Papers

Medical image interpretation is high-volume and time-consuming, and while AI interpretation can reduce workload, fully autonomous deployment carries potential safety concerns and low specificity may in practice lead to increased clinician workload. Learning to Defer (L2D) addresses this by selectively routing cases between autonomous prediction and human experts by learning from input features and AI model and human performance. While theoretical guarantees have been proven for L2D, its performance has not been validated on real-world medical datasets with human reader annotations. We evaluate the predictor-rejector formulation of two-stage L2D, where the AI predictor model is fixed and separate from the trainable routing or rejector model, on Collab-CXR, a multilabel chest X-ray dataset with multiple human annotations per case. This is the first work to look at L2D in the context of real-world medical imaging data with human annotations. We further introduce a new setup, L2D with Guidance, where the decision space is extended to three choices: predict autonomously, defer to a human expert, or defer to a human expert and provide AI guidance. We compare multiple rejector architectures and loss functions, and different input feature availabilities. This is reproduced on two larger datasets, VinDr-CXR and CheXpert. Our results show that two-stage L2D with Guidance outperforms classic two-stage learning to defer, as well as human-alone, AI-alone and AI-guided human baselines. Notably, this performance is achieved with simpler loss functions compared to formally defined L2D surrogate loss functions in current literature.

This contribution presents the new phase-2 CMS baseline tracking strategy currently proposed for online event reconstruction, which can be similarly useful in offline reconstruction under the extreme pileup conditions of the High-Luminosity LHC (HL-LHC). By unifying diverse algorithmic paradigms and integrating machine learning (ML) techniques into a coherent sequence, CMS can maintain high-efficiency and high-resolution tracking while improving the fake track rate and the computational performance, a crucial aspect especially for the HL-LHC. The approach combines GPU-optimized algorithms (Patatrack, LST) with vectorized CPU methods (mkFit), exploiting modern hardware to maximize throughput. This heterogeneous strategy reduces the resource requirements and enhances the physics reach by incorporating a global displaced track reconstruction and extending sensitivity to rare, interesting signatures, such as long-lived particles.

We consider right-linear operators $T$ on a right Hilbert module $V$ over the Clifford algebra R_n, whose S-spectrum lies in an acute double sector. For these bisectorial operators, we introduce spectral projectors P_\pm associated with the two cones of the double sector. They decompose the Hilbert module into two submodules V=V_++V_-, and the bisectorial operator T into two sectorial operators T|_\pm. A crucial but non-trivial, cornerstone in this theory is the boundedness of the projectors P_\pm, which is, in turn connected to a bounded H^\infty-functional calculus of the operator T. We provide two practical criteria: either the squared operator admits a bounded H^\infty-functional calculus, or the operator is m-accretive. Finally, we apply these results to the gradient operator \nabla_a with nonconstant coefficients. For the particular gradient with constant coefficients, we are even able to derive explicit representations of the submodules V_\pm and the projectors P_\pm in Fourier space. Moreover, we identify the sign of the gradient operator with the Clifford-Hilbert transform. This sign plays a central role in the fractional powers of vector operators, which are used, for instance, in the non-local Fourier law of heat propagation.

Transforming scholarly PDFs into machine-readable fulltext remains a bottleneck for large-scale information systems. Recent vision-based parsers improve accuracy, but need GPUs and may introduce noise into the extracted text. GROBID, a modular font-stream parser running on CPU, is the de-facto standard for structuring scientific articles and underpins several of the largest open scholarly corpora. We pair it with a lightweight CPU detector localising figure, table, and paratext (header, footer, page number) regions, encoded as typed-area masks whose tokens are routed to GROBID's specialised models or discarded. On two PMC corpora, Bioinformatics (1,926 articles) and Materials Science (2,595), scored against JATS with a section-aware structural protocol, our extension improves over plain GROBID on most metrics (NS $+0.025$/$+0.013$; $+0.086$ paragraph recall on Materials Science, $d_z{=}1.08$), and caption-linked figure recovery improves on both corpora. On the external Table-BRGM benchmark, table detection recovers F1 $0.16 \to 0.94$ and table structure follows (GriTS-Top $0.27 \to 0.78$, below the strongest GPU system). On body text, against four vision-based systems (Docling, MinerU, olmOCR, dots.ocr), it has the best paragraph precision on both corpora, the best section detection on Materials Science, and a character error rate within 0.004 of the best GPU parser. End-to-end on CPU, it costs $2.7$–$3.2\times$ less than the cheapest GPU system (Docling) and $10$–$14\times$ less than generative parsers.

X-ray polarization models of accreting compact objects often assume the so-called Chandrasekhar-Sobolev polarization prescription for elastic Thomson scattering in electron-scattering semi-infinite atmospheres. In this letter, we examine the effects of inelastic Compton scattering. We assume homogeneous slabs of cold and warm (up to 1 keV) electrons. We use a Monte Carlo approach. We discuss the polarization difference arising in the X-rays between the elastic and inelastic scattering assumptions for plane-parallel transmission and reflection, and the role it can play in the interpretation of X-ray polarimetric observations of accreting compact objects.

Motivated by the recent observation of the $B_c^*$ meson by the ATLAS Collaboration and its promising discovery potential at future high-luminosity colliders, we investigate the semileptonic decays $\bar{B}^*_{c} \to (P, V) \ell^- \bar{\nu}_{\ell}$ ($P=D$, $B_{d,s}$, $\eta_c(1S,\,2S,\,3S)$; $V=D^*$, $B^*_{d,s}$, $\psi(1S,\,2S,\,3S)$) in the Standard Model. Using the relevant form factors obtained in the self-consistent covariant light-front quark model, we present theoretical predictions for several physical observables, including the branching fraction, the ratios of branching fractions $R_{(L)}$, the longitudinal polarization fraction of the daughter vector meson, the $\tau$ lepton spin asymmetry and the forward-backward asymmetry. It is found that the branching fractions for the Cabibbo-favored $\bar{B}^*_{c}$ $\to$ $\bar{B}^*_{s} e^- \bar{\nu}_e$ and $\bar{B}^*_{s} \mu^- \bar{\nu}_\mu$ decays reach the order of ${\cal O}(10^{-6})$, which might be firstly observed at the future high-luminosity Large Hadron Collider.

Successful mobile manipulation requires coordinated base and arm motion while maintaining accurate spatial positioning. However, demonstration-trained policies can struggle to realise the intended base motion reliably, leading to spatial misalignment and subsequent manipulation failures. We present MAVP (Map-Aware Visuomotor Policies), a framework that improves execution reliability by predicting explicit base-pose targets and tracking them using localisation feedback. MAVP reconstructs a static map from teleoperated demonstrations and expresses demonstrated base trajectories in a shared map frame, providing consistent spatial supervision across demonstrations. At execution time, the policy receives RGB observations, joint states, and the robot's current map-frame base pose, and jointly predicts target base poses, arm actions, and gripper actions. A low-level controller tracks the predicted base targets using feedforward motion and pose error feedback, enabling correction of execution deviations. We additionally use pose-noise augmentation during training to improve robustness to errors in the policy's pose input. Across six real-world manipulation tasks and three policy families, MAVP achieves higher task success rates than unanchored velocity control in all tasks. Videos and additional results are available at https://123qwedsa123.github.io/mavp/.

Fairness-aware distributed learning prioritizes clients with large losses to reduce performance disparities, but label poisoning can create large losses, thereby inducing a fairness–robustness conflict. We propose FairMean to manage this conflict. FairMean weights client gradients using a bounded, nondecreasing function of local loss. The increasing weights prioritize high-loss clients to promote fairness, while the upper bound prevents excessive loss-induced amplification of poisoned-client gradients. In the absence of label poisoning, we show that minimizing the FairMean objective is more conducive to solution fairness than minimizing the standard average-loss objective. Under label poisoning, we establish an average-stationarity bound whose attack-dependent term is proportional to the square of the poisoned-client fraction. Experiments show that FairMean promotes fairness by reducing accuracy variance while improving worst-client accuracy.

Charged particle track reconstruction is one of the heaviest computational tasks in the event reconstruction chain at LHC experiments. Furthermore, projections for the High Luminosity LHC (HL-LHC) show that the required computing resources for single-threaded CPU algorithms will exceed those that are expected to be available. It follows that experiments at the HL-LHC will need to employ novel and evolved track reconstruction algorithms, within heterogeneous computing systems that include many-core CPUs as well as GPUs, in the attempt to maximize the computational performance while retaining the best possible reconstruction efficiency. In the context of the CMS High Level Trigger (HLT) at the HL-LHC, the mkFit algorithm, already in use for the CMS track reconstruction during the LHC Run 3, will exploit its parallelized and vectorized nature on CPUs to perform pattern recognition using seed tracks produced with algorithms that are designed to be fully parallelizable and hardware agnostic, thus suitable for heterogeneous systems: the Patatrack and the Line Segment Tracking (LST) algorithms. Patatrack is an established algorithm, already used for the CMS pixel track reconstruction at HLT during the Run 3 of the LHC, while LST is a novel algorithm, recently integrated in the CMS software, targeting the reconstruction of tracks in the outer tracker of the HL-LHC CMS detector. The state-of-the-art performance for the CMS HLT track reconstruction at the HL-LHC is presented, obtained using the combination of the mkFit, Patatrack, and LST algorithms, that in turn use machine-learning (ML) techniques such as deep neural networks and multi-objective particle swarm optimization to suppress duplicate and misreconstructed tracks. Prospects of further improvements are also presented, with a focus on the usage of ML techniques for track reconstruction at CMS.

Commercial image editing requires product identity preservation, accurate text rendering, and user appeal alongside general editing quality. We present KwaiMind, an image editing system combining general capabilities with e-commerce specialization. An agent-based data engine maintains approximately 1.8 million high-quality editing pairs. Built on a multimodal diffusion transformer, KwaiMind undergoes continued pre-training and supervised fine-tuning, followed by preference optimization and online reinforcement learning. A general-purpose vision-language judge and specialized rewards for click-through rate (CTR), text rendering, and product consistency guide specialized policies, which are consolidated through on-policy distillation. We introduce Ecom-Bench, covering 11 commercial editing tasks with task-specific visual evaluation and CTR-based ranking. KwaiMind achieves the strongest overall scores among evaluated open-source editors on ImgEdit, GEdit, both language splits of REDEdit, and Ecom-Bench visual quality, and the highest aggregate CTR ranking score among compared systems. Offline, CTR-guided optimization increases the proportion of generated images whose predicted CTR exceeds that of the original product image from 12.16% to 37.41%. In an online A/B experiment, CTR-based selection of product main images yields an approximately 2.44% relative increase in actual CTR. These results demonstrate the value of domain-specific data and reward-driven alignment for commercial image editing.

This work presents a novel approach to detect misclassification errors in CNNs caused by soft errors in their parameters. We propose an architecture that uses inter-class relations induced by the CNN that needs protection. The architecture has minimal resources overhead and does not require modifying the CNN, which makes it a competent solution that can be used with other error protection techniques. We have validated the architecture with five different combinations of modern dataset-model pairs: ImageNet-1K for ResNet-50 and EfficientNetV2-Small; CIFAR-10 for MobileNetV3, ShuffleNetV2-Small with 2.0x output channels and MNASNet with depth multiplier of 1.3. The validation process was done rigorously with statistical significance, from the creation of the datasets used by the architecture to the acquisition of experimental results. Results show great performance with over 90% accuracy in detecting single errors and great error detection over multiple Bit Error Rates, which can be potentially increased with hyperparameter tuning.

Topological spin textures have attracted considerable attention through their emergent electromagnetic properties, including the topological Hall effect and the topological Nernst effect. Among them, magnetic hedgehog lattices (HLs) are three-dimensional topological spin textures that host emergent magnetic monopoles and antimonopoles. HLs have been identified in metallic compounds, such as $\mathrm{MnSi}_{1-x}\mathrm{Ge}_x$ and $\mathrm{SrFeO_3}$. Various theoretical models have been proposed, with either short-range or long-range interactions, which are typically regarded as effective descriptions of insulating and metallic systems, respectively. However, the role of the interaction range in stabilizing the HLs remains only partially understood. To address this gap, we investigate the stability of HLs by systematically varying the range and spatial decay of the exchange interactions, thereby interpolating between interaction profiles commonly associated with localized-spin insulating models and itinerant-electron metallic models. Based on extensive Monte Carlo simulations, we find that HLs are stabilized over a broad interaction range. Intriguingly, among the interaction profiles examined, intermediate-range interactions are particularly favorable for stabilizing HLs. They give rise to two distinct types of $3Q$-HLs that differ in the number of monopole-antimonopole pairs per magnetic unit cell, and their relative stability depends sensitively on the interaction range. Our results provide a unified framework for tuning phase diagrams from microscopic interaction models, serving as a guideline for interpreting experiments in candidate materials.

Simulating the dynamics of dissipative quantum many-body systems governed by local Lindbladians is a fundamental task in quantum computation. While the near-optimal gate count has been achieved for Hamiltonian simulation, comparable results for Lindbladian simulation remain elusive. In this work, we develop quantum algorithms for lattice Lindbladians by exploiting their locality. First, we consider sparsely dissipative systems, in which the dissipation is sparsely located, including boundary-driven systems. We establish a near-optimal quantum algorithm for their dynamics with gate count $O(Nt \, \mathrm{polylog}(Nt/\varepsilon))$, where $N$ is the system size, $t$ is the evolution time, and $\varepsilon$ is the allowable error. We then consider generic lattice Lindbladians with finite-range interactions and dissipation, and develop an algorithm for simulating time-evolved observables with gate count $O((Nt)^{4/3} \, \mathrm{polylog}(Nt/\varepsilon))$ per sample with the sampling complexity $\Theta(\varepsilon^{-2})$. The gate count has the smallest known dependence on the system size among algorithms retaining polylogarithmic dependence on $1/\varepsilon$. Our algorithms are based on two techniques that exploit locality: patching and merging. Patching decomposes dissipative dynamics into dynamics on subsystems with exponentially small error, generalizing a key idea underlying the Haah-Hastings-Kothari-Low algorithm for near-optimal Hamiltonian simulation. Merging absorbs reversed dissipative dynamics into other parts of the evolution, substantially reducing the overhead associated with quasi-probabilistic sampling. These results demonstrate that locality can be fully exploited to achieve fast quantum simulation of dissipative many-body dynamics, opening the way to applications such as predicting nonequilibrium phenomena and preparing desirable quantum states.

We establish lower bounds for the critical CR GJMS operator on closed pseudo-Hermitian manifolds. In dimension $2n+1$, assuming embeddability, we obtain a lower bound involving the Heisenberg Sobolev norm of order $n+1$, modulo an $L^{2}$ term, on the orthogonal complement of the kernel. In dimension three, we prove that the CR Paneitz operator is bounded below without assuming embeddability. The proofs combine the Heisenberg calculus with a square-root construction based on the joint functional calculus of the sub-Laplacian and the Reeb vector field on the Heisenberg group. As an application, we show that the infinitely many negative eigenvalues of the CR Paneitz operator on the Rossi sphere have zero as their only accumulation point. Consequently, this operator does not have closed range.

We develop a new endgame for numerically computing singular solutions to complex analytic systems by means of a homotopy path whose endpoint is the solution to be computed. When the solution is singular, prediction-correction path tracking may fail at the endpoint. Instead of tracking the path all the way to its end, endgames sample nonsingular points on the incoming path and approximate the endpoint by interpolating the samples with a truncated Puiseux series. By collecting sample points along a logarithmic spiral in the complex plane, the proposed spiral endgame balances between the superior numerical conditioning of a circular sample (known as the Cauchy endgame) and the update efficiency of a linear sample (known as the power series endgame), these being tied together as two extremes of the new approach. The numerical conditioning of the methods are analyzed and compared, and computational examples illustrate the effectiveness of the new spiral sampling method compared with the circular sampling method of the Cauchy endgame.

In this paper, we study the minimum-cost bipartite matching with penalties problem in metric spaces with bounded doubling dimension: Given two disjoint sets $R, B$ in a metric space $\mathcal{M}$ with $|R|+|B|=n$ and a penalty function $p \colon R \cup B \to \mathbb{R}_{\ge 0}$, the goal is to select a set of pairs in $R\times B$ so that every point belongs to at most one pair and the sum of the distances of the selected pairs and the penalties of the points not belonging to any pair is minimized. While near-linear time approximation algorithms are known for the minimum-cost perfect matching problem in geometric settings, no such algorithm was previously known for the penalty setting. We present a randomized algorithm that computes a $(1+\varepsilon)$-approximate minimum-cost bipartite matching with penalties in $O(n \mathrm{poly}(\log n, 1/\varepsilon))$ time with high probability. To the best of our knowledge, this is the first near-linear time approximation algorithm for the problem in the penalty setting.

Long-horizon and multi-turn agents typically generate short actions and process long observations from tools and environments. This growing context demands efficient prefill, compact KV-cache storage, and accurate long-context retrieval. To meet these demands, we introduce HySparse2, a hybrid sparse attention architecture with two-level KV sharing. At the outer level, KV Bridging adopts a YOCO-style self-decoder and cross-decoder structure, but bridges only full-attention layers. The self-decoder uses hybrid sliding-window attention (SWA), while the cross-decoder uses hybrid sparse attention. The KV caches for full-attention layers in the cross-decoder are generated from the hidden states of full-attention layers in the self-decoder. At the inner level, HySparse2 retains HySparse's core KV Reuse design with two refinements. First, it replaces block-level sparsity with token-level sparsity for finer long-context retrieval. Second, it removes the separate SWA branch from sparse layers and instead forces a sliding window of recent tokens into the sparse selection. This two-level KV sharing allows all cross-decoder KV caches to be constructed from self-decoder hidden states. Prefill can therefore exit after the self-decoder, skipping all cross-decoder layers. On an 80B-A3B MoE model, HySparse2 outperforms HySparse and Hybrid SWA on long-context retrieval and multi-turn agentic tasks, while substantially reducing prefill computation and KV-cache storage.

The realization of entanglement distribution (ED) in large-scale quantum networks (QNs), as a fundamental theoretical framework of quantum communication, faces significant challenges. Percolation theory, drawn from statistical physics, has been identified as a potential solution to this problem. Nevertheless, when addressing ED, percolation theory primarily relies on numerical simulations and approximate algorithms. This approach often lacks interpretability and controllability for many phenomena occurring during the ED process in large-scale networks, including emergent phenomena near the threshold. To address these limitations, in this paper, we {focus on} leveraging operator theory to construct a mathematical framework for entanglement percolation in series-parallel QNs. Through the application of operator-theoretic methodologies, we analyze several critical issues within the entanglement percolation process.

Inspired by the static problem $(\alpha,\beta)$-Dominating Set, we propose a general temporal domination problem, called $(\alpha,\beta)$-Temporal Dominating Set ($(\alpha,\beta)$-TDS). We show that this problem encompasses Temporal Dominating Set, and additionally provides first temporal extensions of problems such as $k$-Dominating Set and $\alpha$-Dominating Set. In this paper, we study the parameterised complexity of $(\alpha,\beta)$-TDS with respect to temporal neighbourhood diversity (TND), temporal modular-width (TMW), and temporal cliquewidth (TCW). We obtain fixed parameter tractability results for all values of $\alpha$ and $\beta$ with respect to TND; W[1]-hardness with respect to TMW and TCW whenever $\beta$ is in the problem input, or whenever $\alpha\in (0,1)$ and $\beta$ is a fixed constant; and para-NP-hardness with respect to TCW when $\alpha = 0$ and $\beta = 1$, or $\alpha = 1$ and $\beta = 0$.

The rapid growth of large language model training and serving is driving AI data centers (AIDCs) toward gigawatt scale. Unlike conventional commercial loads, AIDCs possess significant operational flexibility through dynamic voltage and frequency scaling (DVFS) of training and inference workloads, while periodic model checkpointing can induce abrupt power drops and rebounds that erode operating reserves and increase transmission congestion risks. Coordinating AIDC operation with grid scheduling under these unique operational characteristics is challenging because grid and AIDC operators are generally unwilling to share proprietary data and decision-making authority. This paper proposes a hierarchical privacy-preserving coordinated operation scheme between the power grid and AIDCs to address this gap. The proposed scheme contains three phases. In Phase I, the grid operator computes a certified inner approximation of the AIDCs security region for subsequent coordination. In Phase II, the AIDC operator coordinates training and inference AIDCs to optimize workload allocation within the certified security region and generate power schedules and checkpoint alerts. In Phase III, the grid operator solves a checkpoint-aware two-stage robust optimal power flow (OPF) considering renewable generation and checkpoint uncertainties. By exchanging only compact interface information, the framework preserves the privacy of both grid and AIDCs, avoids frequent iterative communication, and enables secure coordination with guaranteed feasibility. Numerical studies on a modified IEEE 14-bus system and a modified NYISO system demonstrate the effectiveness, robustness, and security of the proposed framework.

Optimal sizing problem (OSP) for battery energy storage system (BESS) under uncertain inputs is often formulated as a two-stage stochastic program (2SP), which typically introduces computational and RAM (memory) bottlenecks for large scenario sets. Benders Decomposition (BD) is a popular approach for tackling this problem, but it suffers from slow convergence to the exact solution. To address this problem, we propose an accelerated online learning-based adaptive hybrid BD algorithm for a risk-averse 2SP formulation. The proposed method avoids getting stuck in the infeasible region during early iterations by explicitly embedding a carefully selected scenario in the master problem, while a tail-relevant scenario selector based on online learning helps to avoid solving the entire scenario set at every iteration. The OSP is formulated to select a behind-the-meter BESS in a multi-site energy community with existing renewables. The use case explores an interesting middle-ground between deterministic acceleration methods and training-heavy ML models, showcasing the potential of ML-assisted decision-making. Compared with vanilla BD, OLAH-BD reduces subproblem evaluations and total wall time by up to 80% under the same tolerance settings.

Real-time avatar services run long-lived, GPU-backed sessions that transform a continuous text stream into speech, facial motion, rendered video, and RTC media. Elastic operation repeatedly reassigns these sessions across Workers. A reassignment updates shared placement state while the previous GPU runtime and RTC connection remain alive, opening a gap between control-plane assignment and execution authority. A stateful controller serializes ownership by retaining per-session authority, placing that state on the controller's scaling and failover path. We present DHSched, a control plane that manages long-lived stateful sessions through stateless peer Dispatch replicas. DHSched externalizes session ownership as a generation identified by (WorkerID, epoch). Dispatch conditionally commits each generation. Infer-Controller claims and revalidates the same generation before releasing state-advancing input. A stale runtime loses execution authority even if it remains alive. DHSched ranks candidates from periodically refreshed load views while keeping capacity admission authoritative. Recovery advances the same generation path via source-epoch guards and source-independent reconstruction. DHSched runs in our production avatar-serving stack, which reached 49,987 concurrent live sessions. On 10 June 2026, it completed 9,927 session migrations with no observed dual-owner conflicts or capacity overshoots. At the production peak, CreateLive latency was 98.3 ms at P99 and PushText latency was 16.5 ms at P99. Across 100,000 randomized ownership runs, assignment-time, claim-only, and WorkerID-only protocols permit stale execution, while DHSched records no dual-owner or stale-generation release.

Entanglement embezzlement asks how much entanglement can be borrowed from a many-body state by local operations and classical communication while returning that state with a small error. We uncover a conservation-law mechanism that redistributes the dominant probability mass of the Schmidt spectrum in the logarithmic Schmidt-rank coordinate and thereby controls this operational resource. For typical random pure states at fixed U(1) charge, we prove a finite-error conversion law: away from half filling, the charge bias converts $O(\sqrt L)$ charge fluctuations into $O(\sqrt L)$ borrowable entanglement, whereas particle–hole symmetry at half filling removes this contribution entirely. We then show how the resource develops dynamically in charge-conserving random circuits. Combining hydrodynamic analysis with large-scale replica tensor-network calculations, we find that diffusion broadens the operationally relevant distribution in the logarithmic Schmidt-rank coordinate on the scale $t^{1/4}$ and increases the amount of entanglement that can be borrowed. Charge transport therefore continues to reorganize the entanglement spectrum and activate embezzlement after the leading volume-law entropy has saturated.

With the rapid pace of AI research and the hundreds of daily new publications, staying up-to-date with the latest developments has become increasingly difficult. For researchers, quickly identifying impactful work is essential, yet manually reviewing each new publication is impractical. Automated impact prediction methods help address this challenge, usually by combining various information sources available, such as a paper's content or citation history. In this work, we propose using GitHub engagement as an additional source and demonstrate that it provides both a timely and accurate signal. To this end, we introduce GitScholar, a novel dataset that links GitHub activity from 444,000 repositories to over 558,000 AI arXiv papers. Our experiments show that GitHub reactions improve early prediction precision by up to 12% over a strong academic baseline. Additionally, we find that GitHub signal offers near-complete coverage of high-impact AI papers, and consistently correlates with future academic success. GitScholar is publicly available at https://huggingface.co/datasets/huawei-csl/GitScholar.

Embodied Question Answering (EQA) requires an agent to explore a previously unseen environment, gather relevant information, and answer questions about the scene. Recent approaches leverage Vision-Language Models (VLMs) together with semantic maps or scene graphs to guide exploration. However, exploration is typically driven only by local observations, while structural priors about the environment remain largely unused. We propose HFLEX-EQA, a hierarchical EQA framework that combines online scene graph construction, VLM- based planning, semantic frontier exploration, and floorplan priors. The system incrementally builds a hierarchical scene graph and an open-vocabulary occupancy map from RGB-D observations, enabling a VLM to jointly reason over the scene graph, task-relevant visual observations, exploration history, and an estimated topological floorplan. Furthermore, we introduce a room-discovery strategy that leverages the floorplan and open-vocabulary frontier semantics to guide exploration toward semantically relevant yet currently unobserved room types. We evaluate HFLEX-EQA on the OpenEQA and ExploreEQA benchmarks and demonstrate deployment on a quadruped robot in real indoor environments. Our results demonstrate the benefit of combining VLM-based hierarchical planning with structural floorplan priors for the EQA task.

Tensor-based modulation (TBM) schemes are a promising approach for unsourced random access (URA), where user separation relies on decomposing the received signal tensor via the canonical polyadic decomposition (CPD), typically computed with alternating least squares (ALS). Standard ALS, however, treats the factor matrices as unstructured and fails to exploit the discrete structure of the tensor sub-constellations. We propose DVB-ALS, a discrete variational Bayesian CPD framework with specific priors, tailored to a tensor structure with the corresponding encoding strategy, combined with iterative computation of an approximate posterior distribution. A discrete Gaussian mixture prior on one Grassmannian factor softly aligns the estimates toward the constellation points. The remaining factors are jointly modeled with a structured Gaussian prior whose posterior mean is constrained to the Khatri-Rao product manifold and posterior variance upper-bounded to prevent norm divergence during inference. The resulting closed-form coordinate ascent algorithm jointly estimates all latent factors and their uncertainties. We integrate DVB-ALS into DVB-TBM to design a complete URA receiver with single-user demapping, polar decoding with cyclic redundancy check (CRC) verification, and successive interference cancellation (SIC). Simulation results show significant gains over standard ALS-based decomposition and robust detection performance in URA settings, outperforming state-of-the-art schemes under high system loads.

The Terrapin attack against SSH channel integrity (USENIX Security 2024) used a novel attack vector: attacks on the channel state. Surprisingly, not all AEAD modes of SSH were equally affected by this attack, and it remained an open question if "unaffected" meant "secure". Existing formal models for secure channels are based on stateful encryption. However, these models do not define what the channel state is and how it is used as input to the different AEAD modes. In this paper, we propose a formal model for channel integrity under partially chosen state. Applied to the Terrapin attack, the chosen state is the SSH sequence number. It uses an abstract stateful encryption interface, for which we provide pseudocode descriptions for the eight most prominent AEAD modes used in SSH. By varying the SND oracle, we can model ciphertext-only (CO; the Terrapin attack), known-plaintext (KPA), and chosen-plaintext (CPA) attacks. This allows us to establish concrete bounds on the security of the AEAD modes. We find that all three Encrypt-then-MAC (EtM) modes and ChaCha20-Poly1305 in SSH are insecure in the CO model. AES-GCM is the only cipher secure in all three model variants. Going beyond Terrapin, we show that Encrypt-and-MAC (EaM) with a CBC cipher is secure, even in the KPA model. In particular, we describe a novel BEAST-like chosen-plaintext attack on the channel integrity of EaM-CBC, which separates the KPA and CPA models for this scheme.

For any complete noncompact manifolds $(M^n, g)$ of nonnegative sectional curvature, with $3\le n\le6$, we obtain $\displaystyle \lim_{r\to\infty}r^{2-n}\int_{B_p(r)}\Sc_g\,dV_g\le8\pi\omega_{n-2}(1-\AVR(M^n,g))$ and the corresponding rigidity. The latter bound and its equality classification hold in $(M^n, g)$ with positive-dimensional souls. With a pole, $\displaystyle \lim_{r\to\infty}r^{2-n}\int_{B_p(r)}\Sc_g\,dV_g\le4\pi\omega_{n-2}(1-\AVR(M^n,g))$ follows from an independent distance-sphere proof. We also prove the sharp total scalar-integral bound for closed manifolds whose c

We study the thermal equilibrium of a classical system of identical point charges moving between two walls with parallel surfaces at a distance $d$, charged symmetrically with position-dependent surface charge densities. Specifically, we study the effect of surface charge modulation on the effective force (pressure) between the walls in the Poisson-Boltzmann limit. Based on Monte Carlo simulations at high temperatures, it is predicted that surface charge modulation reduces the pressure between the walls compared to uniformly charged wall surfaces with the same average surface charge density. We restrict ourselves to surface charge modulations in only one direction and, using a general solution of the two-dimensional Liouville equation, we construct exact solutions for the electrostatic potential. The surface charge density on the walls is generated inversely from this potential, which means that our exact results apply to a limited set of models with a specific variation of the surface charge density with distance $d$. Explicit analytical results show that surface charge modulation can both increase (small distances $d$) and decrease (large $d$) the pressure between the walls.

Reinforcement learning has become a central component of large language model (LLM) post-training, yet token-level credit lacks a generally accepted mathematical definition, leaving its relationship to commonly used training signals unclear. We formulate three regularity conditions, namely Completeness, Prefix Consistency, and Neutrality, and prove that they uniquely determine token-level credit. This characterization provides a unified basis for explaining phenomena across existing algorithms and guides the development of an improved actor-critic training procedure. Through this lens, an ideal teacher in On-Policy Distillation (OPD) acts as an implicit critic, yielding an expected policy gradient proportional to that induced by token-level credit. Response-level REINFORCE Leave-One-Out (RLOO) signals match the expected policy-gradient contribution of token-level credit despite their coarser granularity. We further establish approximate credit sparsity under bounded outcome rewards and show how intermediate critic errors in Generalized Advantage Estimation (GAE) can become comparable to the underlying credit. These motivate Policy Aligned Critic Training (PACT), which adopts an Actor-then-Critic update order to apply importance sampling correction to critic training and better align the critic with the updated policy. In agentic mathematical reasoning, PACT achieves 72.87% average accuracy across four benchmarks, outperforming GRPO and PPO by 8.80 and 13.16 percentage points, respectively. On SWE-bench Verified, PACT achieves a pass rate of 67.4%, outperforming PPO, GRPO, and SAO by 2.4, 2.0, and 3.8 percentage points, respectively.