arXiv:2603.15245v2 Announce Type: replace
Abstract: Empathy is central to human connection, yet people often struggle to express it effectively. In blinded evaluations, large language models (LLMs) generate responses that are often judged more empathic than human-written ones. Yet when a response is attributed to AI, recipients feel less heard than when comparable responses are attributed to a human. We built a conversation platform in which participants are asked to offer empathic support to an LLM expressing realistic troubles and conducted a randomized experiment collecting 33,938 messages spanning 2,904 text-based conversations between 968 participants and their LLM conversational partners. We find participants report feeling empathy but systematically fail to express it, but an LLM coaching intervention offering personalized feedback on effective empathic communication significantly boosts it without homogenizing participants' responses. Moreover, we derive a data-driven taxonomy of idiomatic empathic expressions in naturalistic dialogues across personal and workplace trouble scenarios. These results advance the scientific understanding of how empathy is expressed and demonstrate a scalable, AI-based intervention for scaffolding and cultivating it.
Science Journals
arXiv:2603.16446v4 Announce Type: replace
Abstract: When capturing images through glass surfaces or windshields on rainy days, raindrops and reflections frequently co-occur to significantly reduce the visibility of captured images. This practical problem lacks attention and needs to be resolved urgently. Prior de-raindrop, de-reflection, and all-in-one models have failed to address this composite degradation. To this end, we first formally define the unified removal of raindrops and reflections (UR$^3$) task for the first time and construct a real-shot dataset, namely RainDrop and ReFlection (RDRF), which provides a new benchmark with substantial, high-quality, diverse image pairs. Then, we propose a novel diffusion-based framework (i.e., DiffUR$^3$) with several target designs to address this challenging task. By leveraging the powerful generative prior, DiffUR$^3$ successfully removes both types of degradations. Extensive experiments demonstrate that our method achieves state-of-the-art performance on our benchmark and on challenging in-the-wild images.
arXiv:2603.23283v3 Announce Type: replace
Abstract: The uniform electron gas underlies the local-density approximation of density-functional theory, yet correlation contributions beyond the random-phase approximation (RPA) are known mainly through high-dimensional numerical evaluation, not in closed form. We study the screened second-order exchange (``kite'') diagram, in which one interaction line carries a frequency-dependent screened interaction. For a single-pole reference model with a momentum-independent screening scale, we perform the frequency and loop integrals analytically and reduce the diagram to a one-dimensional integral, whose static limit reproduces in closed form the exact Onsager-Mittag-Stephen second-order exchange, fixing the absolute energy scale with no free parameter. A Mellin-Barnes analysis with rigorous remainder estimates gives the behaviour in both screening limits. Using the linearity of the reduced functional in the screened line, we replace the bare interaction by the physical static RPA (Lindhard) screening, so that the density enters only through the Thomas-Fermi scale rather than an assumed map, and its dependence is set by the endpoints of a single geometric kernel with exponents fixed by the diagram. We prove that this kernel is even and that its quadratic coefficient vanishes identically, so the high-density expansion contains no half-integer power; this justifies the integer-power-times-logarithm form used in recent numerical work and identifies a half-integer term there as an interpolation artifact. The construction extends to arbitrary spin polarization, the bare diagram being polarization-independent, and a dynamical adiabatic-connection evaluation normalized only to that limit reproduces the numerically evaluated kite for both the unpolarized and fully polarized gas, including the low-density sign change. The result is a controllable analytic reference for screened exchange beyond the RPA.
arXiv:2603.23800v2 Announce Type: replace
Abstract: We present a novel LLM-informed model-based planning framework, and a novel prompt selection method, for object search in partially-known environments. Our approach uses an LLM to estimate statistics about the likelihood of finding the target object when searching various locations throughout the scene that, combined with travel costs extracted from the environment map, are used to instantiate a model, thus using the LLM to inform planning and achieve effective search performance. Moreover, the abstraction upon which our approach relies is amenable to deployment-time model selection via the recent offline replay approach, an insight we leverage to enable fast prompt and LLM selection during deployment. Simulation experiments demonstrate that our LLM-informed model-based planning approach outperforms the baseline planning strategy that fully relies on LLM and optimistic strategy with as much as 11.8% and 39.2% improvements respectively, and our bandit-like selection approach enables quick selection of best prompts and LLMs resulting in 6.5% lower average cost and 33.8% lower average cumulative regret over baseline UCB bandit selection. Real-robot experiments in an apartment demonstrate similar improvements and so further validate our approach.
arXiv:2603.29693v3 Announce Type: replace
Abstract: A robust decision-making process must take into account uncertainty, especially when the choice involves inherent risks. Because artificial intelligence (AI) systems are increasingly integrated into decision-making workflows, managing uncertainty relies more and more on the metacognitive capabilities of these systems; i.e, their ability to assess the reliability of and regulate their own decisions. Hence, it is crucial to employ robust methods to measure the metacognitive abilities of AI. This paper is primarily a methodological contribution arguing for the adoption of the meta-d' framework as the gold standard for assessing the metacognitive sensitivity of AIs--the ability to generate confidence ratings that distinguish correct from incorrect responses. Moreover, we propose to leverage signal detection theory (SDT) to measure the ability of AIs to spontaneously regulate their decisions based on uncertainty and risk. To demonstrate the practical utility of these psychophysical frameworks, we conduct two series of experiments on three large language models (LLMs)--GPT-5, DeepSeek-V3.2-Exp, and Mistral-Medium-2508.
arXiv:2604.00533v2 Announce Type: replace
Abstract: Large Language Models (LLMs) generalize across tasks through reusable representations and flexible reasoning, yet remain brittle in real deployment when faced with evolving tasks and continual distribution shift. While test-time adaptation addresses this by updating models with unsupervised objectives on test data, prevailing methods are fundamentally limited by their neglect of source knowledge preservation and adaptation signal reliability. Inspired by how Drosophila orchestrates memory update by balancing retroactive and proactive interference via Rac1 and MAPK pathways, we design Synapse Consolidation (SyCo) with two core components: a Rac1-inspired plasticity confiner and a MAPK-inspired update controller. The former dynamically confines plasticity to a tail-gradient subspace that is less critical for source knowledge, enabling rapid specialization while preserving source representations. The latter uses a tiered controller to suppress noisy updates and consolidate useful adaptations under non-stationary streams. To further model real deployments with multiple sources and continually emerging tasks, we introduce Multi-source Open-set Adaptation (MOA) setting, where a model is trained on multiple labeled source tasks and then adapts on open, non-stationary unlabeled test streams mixing seen and unseen tasks with partial overlap in label and intent space. Across 18 NLP datasets under the MOA setting, SyCo consistently outperforms strong baselines, achieving 78.31\% on unseen-task adaptation and 85.37\% versus unseen-data shifts, setting a new state-of-the-art.
arXiv:2604.01577v3 Announce Type: replace
Abstract: We study out of distribution generalization in streaming tasks where models are trained on short sequences but must operate over much longer, unknown horizons under bounded memory. Our focus is on a persistent fast slow recurrent formulation in which a latent state is maintained across observations rather than reset at each stream step. For each incoming observation, the model performs multiple weight-shared latent updates with a recurrent core and then carries the resulting state forward to the next observation. This allows the model to maintain and refine a compact stream-level state without reprocessing a growing context. We evaluate this formulation across symbolic sequence prediction, supervised navigation, and partially observable reinforcement learning tasks. Across these settings, persistent latent recurrence improves OOD generalization over recurrent, state-space, and Transformer baselines. Through recurrent-core ablations, we identify architectural ingredients that are consistently associated with strong OOD performance, including state-dependent transitions and feature-wise nonlinear mixing. Together, these results highlight the value of revisiting persistent recurrence as an architectural bias for more generalizable sequence prediction.
arXiv:2604.05488v2 Announce Type: replace
Abstract: Disruptions are a major concern for future fusion reactors based on the tokamak principle. To ensure machine protection, the thermal loads and vessel forces that arise during disruptions have to be mitigated reliably. For the ITER disruption mitigation system (DMS), the shattered pellet injection (SPI) technology has been selected. It can provide a prompt delivery of the injection material into the plasma core, with the mitigation efficiency depending on fragment size and velocity. A highly flexible SPI system was built and installed at the tokamak ASDEX Upgrade (AUG) to aid the finalization process of the ITER DMS and provide crucial input for modeling. The SPI-induced disruptions in the 2022 AUG experiments follow a typical chain of events, which are discussed in this paper: The first light, main fragment arrival, plasma movement event, MARFE, thermal quench/plasma current spike, current quench, and vertical displacement event phase. Depending on the injection parameters, these phases may vary significantly or some might not be present at all. In this paper, we will focus on the characterization of these disruption phases and figures of merit for the mitigation efficiency, depending on the SPI configuration. With increasing amount of assimilated neon in the plasma - primarily influenced by the neon content in the pellet but also the shattering parameters - the disruptions exhibit different behaviors. This disruption evolution seems to be a continuous process, with the most prominent feature being the changing disruption time scales and plasma current time trace shape during the CQ from convex (poorly or unmitigated) $\rightarrow$ concave (well mitigated/radiation dominated). Depending on the injection, pre-TQ durations between 15 - 0.5 ms and early CQ durations ($\Delta \textrm{t}_\textrm{CQ}^{100 \rightarrow 80}$) between 13.3 - 8.2 ms had been achieved at AUG.
arXiv:2604.07392v3 Announce Type: replace
Abstract: Autonomous agents operating in dynamic environments increasingly demand decision-making systems that are both efficient and interpretable. Hence we propose the Event-Retrieve-Action (ERA) framework, an alternative formulation for embodied decision-making that bridges the gap between black-box imitation and interpretable memory retrieval while enabling online refinement without retraining. The environment is represented as structured semantic events encoded into an interpretable latent representation, and decisions are generated by retrieving relevant prior experiences from a knowledge bank of event-action pairs. Final actions are produced through weighted aggregation of retrieved maneuvers, enabling transparent and physically consistent decision-making. Experiments in UAV navigation demonstrate real-time performance and adaptive behavior in dynamic environments as a representative embodied decision-making application scenario.
arXiv:2604.09320v2 Announce Type: replace
Abstract: Mechanistic understanding and rational design of complex chemical systems depend on fast and accurate predictions of electronic structures beyond individual building blocks. However, if the system exceeds hundreds of atoms, first-principles quantum mechanical (QM) modeling becomes impractical. In this study, we developed FB-GNN-MBE by integrating a fragment-based graph neural network (FB-GNN) into the many-body expansion (MBE) theory and demonstrated its capacity to reproduce first-principles potential energy surfaces (PES) for hierarchically structured systems with manageable accuracy, complexity, and interpretability. Specifically, we divided the entire system into basic building blocks (fragments), evaluated their one-fragment energies using a QM model, and addressed many-fragment interactions using the structure-property relationships trained by FB-GNNs. Our investigation shows that FB-GNN-MBE achieves chemical accuracy in predicting two-body (2B) and three-body (3B) energies across water, phenol, and mixture benchmarks, as well as the one-dimensional dissociation curves of water and phenol dimers. To transfer the success of FB-GNN-MBE across various systems with minimal computational costs and data demands, we developed and validated a teacher-student learning protocol. A heavy-weight FB-GNN trained on a mixed-density water cluster ensemble (teacher) distills its learned knowledge and passes it to a light-weight GNN (student), which is later fine-tuned on a uniform-density (H2O)21 cluster ensemble. This transfer learning strategy resulted in efficient and accurate prediction of 2B and 3B energies for variously sized water clusters without retraining. Our transferable FB-GNN-MBE framework outperformed conventional non-FB-GNN-based models and provided a scalable and accurate route toward interaction energies of large molecular assemblies.
arXiv:2604.16449v2 Announce Type: replace
Abstract: Representing turbulent flow fields in a compact yet physically faithful form remains a central challenge in computational fluid dynamics. We propose a continuous parametric representation based on localized Gaussian primitives, in which the velocity field is modeled as a superposition of kernels with learnable positions, amplitudes, and scales. This formulation yields a compact, grid-independent encoding while enabling evaluation of derived quantities such as vorticity and enstrophy.
The approach is assessed on three-dimensional Taylor-Green vortex fields spanning stages from smooth flow to fully developed turbulence. We quantify the compression-accuracy trade-off using both primary variables and derivative-sensitive diagnostics. The baseline isotropic formulation achieves high velocity accuracy at compression ratios exceeding 1e3-1e4, but exhibits substantial enstrophy degradation due to loss of small-scale structure.
To address this limitation, we investigate structure-aware extensions including adaptive placement, multi-resolution kernels, and anisotropic Gaussians. The anisotropic formulation provides the most consistent improvement, better aligning with elongated vortical structures and recovering intermediate- and high-wavenumber content, while other strategies yield modest gains. A compact-support Beta basis improves enstrophy in some cases but introduces localized artifacts.
Overall, the results indicate that the main limitation of baseline Gaussian representations lies in geometric expressiveness rather than parameter count. The proposed framework provides a compact, interpretable, and continuous representation of turbulent flows, and establishes a foundation for structure-aware and physics-informed flow compression.
arXiv:2607.07460v1 Announce Type: new
Abstract: Navigation for social organisms rarely is a fully independent activity. Group structure and dynamics, as well as embodied interactions, critically influence useful behavior. Individual neural network controlled agents are trained to navigate in different social contexts, where social dependence and behavioral strategy learned is determined by relative task performance and spatial effect. Increasing high quality social information drives phase transitions from individual to following navigational strategy, and to collision avoidance in response to a crowded foraging patch. Predictable, nonstationary environmental dynamics drive behavioral hybridization between individual and social navigation, far and near the patch. Our findings challenge the approach of only inspecting individual behavior for social organisms and highlight the importance of taking a bottom-up approach in understanding how organisms behave.
arXiv:2607.07294v1 Announce Type: new
Abstract: Turn-taking prediction is a key requirement for social robots involved in human-human interaction, particularly in mediator settings, where the robot must anticipate conversational dynamics rather than merely react to pauses. This work presents a Multimodal Voice Activity Projection (MM-VAP) framework that extends the original audio-only VAP formulation to synchronized audio-visual inputs while preserving its self-supervised future-projection objective. The proposed approach builds on pretrained audio-visual backbones originally optimized for speech-related tasks and adapts them through Low-Rank Adaptation to the multimodal turn-taking problem. After independent speaker encoding, an inter-speaker attention stage models the relational dynamics required to project future voice activity. In addition, a semantic consistency loss is introduced to regularize the 256-state output space according to higher-level dialogue activity patterns. Experiments on NoXi and NoXi+J showed improvements over the current baselines, particularly for some turn-taking events. Additional evaluation on the Haru EDR corpus further supported the suitability of this direction for mediation-oriented human-robot interaction.
arXiv:2607.07471v1 Announce Type: new
Abstract: Machine learning models are increasingly deployed in high-stakes domains, raising concerns about both privacy and fairness. Differential Privacy (DP) has become a gold standard for privacy-preserving data analysis, while fairness-aware mechanisms aim to mitigate discrimination against underrepresented groups. However, these objectives can conflict: DP often amplifies disparities across demographic groups, and little is known about whether established fairness interventions remain effective under DP constraints. In this work, we present, to our knowledge, the first systematic evaluation of fairness interventions on differentially private synthetic tabular data. Our benchmark centers on the Adaptive Iterative Mechanism (AIM), identified as the state-of-the-art marginal-based DP synthesizer (Cormode et al. 2025). We thus evaluate fairness interventions across four datasets, multiple group fairness metrics, and three categories of mitigation strategies (pre-processing, in-processing, and post-processing) under a wide range of privacy budgets. We compare four pipeline configurations: (Baseline) training on original data; (DP-only) training on DP synthetic data; (Fair-only) applying fairness mechanisms on original data; and (DP+Fair) combining fairness mechanisms with DP synthetic data. Our results demonstrate that while DP alone can degrade both utility and fairness, applying fairness interventions can partially restore equitable outcomes. Among them, post-processing methods tend to provide more stable fairness-utility trade-offs across privacy budgets and synthesizers, achieving strong fairness improvements while preserving competitive utility relative to other intervention stages. We release all code, data, and experimental artifacts in an open-source repository to ensure full reproducibility and to support future research on the privacy-fairness-utility trade-off.
arXiv:2607.07463v1 Announce Type: new
Abstract: A perennial research question in visualization involves identifying which visual encodings for a particular dataset are most effective for users in performing a specific task. The relative effectiveness of the different encodings are commonly identified through controlled experiments. However, designing an experiment involves making many, often ad hoc, decisions about the experimental setup such as whether to include a training module, whether to provide performance-based incentives to participants, etc. Yet, there is limited guidance on how these decisions should be made, and we do not fully understand the impact of these subjective decisions on empirical results. In this paper, we investigate the impact of one such key design decision: monetary rewards. Specifically, we ask: does providing or not providing participants with performance-based financial incentives affect the results and the conclusions that we draw from visualization studies? We conducted two crowdsourced studies investigating the impact of incentives on (i) a low-level, perceptual task (perception of correlations in scatterplots or parallel coordinate plots), and (ii) a task involving reasoning (decision-making based on a weather forecast represented as intervals or density plots). In each of these studies, we manipulate both the visual representation and the presence of incentives as between-subject conditions. We expected to find no effect of incentives on the perceptual task, but to see an effect for the decision-making task. However, we found no effect on task performance in either study. While these are results of only two studies and should be replicated, they suggest that performance-based financial incentives may not always have the intended effect on participants that we presumed, and calls for a reflection of how incentivized studies should be designed.
arXiv:2607.07652v1 Announce Type: new
Abstract: Search engines have long allocated attention on the web by routing users from queries to websites. AI search changes this arrangement because information needs can be resolved inside the intermediary. Using URL-level Comscore U.S. desktop clickstream, we compare ChatGPT and Google information-seeking occasions and exploit ChatGPT Search access expansions to estimate traditional search displacement. ChatGPT produces outbound clicks in only 5.2% of conversation sessions, far below Google's referral ratio. The remaining clicks are not a scaled-down Google stream: they skew toward specialized destinations and away from ad-supported sites. Wider access cuts search use by 9.4%, with search-referral losses largest for informational categories. Our findings identify a central economic shift in digital intermediation: AI search might satisfy information needs inside the intermediary while weakening the referral bargain that has linked search, traffic, and content production on the open web.
arXiv:2607.07322v1 Announce Type: new
Abstract: Automated crowd counting in Hajj video is difficult not because current models lack capacity, but because the footage violates the assumptions those models were built on: cameras observe the crowd from steep, near-vertical angles, individuals occlude one another extensively, and a single frame can contain well over a thousand people. Benchmarks that test crowd counting in such an environment are either private or not detailed per second. We revisit the HAJJv2 dataset and contribute HAJJv2-CrowdCount: per-second human-annotated crowd counts for its testing videos. Using these annotations, we benchmark three recent zero-shot counting paradigms: an open-vocabulary detector (YOLO-World), a point-based counter (APGCC), and a promptable segmentation-based counter (SAM3Count). SAM3Count attains the lowest overall mean absolute error (MAE 70.4, 95% CI 56.0-86.1), ahead of YOLO-World (92.0) and APGCC (152.9). This ordering reverses, however, in the regime most relevant to deployment: on the densest frames, the detection- and segmentation-based counters both degrade sharply (MAE exceeding 300), while the point-based counter degrades far more gracefully (MAE 114.9). This inversion is decision-relevant for Hajj crowd management, where reliable counts are needed most precisely in the densest and most occluded scenes. The annotations are released to support reproduction and extension of these results.
arXiv:2607.07368v1 Announce Type: new
Abstract: AI control is a family of techniques to prevent an AI with malicious goals from subverting its operator's intent. AI Control usually studies a single agent in one trajectory, but real deployments run many agents over shared infrastructure, and the most severe risks (model-weight exfiltration, training-run poisoning) plausibly need several agents acting in concert. We initiate the empirical study of multi-agent AI control, formalising distributed attacks in which several agents jointly aim for a malicious goal. We develop FakeLab: a synthetic AI-lab codebase (9 services, 86 benign tasks, 4 attack objectives). We evaluate single agent monitoring against distributed attacks, varying the number of agents, their coordination, model capabilities and precise monitoring configuration.
Our central finding is the fragmentation effect: as more agents coordinate to attack, per-agent monitoring becomes less likely to catch any of the attackers. Secondly, this is not driven by the ratio of benign to malicious code -- we argue this is likely dependent on model capability. Third, an explicit planner amplifies the fragmentation effect, raises the rate of completing the attack up to sevenfold, and mildly sharpens executors' poor monitor-theory-of-mind. Finally: a weak "trusted" monitor misses most attacks, while a stronger one cuts undetected success by more than an order of magnitude but remains imperfect. We are happy to share FakeLab, our control setting, with safety researchers on request.
arXiv:2607.07498v1 Announce Type: new
Abstract: Testing is a major effort for the gaming industry, requiring a significant part of development budget and people power. We present a case study on a development version of the ice hockey game EA SPORTS NHL 26, for which human playtesters test the goalie AI for behavioral exploits. To reduce the effort of re-testing the goalie AI after every game or behavior modification in the development phase, we propose Reward-Adaptive Iterative Discovery (RAID), a novel approach to automatically find exploits using an iterative Reinforcement Learning (RL) approach that trains a population of goal scoring agents. While previous approaches can already successfully find exploits, RL algorithms tend to overfit to a single solution. We introduce a simple extension on top of existing RL algorithms, such that they find multiple diverse high-quality solutions. For our first deployment of this approach, within a single experiment we were able to find six hockey scoring exploit strategies that were qualitatively similar to those that playtesters had found in hours-long manual testing sessions.
arXiv:2604.17519v2 Announce Type: replace
Abstract: Quantum devices increasingly expose a fundamental gap between compiler-modeled noise and hardware execution. Today's compilers approximate noise as calibration-derived costs over gates, qubits, and coupling edges, yet semantically equivalent circuits with identical compiler-visible costs can still produce unexpected hardware noise on real devices. The missing factor is execution context: short ordered gate sequences on specific qubit neighborhoods can induce excess error beyond current quantum noise models. Isolating these execution-dependent abnormal patterns is naturally reminiscent of delta debugging, but direct reduction is unsound in this setting: the failure signal is stochastic and drifting across calibration windows.
To address this problem, we propose QRisk, a delta-debugging framework for isolating recurring abnormal gate patterns on quantum hardware. The key idea is to turn hardware-model discrepancy into a size-normalized stochastic fault signal: QRisk compares real-device error against calibration-based error from current quantum noise model, so reductions are judged by excess discrepancy rather than raw error decrease. It then performs segment-level reduction to localize compact gate fragments and validates candidates through independent rediscovery across calibration windows, retaining only recurring backend-specific gate patterns.
On three IBM Heron r2 backends across 30 qubit layouts, QRisk finds 25 recurring abnormal gate patterns. Controlled experiments show that eliminating these gate patterns in compiled circuits reduces excess hardware noise by 24% on ibm_fez (Spearman \r{ho} = 0.515, p = 0.0007) and by 45% on ibm_marrakesh (\r{ho} = 0.711, p < 0.0001).
arXiv:2604.18360v3 Announce Type: replace
Abstract: Audio-text retrieval systems based on Contrastive Language-Audio Pretraining (CLAP) achieve strong performance on traditional benchmarks; however, these benchmarks rely on caption-style queries that differ substantially from real-world search behavior, limiting their assessment of practical retrieval robustness. We present Omni-Embed-Audio (OEA), a retrieval-oriented encoder leveraging multimodal LLMs with native audio understanding. To systematically evaluate robustness beyond caption-style queries, we introduce User-Intent Queries (UIQs) - five formulations reflecting natural search behaviors: questions, commands, keyword tags, paraphrases, and exclusion-based negative queries. For negative queries, we develop a hard negative mining pipeline and propose discrimination metrics (HNSR, TFR) assessing models' ability to suppress acoustically similar distractors. Experiments on AudioCaps, Clotho, and MECAT show that OEA achieves comparable text-to-audio retrieval performance to state-of-the-art M2D-CLAP, while demonstrating clear advantages in two critical areas: (1) dominant text-to-text retrieval (+22% relative improvement), and (2) substantially superior hard negative discrimination (+4.3%p HNSR@10, +34.7% relative TFR@10), revealing that LLM backbones provide superior semantic understanding of complex queries.
arXiv:2604.24080v3 Announce Type: replace
Abstract: A compressed self-index stores a string in compressed form while supporting locate queries without decompression. For highly repetitive strings, such as those arising in web crawls, versioned documents, and genomic collections, static self-indexes can match the $\delta$-optimal lower bound of $\Omega(\delta \log(n \log \sigma / (\delta \log n)) \log n)$ bits up to constant factors, where $n$ is the string length, $\sigma$ is the alphabet size, and $\delta$ is the substring complexity. Their dynamic counterparts, however, remain scarce: every existing dynamic self-index either fails to attain $\delta$-optimal space, pays $\Omega(\log n)$ time per reported occurrence for locate queries, or has an update time that grows with the maximum value in the longest common prefix (LCP) array of the text. We present the dynamic RR-index, a dynamic grammar-compressed self-index built on the restricted recompression run-length straight-line program (RLSLP). To our knowledge, it is the first dynamic self-index to attain $\delta$-optimal space. The index uses $O(\delta \log(n \log \sigma / (\delta \log n)) \log n)$ bits in expectation, answers locate queries in expected $O(m + \log m \log^{2} n + \mathit{occ} (\log n / \log \log n))$ time, where $m$ is the pattern length and $\mathit{occ}$ is the number of occurrences, and supports insertion of a length-$m'$ string and deletion of a length-$m'$ substring in expected amortized $O(m' \log^{2} n + \log^{3} n)$ time, with no dependence on the maximum LCP value. On eleven highly repetitive corpora, including a $37$ GB Wikipedia dump and a $59$ GB human-chromosome collection, the dynamic RR-index is up to $77\times$ faster than the dynamic r-index for updates and up to $11\times$ faster than other dynamic indexes for locate queries.
arXiv:2607.07608v1 Announce Type: new
Abstract: Mainstream Vision-Language-Action (VLA) models predict actions primarily from the current observation under a Markovian assumption, thus struggling with long-horizon, temporally dependent tasks. Existing memory-augmented VLAs either expand the observation window or retrieve history from the memory bank as auxiliary policy-side context. However, they leave memory outside the native latent embedding space of VLA reasoning, preventing historical experience from being fluidly interleaved with multimodal reasoning and action formation. To this end, we introduce LaMem-VLA, a latent-memory-native framework that reconstructs historical experience into latent memory tokens and directly interweaves them with VLA reasoning. At its core, LaMem-VLA introduces four coordinated components: (i) a curator that organizes historical experience into two complementary short-term and long-term memory vaults; (ii) a seeker that queries both vaults using the multimodal cognition to retrieve context-relevant evidence; (iii) a condenser that reconstructs the retrieved evidence into compact short-term and long-term latent memory tokens; and (iv) a weaver that injects these memory tokens with the current observation and instruction into one continuous embedding sequence. By representing, retrieving, and consuming historical experience entirely in the same continuous latent space, LaMem-VLA enables memory to directly participate in VLA reasoning and guide action generation under a bounded context. Extensive experiments on SimplerEnv and LIBERO demonstrate the superiority of our LaMem-VLA.
arXiv:2607.07611v1 Announce Type: new
Abstract: Background: Graph neural networks improve computational prediction of polypharmacy side effects, but standard binary cross-entropy training allocates equal capacity to well-classified and difficult examples, potentially missing clinically significant interactions. We evaluated whether an asymmetric focal objective could improve multi-relational drug-drug interaction (DDI) prediction by emphasizing difficult positive interactions. Methods: ClinicalFocal loss was integrated into a relation-aware graph convolutional network using molecular fingerprints, physicochemical descriptors, and learned embeddings. The model was evaluated on TWOSIDES using five-fold cross-validation with identical experimental conditions (architecture, features, data partitions, hyperparameters, and random seeds) for ClinicalFocal loss and binary cross-entropy baseline. Results: ClinicalFocal loss increased accuracy from 0.699 to 0.892 (+19.3 percentage points) and F1 score from 0.700 to 0.894 (+19.4 percentage points). AUROC increased from 0.766 to 0.914, and AUCPR increased from 0.714 to 0.860. The false-negative rate decreased from 29.8% to 9.1%, while specificity increased from 69.6% to 87.5%. Overall classification error decreased from 30.1% to 10.8%, corresponding to a 64.1% relative reduction. Improvements were consistent across all five folds. Conclusions: Asymmetric focal optimization improved classification and ranking performance while achieving 90.9% recall for observed interaction triples, without modifying the underlying architecture. Loss-function design is a direct, tunable lever for improving graph-based DDI prediction.
arXiv:2607.07308v1 Announce Type: cross
Abstract: Friedman and Wigderson (1995) introduced a notion of second eigenvalue for hypergraphs that generalizes the second eigenvalue of the adjacency matrix of a graph. We show that $r$-uniform Erd\H{o}s-R\'enyi hypergraphs on $n$ vertices exhibit a spectral gap as soon as their expected number of hyperedges $m$ satisfies $m \gg n^{r/2}$. Prior work identified this scale only up to logarithmic factors; removing these factors is the main technical challenge.
Our proof overcomes this obstacle through an explicit decomposition of an associated selector process, inspired by a generic decomposition theorem of Talagrand (2021). As a consequence of our techniques, we obtain improved norm bounds for sparse random tensors with independent entries. Finally, under a mild moment equivalence assumption, we extend to tensors a seminal result of Seginer (2000) for random matrices with i.i.d. entries.