscopic alternatives does it decide first, and when? We study this for bridge-diffusion models of the Path Integral Diffusion (PID) type started from a point source. We show that every isotropic PID is a time-changed Gaussian stochastic localization, and conversely every signal-to-noise schedule is realized by a PID protocol. What a continuous-time generator decides therefore does not depend on the protocol, which only sets when and at what cost. We describe the generator's state by its partial commitments: the probabilities it assigns, at each moment, to nested classes of configurations fixed by the target. The partial commitments are martingales that move only along class contrasts. Their ensemble average gives the information about each level at an exact rate set by squared magnetization contrasts. In the balanced, effectively binary approximation, a split reaches half of its class information near \gamma=1.044/\lambda_P, where \lambda_P is the leading eigenvalue of the parent covariance. The theory is applied to Bethe-lattice and Sherrington–Kirkpatrick (SK) glasses with up to 24 spins. It is observed generically that a trajectory first shifts its weight to one of the two mirror halves of the zero-field measure and then to the glassy classes inside it. These classes appear near the glass transition, keep their identity on cooling and are absent in a ferromagnet on the same graphs. Sparse (Bethe) and dense (SK) glasses differ in how often and how strongly such classes exist, not in how trajectories commit to them. For a generator that takes finitely many steps, the protocol matters through the late, local stage rather than the macroscopic decisions. Sampled posteriors at N=32,64 and small trained denoisers show the same sequence.
Read the full abstract
Artificial Intelligence
ACROSSVAM3.0: Resource-Bounded 4D Scene Construction and Adaptive Surface Association
Combining video generation, dynamic reconstruction and monocular SLAM requires consistent coordinates, explicit handling of missing outputs and predictable memory use. We present AcrossVAM3.0, an offline system that converts text, images or monocular RGB video into a versioned point representation. Frozen components run sequentially on one GPU; map corrections update cameras, points and trajectories together. The coordination layer adds no neural parameters. Three fresh input-mode runs completed with a maximum sampled task-process-tree GPU footprint of 17.14 GB. In a frozen 16-scene video suite, every scene yielded a 4D reconstruction and 15 yielded a non-empty SLAM map. Strict cross-window matching reduced local surface error but lost long-range track coverage on TAP-Vid-DAVIS. We therefore developed multi-hypothesis adaptive surface chaining (MHAC), which retains several candidate matches and an explicit missing state. After selection on disjoint development data, MHAC improved Average Jaccard over a position-only baseline by 0.1394 on a frozen ten-video holdout (95% bootstrap interval [0.1074,0.1689]). Across 265 RoboTAP videos, the mean paired gain was 0.0610 ([0.0510,0.0715]), with 215 wins, one tie and 49 losses. These results support resource-bounded scene construction and an external tracking gain under the tested protocol. They do not establish metric scale, semantic identity or high-fidelity novel-view rendering.
Read the full abstract
Computer Vision
When to Trust Adaptation: Contextual Fusion for Transferable Vision-Language Prompting
When a vision-language model encounters a new class or domain, how much should it trust its learned prompts? Because source adaptation can correct pretrained errors on some target images while disturbing valid predictions on others, we introduce Contextual Adaptive Fusion (CAF), which retains four prediction branches by pairing frozen and prompted visual and textual representations. To arbitrate when branches disagree on an image, CAF combines each branch's predictive confidence with how frequently the branches favor its proposed class across the unlabeled target collection, placing greater trust in predictions that are locally confident yet less dominated by collection-wide frequency bias. The fusion coefficients are learned once on source data and remain fixed at inference. Trained on ImageNet, CAF achieves 69.96% average accuracy across ten cross-dataset benchmarks (1.92 percentage points above ManiPT) and remains highly competitive across four ImageNet variants, setting a new best of 52.21% on ImageNet-S. In base-to-novel transfer across eleven datasets, CAF further reaches 79.77% novel-class accuracy and an 82.81% harmonic mean. Code is available at https://github.com/zhenyuanchenai/CAF.
Read the full abstract
Hardware Architecture
AscSpMM: Efficient SpMM on Ascend NPUs via Pipelined Vector-Cube Collaboration
Neural processing units (NPUs) provide high compute density and energy efficiency through specialized matrix engines and software-managed on-chip memory. However, their libraries primarily target dense computation. Sparse matrix-dense matrix multiplication (SpMM), a key kernel in graph analytics, sparse neural networks, and scientific computing, requires index-driven operand selection, indirect addressing, scattered memory accesses, and load balancing. These requirements conflict with the fixed shapes and contiguous layouts preferred by NPU matrix engines. GPU Tensor Core SpMM methods cannot be directly transferred to NPUs with decoupled vector and matrix engines. We present AscSpMM, which separates irregular operand preparation from regular matrix computation through a parameterized Vector-CSR representation. The vector engine decodes indices and gathers and reorganizes dense operands, while the Cube engine performs matrix multiply-accumulate operations. A double-slot ring buffer enables dependency-aware asynchronous pipelining, while operand reuse and runtime-adaptive tiling reduce cross-engine stalls. On an Ascend 910B3, AscSpMM achieves geometric-mean speedups of 2.36x over GEMM, 19.92x over an existing SpMM implementation, and 5.82x over cuSPARSE on an A100. It reduces matrix-input storage and peak HBM footprint by 48.24% and 22.98% on average, while remaining competitive with specialized GPU Tensor Core SpMM kernels. The code is available at https://github.com/LfLab-AI/Ascend-NPU-SpMM.
Read the full abstract
Quantum Physics
Long-distance quantum secure direct communication with Gottesman-Kitaev-Preskill qubits
The Gottesman-Kitaev-Preskill (GKP) code possesses a lattice periodic structure that enables the effective correction of small displacement errors induced by noise in phase space, which is indispensable for quantum computation. The GKP code was also used to construct a long-distance entanglement channel in quantum repeaters and distribute random keys in quantum key distribution (QKD). Using GKP encoding, we propose a quantum secure direct communication (QSDC) protocol that not only transmits the secret message directly, but also helps protect the safety of communication personnel. The exponential attenuation loss of photons can be overcome by the third-generation quantum repeater. Numerical simulation results demonstrate that secure transmission distance exceeding 1000 km is achievable with a 23 dB squeezing level. Furthermore, we establish a cost-based optimization framework to characterize the trade-off among secure communication distance, secrecy capacity, and repeater resources, enabling optimized repeater deployment and providing a systematic design strategy for scalable quantum repeater architectures. The results demonstrate the potential of combining GKP error correction for long-distance QSDC and quantum networks.
Read the full abstract
Computer Vision
Direct Trajectory Optimization for Inference-Time Alignment of Text-to-Image Diffusion Models
Alignment of text-to-image diffusion models with reward functions is a key challenge in controllable generation. Existing approaches typically rely on optimizing model inputs or fine-tuning denoising networks, which requires computationally expensive backpropagation through large denoising models. In this work, we propose Direct Trajectory Optimization (DTO), a training-free framework that performs alignment by directly optimizing the sampling trajectory of diffusion models. Specifically, DTO treats intermediate denoising scores and sampled noise as the optimization variables, eliminating the need for gradients through the denoising network. To mitigate the out-of-distribution problem, we introduce Reward-Optimized Score Guidance, which interpolates between the optimized scores and the pretrained model predictions to stabilize sampling, while providing a provable theoretical guarantee of reward improvement under constrained distributions. Empirically, DTO improves alignment performance across three diffusion models on two benchmarks, outperforming existing inference-time alignment methods while maintaining high efficiency. These results suggest that trajectory-level optimization is a promising direction for efficient and effective diffusion model alignment.
Read the full abstract
Distributed Computing
Exploiting Native Hardware Heterogeneity: Custom GPSIMD Operators for Accelerating Emerging LLM Memory on AWS Trainium
Emerging memory modules of large language models, such as Engram in DeepSeek-V4.1-Flash, mix irregular integer work (a 64-bit n-gram hash and a sparse table lookup) with dense computation (a GEMM and a gate). When deploying on a GPU-FPGA system, we can disaggregate operations to different devices for Engram to speed up by 5.9$\times$. AWS Trainium provides a similar heterogeneity natively: each Trainium NeuronCore pairs dense engines with a customizable GPSIMD Engine, and all engines share the on-chip SBUF. However, the Neuron Kernel Interface (NKI) exposes the GPSIMD cores only through fixed built-in instructions, and the official GPSIMD custom-operator toolchain does not support Trainium 2 or 3 and cannot be used with NKI. In this work, we build a toolchain that compiles C++ custom operators for the GPSIMD cores and injects them into the compiled NEFF from the NKI kernel on Trainium 3, expanding the customizability of GPSIMD with the current software stack. Over a hand-optimized NKI baseline of DeepSeek-V4.1-Flash, two GPSIMD operators speed up the n-gram hashing by 1.79$\times$ at decode, the table retrieval by up to 4.95$\times$, and the whole Engram computation by up to 1.14$\times$.
Read the full abstract
Artificial Intelligence
When Should AI Overturn a Human Decision? Controversy at the 2026 World Cup and a $\sqrt{2/\pi}$ Solution
We study how to combine a human binary call (a referee's offside or onside decision) with an AI numerical measurement (VAR's estimate of an attacker's position relative to the offside line). We derive an explicit maximum-likelihood rule that determines when the referee's call should stand and when a conflicting VAR measurement should overturn it. Our model treats the true offside margin as fixed and unknown and assumes independent additive Gaussian errors in the human's perception and the AI measurement. When both errors have zero means and known standard deviations $\sigma_H$ and $\sigma_{AI}$, respectively, the reversal threshold is $\sqrt{2/\pi}\,\sigma^2_{AI}/\sigma_H$. An onside call is overturned precisely when VAR's measured offside margin exceeds this threshold—not merely because VAR places the attacker beyond the offside line. We also derive the rule when either source has a systematic bias. We prove that the maximum-likelihood estimate of the true margin exists and is unique, while the final binary decision requires only a threshold comparison, without computing that estimate numerically. We illustrate the rule through numerical examples and show how to estimate the error parameters from calibration data. The same analysis applies beyond soccer to binary classification problems in which a human call and an AI measurement satisfy the stated model. This paper is intended for educational purposes: we aim to illustrate how undergraduate students and advanced high school students familiar with basic probability and calculus can investigate research questions in AI and its applications.
Read the full abstract
Analysis of PDEs
On the free-boundary incompressible porous media equations with surface tension
We study the free-boundary incompressible porous media equations with surface tension in dimension $1+d$. For fixed positive surface tension, we establish local well-posedness for arbitrary initial interfaces in $H^{s+1}$ and density perturbations in $H^s$, with $s>(d+3)/2$. When $s>(d+5)/2$, we justify the local vanishing-surface-tension limit for stable interfaces and sufficiently small density perturbations. In a finite-depth strip over $\mathbb{R}^d$ with a flat bottom and fixed positive surface tension, we prove global well-posedness and uniform anisotropic energy bounds for small compatible perturbations of an equilibrium with strictly stable stratification, under a horizontal low-frequency condition.
Read the full abstract
Cryptography and Security
VeilTune: Secure Fine-Tuning of Private LoRA Adapters on Public Pretrained Language Models
Fine-tuning public pretrained language models with lightweight adapters such as LoRA has become the mainstream way to adapt them to downstream tasks. A service provider often trains its adapter on a data owner's samples, so two assets need protection: the data owner's samples and the provider's adapter. Existing secure fine-tuning protocols protect the samples but reveal to the data owner intermediate results of every training step, such as the hidden representations or logits of its inputs. We show that a single step suffices to copy the adapter: from the hidden representations of one mini-batch of 32 samples, the data owner fits an adapter that agrees with the provider's on 99.2% of test inputs, and logits alone yield about 94%. Protecting both assets therefore requires that no value that depends on the adapter be revealed throughout training. We present VeilTune, to the best of our knowledge the first protocol that fine-tunes a pretrained language model with backpropagation entirely in secure multi-party computation (MPC). VeilTune computes only what the loss depends on, replaces generic MPC operators with training-aware nonlinear operators that stay accurate on the values that training produces, and obliviously skips the steps corrupted by truncation failures. On four few-shot tasks with RoBERTa-large, VeilTune nearly matches plaintext training and is 5.3 points more accurate on average than SecP-Tuning, the state-of-the-art secure fine-tuning protocol. On SST-2, it achieves a 3.5x end-to-end speedup and a 3.6x reduction in communication over SecP-Tuning, and 26x and 17x over full fine-tuning in MPC.
Read the full abstract
Mesoscale and Nanoscale Physics
The Photonic Bath at Infinity: Foundations, Derivations, and Consistency Checks
The photonic bath at infinity represents the unbounded electromagnetic environment as an effective reservoir in the nonequilibrium Green's function (NEGF) formalism. It accounts for outgoing radiation and incoming electromagnetic fluctuations, allowing radiative transport to be described consistently. We clarify how environmental fluctuations are incorporated without double-counting the outgoing radiation response already contained in the free retarded Green's function. We develop four complementary approaches to deriving the bath self-energy: infinitesimal environmental absorption, semi-infinite leads, inversion of the free Green's function, and sesquilinear forms with outgoing radiation boundary conditions. In particular, we present an explicit three-site matrix inversion for a planar geometry and introduce a sesquilinear form that relates the bath self-energy to the Silver–Müller radiation condition. We establish the connections among these approaches and obtain explicit bath self-energies for one-dimensional, planar, and spherical geometries. The resulting bath self-energies reproduce the Planck spectrum, the Stefan–Boltzmann law, and radiation pressure, and provide the radiative contribution required for energy conservation in photon NEGF.
Read the full abstract
Systems and Control
Generalizable Output Feedback Optimal Control via Transformers Without Explicit State Estimation
Classical optimal control is used to design a separate controller for each plant. This paper asks a different question. Can one learned policy serve a whole family of partially observed, noisy linear systems? We show that one transformer can. The policy is trained to imitate optimal LQG control on a heterogeneous family of Linear Time-Invariant (LTI) systems with Multiple Inputs and Multiple Outputs (MIMO), whose dimensions, cost matrices, and noise levels all vary. A shared representation with system-wise standardization, zero-padding, masking, and an encoding of the cost and noise covariances lets one set of weights control every member of the family. At run time, a short window of noisy measurements is all the policy needs to produce a control action, with no plant matrices and no state estimator. We characterize the relation between the optimal LQG policy (represented as a dynamic state-space model) and a static transformer acting on a finite window of measurements. We then express the closed-loop excess cost of any causal policy as the energy of its deviation from the optimal action. Finally, we evaluate the approach on 28 simulated systems. The trained policy reaches the optimal cost to within a median gap of $0.4\%$ for seen systems. For unseen systems, a light fine-tuning step brings the median gap to $1.0\%$.
Read the full abstract
Optics
Lensless single-pixel imaging through dynamic coded-aperture illumination
We demonstrate a lensless single-pixel imaging (SPI) system combining dynamic coded-aperture illumination with single-pixel detection. Binary digital micromirror device (DMD) masks generate tailored grayscale intensity fields through coherent propagation, without relay projection optics or temporal grayscale synthesis. Six target-field configurations—uniform, Rayleigh speckle, squared-Rayleigh speckle, and sparse binary fields with $p=10\%$, $5\%$, and $1\%$—are evaluated using binary and grayscale objects. Intensity-correlation and nonnegative total-variation-regularized reconstructions reveal that the preferred illumination depends on the object, reconstruction method, and evaluation metric. By combining rapid binary modulation, programmable grayscale illumination, and single-pixel readout, this implementation provides a flexible architecture for lensless computational imaging and joint design of optical encoding and reconstruction.
Read the full abstract
Combinatorics
The minimum forcing number of graphs and involutory weighted adjacency matrices
Let $G$ be a graph with a perfect matching. The minimum forcing number $f(G)$ of $G$ is the smallest size of a matching in $G$ that is contained in exactly one perfect matching. For the $d$-hypercube $Q_d$ ($d\geq 2$) Diwan (2019) determined $f(Q_d)=2^{d-2}$ by using a linear algebraic method, and Mohammadian (2022) generalized this method and showed that $f(G\square K_2)=n/2$ when a bipartite graph $G$ admits order $n$ and an involutory weighted adjacency matrix over a field $\mathbb F$, where $\square$ denotes the Cartesian product. In this paper, we further prove that if $\operatorname{char}\mathbb{F}\neq 2$, then $\frac{n}{2}\le f(G\mathbin{\square} P_k)\le \frac{n}{2}\lceil\frac{k}{2}\rceil$ for every $k\ge 2$, where $P_k$ is the $k$-vertex path; this recovers Mohammadian's result when $k=2$, and the lower bound is attained for $k=3$. Moreover, we extend this method to non-bipartite graphs. Recursively constructing the adjacency matrix of the folded hypercube $FQ_n$ over the binary field $\mathbb F_2$, we prove that $f(FQ_n\mathbin{\square} K_2)=2^{n-1}$ for every even $n$. Combined with Mohammadian's result for odd $n$, this establishes the same formula for all $n\geq 1$.
Read the full abstract
Computation and Language
CaseLaw: A Policy-Decoupled Framework for Training-Free Content Moderation
Content moderation is inherently policy-dependent: moderation boundaries are often subtle, subjective, and frequently updated, while conventional rule based or fine-tuning based systems require costly rule rewriting, model retraining, or system redeployment. We propose CaseLaw, a policy-decoupled and training-free framework for adaptive content moderation. CaseLaw reformulates moderation as precedent-based retrieval and voting, where multimodal content is transformed into evidence-preserved structured semantics, embedded into a stable semantic space, and classified by thresholded voting over retrieved precedent labels. We train NoteCap on a large-scale corpus of structurally annotated notes to ensure consistent semantic extraction and freeze it during subsequent policy adaptation, which requires no additional training. Policy adaptation is achieved by updating the semantic case library, including case labels and mappings, without retraining models or modifying prompts. To evaluate CaseLaw, we construct MFO-RedNotes, a benchmark for the male-female opposition risk domain. We additionally provide a five category CoPro comparison. Experiments show that CaseLaw consistently outperforms competing baselines in both precision and recall across policy-dependent risk levels and broader safety categories, with category-wise precision/recall reaching up to 86.8%/90.1% on MFO-RedNotes across both policy settings and 94.5%/92.3% on CoPro. Our work highlights the potential of precedent-based reasoning for policy adaptive content moderation and offers a practical direction toward more stable, transparent, and training-free moderation systems.