Papers

Systolic arrays are widely used to accelerate neural-network inference, yet fixed two-dimensional arrays struggle with the substantial diversity in both matrix geometry and available parallelism across CNNs and Transformer-based large language models. This mismatch is particularly severe for skinny matrices, small-batch autoregressive decoding, and operators containing multiple independent partitions, leading to low processing-element (PE) utilization and repeated spatial folding. Existing reconfigurable arrays typically improve either shape adaptation or distributed execution, whereas supporting both through fine-grained routing incurs considerable interconnect, buffering, and control overheads. We present LizTail (named after lizard tail autotomy, where predefined fracture planes allow segments to detach and remain independently active, analogous to LizTail's predefined reconfiguration boundaries), a multi-mode systolic array that enables constrained segment-level reconfiguration by partitioning a $128\times128$ array into 32 regular $128\times4$ segments and adding connectivity only at predefined endpoints. The resulting substrate supports trimming, long-path extension through $180^\circ$ turn-back connections, and distributed execution in which each subarray can itself be trimmed or extended. Differentiated PE datapaths and segment-aware CEB/SIB buffers provide the required compute and data-delivery flexibility without unrestricted PE-level routing. A hardware-constrained mapper selects the execution organization, dataflow, task partition, and tiling. Across representative workloads, LizTail achieves geometric-mean speedups of $5.02\times$ on CNN inference and $25.67\times$ on LLM decoding over a TPU-like array, while reducing EDP by up to $38.23\times$. LizTail is open-sourced at https://github.com/kmjbwxa/LizTail.

Conflict-graph search can stagnate despite effective operators when interventions fail to address the conflicts blocking progress. A disruptive intervention may also require further search before its value emerges. We propose CIPHEUR, a large language model (LLM)-guided framework that connects conflict evidence to coordinated intervention and recovery through Structured Intervention Units (SIUs). The shared control interface supports offline synthesis of executable policies and online contextual decisions. Asynchronous online control sustains native search during inference, while all policies preserve the original objective and feasibility constraints. Across eight satellite-ground scheduling conflict views with five runs per view and a 900-second evaluation cap, CIPHEUR-online achieves the highest mean objective among 11 configurations and increases mean scheduled contact duration by 14.84% over the greedy baseline GWMIN. It exceeds the strongest external per-view mean in seven of eight views and outperforms CHILS in all 40 paired runs. Within the same runs, its normalized objective improvement from 360 to 900 seconds is 1.50 to 1.86 times that of the two programmatic controllers. A separately frozen, one-shot LLM-generated controller also improves the mean objective over CHILS without runtime model calls. Together, these results demonstrate how structural search control translates LLM design knowledge and contextual reasoning into improved solution quality and sustained search progress.

We identify Gaussian curvature as a local mechanism for cubic cancellation of Laplace eigenfunctions on Riemannian surfaces, without arithmetic or dynamical assumptions. For $L^2$-normalized local solutions at frequency $\lambda$, cubic integrals against smooth compactly supported weights tend to zero whenever the curvature is nonzero almost everywhere where the weight is nonzero. Finite-order nondegeneracy yields sharp polynomial decay rates. In particular, on closed surfaces with nowhere-vanishing curvature, we obtain the rate $O(\lambda^{-1/6})$, attained by spherical harmonics on the round sphere. This proves the cubic case of the Gaussian moment conjecture for all normalized real eigenfunctions on closed negatively curved surfaces. The local proof rests on an exact identity relating the reduced phase at a Fermat point to Gaussian curvature. On complete hyperbolic surfaces with positive systole, we also prove global $O((\log\lambda)^{-3/2})$ decay for $L^2$-normalized eigenfunctions and smooth weights with bounded covariant derivatives through order two, without assuming finite area.

The nearly pure TeV-scale Higgsino dark matter provides a natural explanation for the high-energy nuclear-recoil excess reported by LUX-ZEPLIN. Such a heavy and compressed Higgsino spectrum makes direct searches at the LHC challenging. We therefore investigate complementary probes through the production and decay of heavier sparticles. We perform a comprehensive analysis and obtain the following observations: (i) For a TeV Higgsino, the current LHC Run-2 searches exclude gluino up to approximately 2.2~TeV, while the HL-LHC can give an exclusion reach of 2.5–2.6~TeV and a discovery reach of 2.3–2.4~TeV; (ii) The light-flavor squark coannihilation window lies beyond the Run-2 reach, but its lower end could be excluded at the HL-LHC under optimistic assumptions; (iii) The Run-2 search does not exclude the stop or sbottom benchmarks; (iv) The slepton and heavier-electroweakino productions are too small to access at the HL-LHC.

Interactive-agent evaluations commonly rely on standardized LLM-based user simulators that exhibit broadly cooperative behavior, potentially overlooking failures that arise under diverse but realistic user interactions. This can lead to overly optimistic estimates of agent robustness and provide limited failure signals for training agents that generalize across users. Existing approaches largely rely on predefined user behaviors, which may miss agent- or domain-specific weaknesses, while failure-oriented search can reduce task success by invalidating the underlying task rather than exposing genuine robustness failures. To fill this gap, we introduce TIPER, a task-integrity-preserving persona evolution method that automatically discovers agent-specific user policies exposing robustness weaknesses while keeping the underlying task valid and solvable. TIPER evolves persistent user personas through reflection-guided behavioral mutations and task-wise Pareto selection under a task-integrity constraint. Across four interactive-agent benchmarks, TIPER yields lower held-out avg@3 scores in all 12 GPT-5.4 settings, with observed reductions of 5.8 to 33.7 percentage points, and in all 28 agent-domain pairs spanning seven agents, with a median observed reduction of 22.9 points. Further analysis yields a behavior-to-failure taxonomy that connects discovered user behaviors to recurring agent failure modes. Together, these results show that TIPER enables more diagnostic robustness evaluation of interactive agents and provides informative failure analysis for future agent robustness improvement. Code repository: https://github.com/UCSB-NLP-Chang/TIPER.

We determine the Poisson algebras generated by partial Casimirs of undeformed and non-standard $q$-deformed $\mathfrak{sl}(2)$ coalgebras, respectively. The derivation is performed on the abstract algebraic level and does not depend on any particular symplectic realization. For the 3- and 4-site cases the constructions associated with the undeformed $\mathfrak{sl}(2)$ produce the rank-one and rank-two Racah Poisson algebras, respectively, while those with the non-standard deformed $\mathfrak{sl}$ yield finite quadratic Poisson algebras generated by connected partial Casimirs. At both the three and four sites, the final closure coefficients do not explicitly depend on the deformation parameter. The results hold at the Poisson coalgebra level and hence apply to any realizations which preserve the defining brackets and coproducts. Canonical symplectic realizations give rise to Hamiltonians in spaces of non-constant curvature. For the Stäckel $q$-deformed Smorodinsky-Winternitz model we show explicitly how the universal coalgebra sector is completed to maximal superintegrability. An exact two-block identity, proved directly in the localized Poisson algebra without appealing to a symplectic realization, controls the additional Stäckel integrals. In three dimensions, it generates a quadratic Poisson block with its own Casimir relation; in four dimensions, its coupling to the full left/right coalgebra family is organized into three nested quadratic blocks, and the same identity extends this nested coupling to arbitrary $N$. Explicit higher-site partial Casimirs and factorized bracket generators are collected in the appendices.

Label noise is prevalent in medical image datasets, arising from observer disagreement and ambiguous cases and deep networks memorize the mislabeled examples at the cost of generalization. Current solutions such as loss correction and sample selection add training complexity and generally assume the noise rate is known or estimable. This paper investigates whether robustness can be obtained architecturally, by routing a frozen backbone's features through a low-dimensional bounded bottleneck and classifying them with a head whose function class is intrinsically smooth. Variational quantum circuits with angle encoding and a linear cost function realise exactly such a class. Their use in medical imaging, however, has been limited by the small number of qubits available on noisy intermediate-scale devices, which typically restricts inputs to low-resolution grayscale images. We therefore couple a frozen ImageNet-pretrained ViT-B/16 backbone to a trainable linear bottleneck and an eight-qubit variational quantum circuit, allowing 224x224 RGB inputs to be processed. Under clean labels the model leads five of six MedMNISTv2 benchmarks against published hybrid quantum-classical models, with margins of 17.4 and 15.7 percentages on BloodMNIST and PathMNIST. The advantage emerges under corrupted labels. At symmetric noise rates of 30% and 50% the model outperforms many published quantum and noise-robust classical baselines on all six datasets without any explicit noise-handling component. Against a control that shares the backbone, bottleneck, parameter count and optimizer and replaces only the circuit with a ReLU layer, it retains 89.75% accuracy on PathMNIST at 50% noise where the control falls to 78.96%. Direct measurement of training dynamics shows that this advantage is not an inability to fit corrupted labels, which the circuit retains, but greater stability under prolonged training on them.

Although post-training significantly enhances model performance, the use of massive undisclosed data raises concerns about benchmark contamination and copyright infringement. This drives growing interest in training data detection for post-trained models. Existing evaluations for training data detection often leverage benchmarks constructed post hoc, which can introduce distribution shifts. Alternatively, controlled evaluations rely on small-scale training setups that diverge from realistic post-training settings. As a result, their reported performance may not faithfully reflect the effectiveness of detection methods in real-world scenarios. To this end, we introduce \method, a reliable benchmark designed to evaluate detection methods in realistic, large-scale post-training settings. Built on a multi-stage pipeline, \method\space encompasses stage-specific models throughout the post-training lifecycle, spanning SFT, DPO, and RL. To perform a reliable evaluation, we construct model-specific evaluation sets across four diverse tasks, featuring distribution-matched members and non-members with verifiable membership labels. Our results reveal that even the best-performing method achieves an average AUC of only 0.529 on our benchmark, which approaches random-guessing level (0.5). Furthermore, our analysis suggests that evaluation protocols can significantly affect measured detection performance. These findings highlight the significant potential for developing effective, practical, and robust detection methods under realistic scenarios.

A path indistinguishable from its own reverse carries no arrow of time, so time irreversibility is a necessary condition for a directed mechanism to be readable from trajectories alone. In finance it has been measured only on scalar series such as returns and volatility. We define the irreversibility rate of a coarse-grained limit order book path as the forward-to-reverse log-density ratio per event, and estimate it from three weeks of Nasdaq market-by-order data with an order-invariant MLP and a temporal convolutional network that receive identical blocks. For NVDA, the order-invariant and order-sensitive estimates are 0.036 and 0.054 nats per event, respectively. Their difference, the ordering increment, is 0.018. A surrogate preserving the coarse one-step transition table recovers 30 percent less irreversibility than the observed paths. The increment is positive for all twelve symbols on every held-out daytime date, and for NVDA in each session window. It changes more than tenfold with the coarse-graining scale. Neither a linear score nor any single feature group reproduces the full fitted rate. A queue-reactive simulator calibrated on the same days matches the market's ordered rate, 0.052 against 0.054, yet its ordering increment is 0.0006 against 0.018, building temporal asymmetry out of block composition instead of block order. We propose the ordering and scale profile of path irreversibility as a benchmark for order book simulators. Code and audit materials are available at https://github.com/hjyeo30/lob-irreversibility.

Predicting quantum-photonic circuit performance requires retaining source and device physics without making photon-counting calculations prohibitively expensive. Here we present PhoQuS, a framework that combines frequency-resolved source models, electromagnetic compact models and spatially correlated fabrication variation to simulate quantum photonic circuits. Covariance-Gram reduction after full-resolution Gaussian propagation lowers readout cost while retaining dominant spectral correlations. For continuous-variable state preparation, spectral reduction accelerates photon counting by 60.4-fold in an approximate Gottesman-Kitaev-Preskill circuit. Frequency-resolved modeling guides the use of cascaded squeezing with temporal walk-off compensation, yielding 0.99 fidelity to the single-mode circuit reference and 0.98 conditional purity in lossless simulations. For discrete-variable state preparation, simulations of three-photon Greenberger-Horne-Zeilinger circuits across a virtual wafer show that component robustness alone does not predict circuit fidelity, which depends jointly on source spectra, circuit architecture and control. PhoQuS connects device physics to protocol-level performance through a unified workflow that retains spectral detail while reducing readout cost.

Learning equilibria from data is difficult when several decision makers interact indefinitely and the system state has no prescribed bounds. We examine discounted general-sum stochastic games with compact action sets and an unknown nonlinear drift observed through Gaussian transition noise. The proposed procedure uses a single uninterrupted trajectory to update Gaussian-process models of the drift and to solve successively refined finite games. We prove a perturbation result showing that stability of equilibrium selection, together with uniform KL accuracy of Gaussian transition models, is sufficient for regional uniform convergence of the induced values and approximate Nash behavior in the original game. The region specifies where the guarantees are evaluated; strategies and dynamics continue to act on the entire Euclidean state space. For GP estimators, simultaneous confidence bounds yield a rate controlled by posterior uncertainty, kernel information gain, and joint-state dimension. Numerical studies illustrate how kernel choice and dimension appear in the predicted convergence behavior.

A ghost-cell immersed boundary method is proposed for the two-dimensional shallow water equations with a variable bed (topography), solved with an explicit Godunov-type finite volume scheme (unsplit second-order MUSCL, HLLC flux, hydrostatic reconstruction) on Cartesian grids. Ghost-cell methods leak mass through the wall. To find a robust and conservative treatment, eight constructions of the ghost values and a cut-cell form of the continuity equation were applied to seven test problems, which include wetting and drying on sloping beds and a laboratory dam break. In the resulting scheme, the ghost values are interpolated bilinearly at the image point; solid and dry nodes are excluded, and the weights of the remaining nodes are renormalized. Mass is conserved by solving the continuity equation in cut-cell form, while the momentum equations are left unchanged. The proposed scheme is about as accurate as the ghost-cell method without this modification and preserves a lake at rest whose shoreline crosses an immersed wall. Because only the continuity equation is modified, it avoids the treatment of the momentum equations on cut cells (wall pressure, cell merging) that a full cut-cell method requires.

Dynamic sparse training (DST) aims to reduce training costs while maintaining accuracy. Existing DST methods mask out parameters but keep matrix dimensions unchanged. Training speedups therefore depend on specialized support, such as sparse kernels. We introduce BARTER, a structured DST framework that reduces training cost using standard dense kernels by keeping training tensors physically compact while dynamically reallocating capacity across layers. Unlike existing works that require computing gradients for all parameters, BARTER instead uses statistics collected only from active units during ordinary backpropagation to prune and clone attention heads and feed-forward channels under separate fixed global budgets. At each topology update, it resizes the affected parameter tensors and remaps their optimizer states, so subsequent training steps operate directly on the compact model. Training and fine-tuning experiments with Transformer-based vision models on two image classification benchmarks at densities from $0.1$ to $0.7$ show competitive Top-1 accuracy relative to full-tensor DST baselines. On an NVIDIA L40, BARTER achieves speedups of up to $6.29\times$ in measured training segments and reduces peak allocated GPU memory by up to $81.3\%$ relative to these implementations. An iPhone 13 Pro Max training benchmark additionally shows a $2.78\times$ speedup over the full-tensor DST baseline.

We present Unravel, a training-free framework that turns a single image into an interactive 3D scene, where every object is an independent and movable asset. Existing generators either produce one entangled geometry in which no object can be moved, or predict all objects in one feed-forward pass, where a wrong guess cannot be detected or repaired. Unravel instead takes a different approach: it first takes the scene apart, and then puts it back together. Starting from the input image, an agentic harness that iteratively peels the image layer by layer, and adopts point cloud as major modality for spatial relationship recovery. Specifically, a vision-language model plans which objects are fully visible, an image editor removes them and restores the occluded regions. Every edit is verified before being incorporated into the reconstruction. This process gradually turns occluded parts of the scene into direct observation, allowing objects to be reconstructed when they are best exposed rather than hallucinated from the original image alone. The scene is then rebuilt in reverse removal order, so that the supporting objects are naturally placed before the objects they support. Since a single-view image only reveals part of an object, we treat the observed points as a lower bound on its size: an asset may grow to cover its points, but can not shrink away from them. Finally, we physically settle the reconstructed scene to remove floating objects and interpenetrations, followed by an agentic audit that corrects remaining placement errors. Extensive experiments on diverse indoor scenes show that Unravel recovers more objects, places them more faithfully, and produces scenes that are more physically-plausible than existing methods.

Agent frameworks can persist checkpoints, trace tool calls, and limit resource use, but these facilities often assign different identities to execution, governance, and replay. We present Pollard, a framework-independent runtime based on governed semantic execution trees. A step identity commits to canonical input, kind, attempt, and parent identity; the same tree records completion, refusal, resource settlement, and branch history. Strict replay returns a verified stored result without invoking the wrapped operation, while a transactional store coordinates reservations across concurrent workers. Under a signed deterministic freeze, two state-model replicas each executed the evaluation transitions without an invariant violation, retained-lease PostgreSQL groups were tested against exact limits, strict replay avoided external calls, and all semantic mutations were detected with no false alarms on the controls. A multi-pair framework matrix preserved reference behavior and verified the trees. A separate cost-capped hosted study exercised live provider paths, and every completed Pollard trial produced a verified tree. Together, the studies support the core mechanisms under explicit lease, boundary, and seal-custody assumptions.

Serving stacks for large language models manage context through four recurring mechanisms: incremental edits to a growing context, branching and snapshotting for tree search, identification of reusable prefixes for KV-cache reuse, and encoding of repeated segments. Each mechanism is served today by a different specialist structure. We show that all four reduce to a single persistent weight-balanced rope whose every node carries a polynomial hash of its subtree over a Mersenne prime, so that the structure supports logarithmic edits and logarithmic verified content-identity queries at the same time. We validate each mechanism with pre-registered, replicated experiments, then compare against a deployed specialist per mechanism: the Ropey editing library, a faithful PagedAttention copy-on-write block table, and the vendored SGLang radix matcher. On workloads that interleave edits with identity queries, Hashrope is net-superior to Ropey by a factor that grows without bound (369.7x at one million characters). Against the radix matcher on a live vLLM stack serving Qwen2.5-7B-Instruct-1M on four A100 GPUs, Hashrope identifies a one-million-token shared prefix exactly and 1.20x faster, at a context length the dense engine cannot even prefill.

SLAM (Simultaneous Localization and Mapping) is a core technology in autonomous systems that enables them to estimate their own position and build a map of unknown environments simultaneously. SLAM systems in dynamic environments are of great practical importance. However, existing dynamic SLAM approaches still have several limitations like over-removal and high computational costs. To address these limitations, we propose a real-time RGB Dynamic SLAM framework that precisely removes moving objects at the pixel level without relying on predefined labels or segmentation models. Our system combines pixel motion predicted by ego-motion using DUSt3R with point tracking from All-Tracker, removing only actually moving pixels by analyzing discrepancies between the two motion sources. Unlike object-level masking methods, our approach eliminates only currently moving regions, allowing temporarily static parts to be utilized for tracking. Operating at approximately 10 FPS, the system achieves real-time performance on standard 30 FPS video streams by subsampling every 3 frames. Relying solely on RGB inputs, this significantly improves upon state-of-the-art dense dynamic SLAM approaches (0.5-3 FPS), presenting a highly practical solution.

This paper develops a unified framework for distributional regression with arbitrary response variables. Rather than relying on scalar thresholds or Euclidean structure, we model conditional event probabilities indexed by measurable sets, which extends distributional regression to general response spaces while retaining a direct connection to classical event-based formulations. Within this framework, we study two complementary estimators: a preliminary estimator based on separate regression of event indicators, and a constrained estimator that enforces inclusion-consistent predictions through a directed acyclic graph (DAG) over a finite set family. We establish general error bounds for both estimators and further specialize the results to dense ReLU networks, yielding explicit convergence rates overcoming the curse of dimensionality. Simulation studies and real-data experiments further demonstrate the practical effectiveness of the proposed framework.

LLM agents increasingly operate over multiple entities, where reliable response generation requires not only retrieving relevant evidence but also preserving its association with the correct entity. We identify Phantom Merge (PM), an entity-evidence binding failure in which an agent attributes unsupported claims to its committed entity, potentially even when the task is completed correctly. We formalize PM at the claim level and characterize its prevalence and error subtypes across four agent domains and four backbone families. To address this problem, we propose a two-stage framework for claim-level risk estimation and correction. First, Anchor-Grounded Risk (AGR) combines a representational readout of the agent's pre-emission hidden state with Jacobian-based functional information, providing complementary slot-level and anchor-value risk scores. Second, Fixed-Anchor Correction (FAC) audits flagged claims and selectively revises them using observed anchor evidence or removes them when no supported replacement is available. On the held-out Shopping detection cohort, AGR-slot achieves an AUROC of 96.2% and an average precision of 97.3%, outperforming the evaluated detection baselines. In the end-to-end Shopping mitigation evaluation, FAC reduces trajectory-level PM from 51.4% to 6.2%, with an output-to-original claim ratio of 80.7%. These results demonstrate the prevalence of entity-specific grounding failures and the effectiveness of combining internal risk estimation with evidence-constrained claim correction. Our code is publicly available at https://github.com/LabRAI/PhantomMerge.

Transportation planning has traditionally required substantial expertise in network modeling, traffic assignment, and optimization. As a result, strategic analyses such as network design, capacity expansion, and model calibration remain inaccessible to non-expert decision makers. Recent advances in large language models (LLMs) provide an opportunity to bridge this gap by enabling natural-language interaction with complex transportation planning tools. However, existing LLM-based systems primarily serve as interfaces to simulation software and lack the ability to efficiently search, evaluate, and optimize transportation solutions. This study proposes an integrated human-AI transportation planning framework that combines a large language model, a model context protocol (MCP) interface, an open-source transportation planning engine (AequilibraE), and a differentiable user equilibrium (UE) assignment layer. The framework allows non-expert users to execute network analyses and improve solutions for optimization. A differentiable UE layer is developed by directly differentiating the Karush-Kuhn-Tucker (KKT) conditions of the Beckmann traffic assignment problem. The proposed framework is demonstrated through two representative planning tasks: calibration of travel time functions from observed traffic counts and network design under budget constraints. The examples illustrate how differentiable transportation models and LLM agents can jointly democratize transportation planning and substantially reduce barriers between domain expertise and transportation decision making.

Learned world models for autonomous driving operate in unconstrained latent spaces, yet the physical systems they represent are governed by strict non-holonomic constraints: ground vehicles cannot realize instantaneous pure lateral motion. We study long-horizon latent rollout modeling for autonomous driving and ask whether rollout quality can be improved by assigning a motion-aware geometry to latent transitions, rather than by redesigning the predictor architecture. Starting from an unconstrained latent rollout baseline, we introduce Directional Pullback Regularization (DPR), a training time method that defines a kinematics-aware anisotropic Riemannian metric in decoded state space and pulls it back through the decoder Jacobian to the latent manifold. The resulting latent-space metric assigns higher cost to latent directions that decode to lateral vehicle motion, while preserving the original encoder-dynamics-decoder backbone. This yields a clean three-way comparison among an Unconstrained Latent Rollout model (ULR), an Euclidean Pullback Regularization ablation (EPR), and the full directional method (DPR). We evaluate all models on open-loop ego-motion rollout over a common 10-second test horizon, using training rollout lengths of 1, 3, 5, and 10 steps. Across all four training regimes, DPR achieves lower 10-second final displacement error than both ULR and EPR. The strongest relative accuracy gain appears for models trained with 3-step rollout supervision, where DPR reduces the 10-second final displacement error by approximately 44 percent relative to ULR. The strongest overall control-relevant tradeoff is obtained for longer rollout-trained models, especially at 10 steps, where DPR simultaneously improves displacement error, lateral motion consistency, and forward motion preservation relative to both ULR and EPR.

We consider an orthogonal basis of $\mathbb C^2\otimes\mathbb C^2$ with three entangled states and one product state. This basis is not perfectly distinguishable by LOCC or even by separable measurements when a single copy of the unknown state is available. We show that with two copies no adaptive separable protocol, which measures the copies one at a time, distinguishes the basis perfectly. We then show that a collective separable measurement on the two copies does. Whether two copies suffice for LOCC remains open.

Memory ranking is the stage in personalized Large Language Models (LLMs) that decides which of the candidate memories returned by a retriever enter the limited context, so that the answer is grounded in the right parts of the user's history. Existing memory rankers operate in a single pass, scoring each candidate once against the raw query and cutting the ranked list at a fixed top-K. The single pass misses memories that the answer needs but that look unrelated to the query on their own, and the fixed cut ignores how many memories the query needs. To address both problems, we propose ReMind, a memory ranker that runs as a recurrent and self-terminating loop, following the cognitive theory of memory search: each recalled item joins the cue for the next recall (compound cueing), and recall stops once enough has been collected for the task (metamemory monitoring). For compound cueing, ReMind adds the top-ranked memory to the query context at each step and reranks the remaining candidates under the expanded cue, so that each recall makes the cue more specific and later steps look for what is still missing rather than for what merely resembles the query. For metamemory monitoring, a sufficiency predictor estimates at each step whether the collected memories are enough to answer the query and stops the loop once they are, so that the number of selected memories adapts to each query. Experiments on public benchmarks show that ReMind achieves the best results compared to state-of-the-art memory rankers.

The $r$ largest observations of a block form a dependent random vector whose support moves with all three parameters of the $r$-largest generalized extreme-value ($r$GEV) model. For a fixed number $r$ of order statistics, we show that the model is differentiable in quadratic mean if and only if $k_0<1/2$, where $k=-\gamma$ is the shape parameter; the threshold is thus the same as for block maxima, because only the largest or the smallest retained order statistic can reach the moving boundary. In this range the Fisher information is positive definite and decomposes into the block-maximum information plus contributions of the lower order statistics, and it equals the negative expected Hessian, which we prove without using any estimator. The model is then locally asymptotically normal, whereas for $k_0\ge1/2$ root-$n$ local asymptotic normality fails. Numerically, using the five largest order statistics instead of the block maximum alone lowers the efficient asymptotic standard error by about $45\%$ for the shape parameter and by $20$–$24\%$ for the 100-year return level in heavy-tailed cases ($k_0\le-0.2$), but by only about $2\%$ for the return level when $k_0=0.4$.

Omni-modal large language models (Omni-LLMs) extend LLMs to process heterogeneous modalities (e.g., vision, audio, and IMU signals) alongside language, enabling a broader range of multimodal tasks. Yet we find that these models often fail to express information from non-textual modalities even when they perceive it correctly. We term this disconnect knows-but-cannot-say. To investigate this phenomenon, we introduce Token-level Modality Alignment (TMA), which measures the alignment between individual modality representations and response representations. Our analysis shows that non-textual representations are less aligned with response tokens and receive less attention than textual representations, suggesting a potential explanation for uneven modality preferences. Motivated by these findings, we propose OmniCatalyst, a simple perceive $\rightarrow$ infer strategy that complements the original modality inputs with self-generated text, making non-textual knowledge more accessible. Across seven Omni-LLMs, OmniCatalyst consistently improves performance (+7.29%) and yields more balanced modality preferences, with further gains from diverse training strategies. Further analysis demonstrates that the knowledge extraction phases increase TMA, suggesting that improved alignment makes non-textual information more accessible during response generation. In summary, our findings uncover how Omni-LLMs can better express modality knowledge through language. Our code is available at https://github.com/naver-ai/OmniLLM.

Dataset distillation (DD) aims to retain the training utility of a large dataset in a much smaller set of examples. For skeleton-based action recognition, this requires capturing discriminative motion variation without disrupting the geometry and temporal continuity of articulated bodies. Directly optimizing joint coordinates can violate these constraints, while plausible motion generation alone does not ensure useful training examples. We propose Prototype-Conditioned Motion Dataset Distillation (ProMoDD), which conditions a motion diffusion model on class-wise prototypes of temporally resolved recognition features. We further tailor diffusion training to motion data, enabling the generator to model both skeletal configurations and their temporal evolution. We then construct each compact training set through budget-aware hybrid selection: representative real sequences cover densely populated regions of the pooled recognition space, while selected generated candidates complement them in less-populated regions. This couples prototype-guided generation with feature-space selection, balancing the reliability of observed motions with the variation of synthesis under a fixed per-class sequence budget. Across five difficulty-stratified subsets of the single-person classes in NTU RGB+D 60, ProMoDD achieves its largest gains under tight distillation budgets, reaching 40.32% top-1 accuracy on Full49 with three sequences per class, 4.57 percentage points above class-wise k-means. These gains reflect the complete framework. The same distilled sets train ST-GCN, MS-G3D, and CTR-GCN without regeneration.

Puzzles offer a natural testbed for Large Language Models (LLMs) reasoning because they pose verifiable goals under partial information and require multi-step decisions to reach a solution. However, existing puzzle benchmarks typically focus on narrow environments with rigid mechanics and tightly constrained solution targets. To bridge this gap, we introduce the Active Probing Benchmark (APB), which generalizes puzzle-solving by casting concept identification as an interactive reasoning task that integrates specialized knowledge assessment into the puzzle setting. Specifically, APB evaluates whether models can acquire information via multi-turn active probing and inference to pinpoint 960 target concepts of specialized difficulty across eight diverse academic domains. Our evaluation reveals a counter-intuitive phenomenon: most open-source reasoning models significantly underperform their standard instruction counterparts. This finding suggests that in interactive settings, these models struggle to leverage procedural reasoning to reliably access and apply specialized domain knowledge. Beyond these empirical insights, APB provides a scalable pipeline for generating benchmarks that probe knowledge reasoning under interaction across diverse settings.

Semantic-ID-based generative recommendation unifies retrieval and ranking through autoregressive decoding but differs structurally from discriminative models, which can directly access rich target-item side features. The attributes are compressed into a finite codebook during quantization, while the target item remains undetermined during decoding and its features cannot be looked up. This target-side feature vacuum weakens generation at each semantic-ID level through high conditional entropy and cascading errors, and final ranking through missing target-side feature crossing. Existing methods interact with candidate representations or generate discrete attribute tokens, but none reconstructs a continuous target-side representation inside decoding to support both stepwise generation and ranking. We propose TacGR (Stepwise Target-Side Feature Anticipation and Crossing for Generative Recommendation). Before decoding each codeword, TacGR anticipates a continuous target-side feature vector and crosses it with the semantic prefix and user behavior sequence inside the decoder. The crossed representation conditions current-level generation and is directly used for ranking. Continuous-vector regression risks a learnable target collapsing with its prediction and prediction noise propagating into generation through attention. TacGR addresses these challenges with a semantic-anchoring auxiliary tower and a unidirectional visibility mask, respectively. For extremely long industrial behavior histories, it introduces gated convolutional denoising to suppress noise and adapts the pyramid memory funnel to reduce inference cost. Offline experiments on a large-scale Lazada dataset show significant gains over OneRec, UniRec, and UniSGR. Online A/B testing improves IPV, Order, and GMV by 0.79%, 5.55%, and 3.09%, respectively.

Financial forecasting has traditionally relied on numerical time-series (TS) signals, while textual information such as news remains largely underutilized. We introduce XForecast, a benchmark and competition for news-grounded financial forecasting, where the task is to predict stock price movements four weeks ahead using both historical price data and paired news signals. XForecast combines stock price data with curated financial text generated using FinTexTS, a multimodal dataset generation method. It covers 100 large-cap stocks from 2019–2023, with training on 2019–2022 and evaluation on 2023. To better evaluate models that leverage textual signals, we adopt a weighted hit rate metric that places greater emphasis on instances where news information is expected to have stronger influence. The XForecast Challenge, hosted on Kaggle as part of KDD 2026 Finance Day, aims to benchmark methods for integrating financial TS and textual information.

Gravitational waves from core-collapse supernovae may provide a unique probe of the equation of state (EOS) of dense nuclear matter. In previous work, we assessed the potential of machine-learning methods for EOS classification for rapidly rotating supernovae using individual gravitational-wave detectors. Here, we extend the framework to detector networks and investigate the impact of detector configuration, source distance, arrival time, and sky location on the classification. Using convolutional neural networks, we perform 10-class EOS classification directly from noisy detector data for both second- and third-generation observatories. We find that detector geometry strongly influences classification performance. For Galactic events, current detector networks can achieve promising classification performance, but the reach depends sensitively on the detector network antenna response. Adding detectors improves time-averaged performance and produces more uniform classification across the sky. Third-generation observatories extend the classification reach to 100 kpc for favorable source orientations but do not remove the dependence on detector geometry.