Science & Mathematics

Sparse-view novel view synthesis (NVS) is essential for AR/VR and real-time communication applications, yet remains challenging when reconstructing 3D scenes from limited viewpoints. While 3D Gaussian Splatting (3DGS) excels in dense-view scenarios, it suffers from overfitting and geometry degradation under sparse inputs. Current approaches often tackle these problems from a local perspective, addressing either data quality or training regularization individually, lacking a unified and interpretable framework for systematic improvement. We introduce a new perspective that reframes sparse-view 3DGS from a few-shot learning viewpoint, treating each Gaussian primitive as a neuron and leveraging deep learning regularization techniques. Under this unified framework, our approach generates more effective supervision from scarce observations through two strategies: (1) data-related supervision enrichment via optical flow-guided pseudo-view synthesis and joint camera-scene optimization; and (2) training-time regularization via motion-aware dropout and early stopping, which suppress overfitting while preserving dynamic-scene stability. Extensive experiments on the SIGGRAPH Asia 2025 volumetric video challenge sparse-view dataset demonstrate significant improvements over state-of-the-art optimization-based and feed-forward baselines. Our system won first place in the sparse-view track of the SIGGRAPH Asia 2025 challenge.

Large self-supervised speech models provide powerful representations for speaker verification, but full fine-tuning is expensive. Most conventional adapters insert small trainable modules to modify the output of each layer, while keeping the original residual connections unchanged. Building on our preliminary study of manifold-constrained Hyper-Connections in speaker encoders, we explore parameter-efficient adaptation of both residual paths and feature selection across layers. We first propose a Manifold-Constrained Residual Adapter (MCRA), which splits the hidden dimensions of each Transformer residual state into N groups and uses a small Sinkhorn-normalized matrix to exchange information among them. A learnable gate controls the balance between the original and mixed residual paths. The gate adds only one scalar parameter at each residual connection and initially favors the original path, allowing gradual adaptation. We further propose Frame-Adaptive Layer Aggregation (FALA), an attention-based aggregation method that uses a single learned query shared across all utterances to compute layer weights separately for each speech frame. Because the layer representations depend on the input, the resulting weights vary across both utterances and frames, unlike the fixed weights of a conventional layer-weighted sum. Compared with a frozen WavLM baseline using a fixed layer-weighted sum, our method achieves relative EER reductions of 22.4% on VoxCeleb1-O and 2.9% on CN-Celeb, with trainable adaptation parameters equivalent to only 0.0009% and 0.0016% of the WavLM-Large parameters, respectively.

As physical AI increasingly relies on sensing to understand physical systems, reliable condition monitoring of industrial machinery has become more important. Motors are ubiquitous in industrial systems, and increasing system complexity raises the likelihood of simultaneous faults. Accordingly, compound fault diagnosis that identifies multiple fault components and their severities is increasingly required. Because different fault components jointly affect the same vibration signal, their severity classification can be formulated as related tasks, making multi-task learning (MTL) suitable for exploiting shared information. However, conventional shared-trunk architectures force tasks to share a common representation, potentially causing negative transfer because fault components are characterized by different time-frequency regions. To address this issue, this study proposes a Residual Neural Dimension Reductor with Time-Frequency Channel Attention (RNDR-TFCA), an attention-enhanced cross-talk architecture for multi-output compound fault diagnosis. RNDR-TFCA preserves task-specific feature extractors while selectively exchanging information across tasks. Time-frequency attention emphasizes informative intra-task regions, while channel attention enhances useful inter-task information after nonlinear feature fusion. Experiments on a rotor-bearing system with inner-race fault, outer-race fault, misalignment, and unbalance showed that RNDR-TFCA achieved a macro F1 score of 0.898, outperforming single-task, multi-class, shared-trunk, and cross-talk baselines. Ablation experiments confirmed the benefit of jointly applying both attention mechanisms. RNDR-TFCA required 713 K parameters and 170.42 MMACs, only slightly higher than RNDR. Channel-expanded baselines still underperformed RNDR-TFCA, indicating that the gain was not simply attributable to increased model capacity.

We present a simulation framework for Euclid galaxy clustering measurements that reproduces the essential behaviour of the spectroscopic reduction chain, bypassing the computation required by full pixel-level simulations. The framework comprises two complementary tools, FastSpec, which simulates 1-dimensional slitless spectra with a simplified instrument and noise model, and Pypelid, which computes detection probabilities on input catalogues, accounting for the survey dithering sequence and spatially varying noise. We develop calibrated prescriptions for the survey selection function, redshift errors, and redshift interloper populations. Applied to the Euclid Wide Survey, the baseline selection yields a redshift purity of 74% and a sample completeness of 45%, which improves with the use of Euclid photometric colours. The redshift error distribution is found to have a Gaussian core of width $σ_z = 9.4 \times 10^{-4}$, with no significant trend over $0.9 < z < 1.8$. Emission lines [OIII]5008 and [SIII]9530 misidentified as H$α$ contribute a few to ten per cent depending on redshift, while catastrophic redshift measurement failures dominate at low signal-to-noise ratio. Simulating the Data Release 1 footprint, we find that complex survey selection effects can be corrected with a forward-modelled random catalogue, recovering the monopole and quadrupole of the galaxy 2-point correlation function. The simplified noise model does not include residual contamination from overlapping spectra, or other artefacts seen in early Euclid data; this will be revisited with DR1 data. The framework provides an efficient route for characterising the Euclid spectroscopic selection function at the precision required for galaxy clustering science, and is the basis for systematic error assessments in Euclid galaxy clustering analyses.

Proactive agents can support group collaboration by surfacing useful information without waiting to be asked, but even relevant contributions can be disruptive if poorly timed. In this work, we isolate the timing question: given a fixed candidate contribution such as a clarification or relevant fact, when should an AI agent chime in? We introduce PITCH (Proactive Intervention Timing in CHat), to our knowledge the first benchmark with large-scale human judgments of proactive agent timing collected from participants in the same unscripted multi-party conversations they took part in. To collect high-quality judgments, we design a novel interface in which 2-4 people chat synchronously, author candidate AI contribution messages, place their own and others' contributions into the chat history, and then deliberate as a group about the best placement for each message. PITCH contains 384 candidate contributions and 1,091 judgments, each marking an ideal turn, an acceptable window, or that a contribution should not appear, from 89 participants engaged in 35 conversations. Evaluating 15 models across 28 configurations, we find that while models can place contributions well when given the entire conversation history, they systematically chime in too early when asked to decide live as the conversation unfolds: best-turn accuracy falls from 57.8% to 19.0%, with nearly half of placements occurring before every participant's acceptable window. In comparison, lay humans deciding live are more accurate (23.7% vs 19.0%), and fewer of their placements are premature (39.7% vs 46.4%). Model reasoning traces show that 93% of intervention justifications cite relevance, while almost none consider waiting. By providing a human-grounded benchmark, our work reveals gaps in agents' ability to integrate into group settings and points towards proactive agents that can reason explicitly about appropriate timing.

As jailbreak attacks against large language models (LLMs) become increasingly diverse, red-teaming has come to rely on large libraries of jailbreak prompts and attack strategies to evaluate the robustness of victim models and their defense settings. However, attack effectiveness is highly condition-dependent: an attack that succeeds under one victim-defense setting may fail under another, while exhaustively evaluating the full attack library for each setting incurs substantial token and monetary costs. This motivates studying how jailbreak attacks should be prioritized under a limited attempt budget for efficient robustness evaluation. In this work, we formalize this problem as budgeted attack scheduling over a fixed attack library and investigate three complementary signals under a fixed victim-defense condition: (1) static historical attack success rate (ASR) priors; (2) semantic guidance from LLMs; and (3) online feedback from earlier queries. Building on these signals, we design four scheduling strategies and evaluate them across three victim models and three defense settings on GPT-5.2. Extensive experiments clearly characterize the conditions under which each signal is most effective. Historical ASR provides a strong initialization, but its reliability degrades under defense-induced distribution shifts. Semantic guidance consistently improves or matches static prior-based scheduling in the cold-start regime, while online feedback becomes particularly valuable when transferred priors become stale. Semantic guidance and online feedback can also partially compensate for missing historical priors. Taken together, these findings characterize budgeted red-teaming as an attack scheduling problem in which historical priors, semantic guidance, and online feedback play complementary roles.

Greedy State Boundary Search (GSBS) identifies transitions between relatively stable neural activity patterns in continuous recordings. Repeated template reconstruction and correlation evaluation make analyses across recordings and parameter settings computationally expensive. We present HP-GSBS, an open-source implementation that expresses the matching objective as additive interval contributions and computes t-distance from reusable correlation statistics. Streaming candidate reduction, sequential local refinement, and active-block reuse complement CPU, PyTorch CUDA, and Triton execution paths. Across seven controlled workloads with two input realizations each, median same-input speedups range from 154 to 1,730. All 168 optimized executions reproduce complete reference boundary histories and selected state counts, with a maximum absolute t-distance difference of $1.88\times10^{-11}$. Four publicly supplied BBBD EEG analysis matrices provide complementary cross-backend validation. A naturalistic fMRI application segments 4,230 angular-gyrus voxels in a group response and 16 participants; the 17 CPU fits total 213.10 seconds. Mean pairwise boundary alignment exceeds a circular-shift baseline, illustrating cohort-level use of the saved segmentations. Versioned results, recording-level parallelism, and optional low-memory execution support varied analysis scales. HP-GSBS connects an explicit mathematical reformulation of repeated computation to efficient and reproducible neural-state analysis.

Software engineering (SWE) benchmarks are the primary way to measure the capabilities of coding agents: an agent is placed in a code repository or machine environment, given a request, and judged by hidden tests. These benchmarks almost always start the agent from a clean state that contains no earlier work on the task. In practice, however, agents often continue work that a developer or another agent has already done and reported as complete, with no indication of what remains incorrect or missing. We introduce SWE-Residual, a benchmark for Starting-State Robustness: the ability to maintain performance on a task when the starting state already contains earlier work on that task. SWE-Residual contains 198 Residual States drawn from 111 tasks in DeepSWE and Terminal-Bench, where the earlier work modifies a code repository or a whole machine environment. Each Residual State is the state a predecessor agent left behind after submitting its work as complete, although the hidden tests still fail. The evaluated agent receives only the original request and this state, without the predecessor's trajectory or any sign that the attempt failed; because the request and hidden tests are unchanged, the starting state is the only difference from the source benchmark. Across five frontier agents, success drops by 20.6--53.5 percentage points on DeepSWE and 20.6--34.9 points on Terminal-Bench relative to a Clean Start, and agent rankings on DeepSWE reverse. Failed continuations show a pattern we call acceptance bias: agents treat the inherited work as the definition of what remains to be done instead of checking it against the request. On DeepSWE, in 58.9--88.2% of failed runs, they submit with exactly the same tests failing, although they pass most of these tests from a Clean Start. SWE-Residual thus exposes a capability gap that clean-start evaluation does not capture.

Out-of-distribution detection in continual learning (OOD-CL) is highly challenging, as it requires (i) learning a sequence of tasks without catastrophic forgetting, and (ii) identifying unknown samples (OOD) without access to known (in-distribution (ID)) data from past tasks, where standard confidence-based OOD scores tend to degrade. Existing methods largely rely on a rehearsal buffer of past data or assume closed-world settings, limiting their scalability and reliability in real-world scenarios. In this paper, we propose a novel OOD detection framework for rehearsal-free continual learning, inspired by Kolmogorov-Arnold Networks (KAN), which mitigate catastrophic forgetting via their inherent local neuroplasticity. To better leverage KAN in OOD-CL, we introduce a task-adaptive classifier architecture, TA-KAN, which mitigates recency bias and enhances ID--OOD separability through task-specific control of activation locality. Furthermore, we propose a geometry-guided OOD score that complements TA-KAN classifier confidence without relying on past ID data. Our method significantly improves OOD detection while maintaining strong ID classification performance, achieving state-of-the-art results in OOD-CL.

Automated Gleason grading of prostate whole slide images (WSIs) enables faster and more consistent cancer risk stratification. However, it remains a challenging task due to the subtle morphological differences between grades, the severe class imbalance arising from rare patterns, and the diagnostic features that span multiple spatial scales. Pathology foundation models (FMs) extract strong patch representations but often operate at a single-scale, which limits spatial context and disproportionately affects classes that depend on broader context. The challenge of fusing multiple scales and complementary models, with minimal data and computational resources, remains unresolved. We propose a multi-scale classifier built on state-space models. Our method combines two complementary feature sources: a frozen foundation model, and a lightweight FastVim, pre-trained on ImageNet and distilled from the foundation model, that reads the raw image crops. At each scale, a Mamba model encodes the features from both sources with linear-time spatial context, and a jointly trained head fuses the encoded representations into one prediction. On the PANDA dataset, our method reaches macro F1 of 0.894 and Gleason 5 F1 of 0.802 under 5-fold cross-validation, a 12% and 21% relative improvement over the best single-scale model. A controlled study over 5 foundation models, 5 base classifiers and 3 scales confirms that the gain holds across encoders, with larger improvements observed at smaller scales.

An annotated graph is a pair $(G,S)$, where $G$ is a graph and $S$ is a subset of vertices of $G$. We present a unified approach for definitions of annotated versions of classical graph parameters such as treewidth, pathwidth, treedepth, etc. A generic way to incorporate a distinguished set of vertices into a graph parameter is via torsos. Given a graph $G$ and a subset $S'\subseteq V(G)$, we define $\mathrm{torso}_G(S')$ to be the graph with vertex set $S'$ in which distinct vertices $u,v\in S'$ are adjacent if and only if there exists a $u$-$v$ path in $G$ whose internal vertices lie outside $S'$. Now let $\mathrm{p}$ be any graph parameter. For an annotated graph $(G,S)$, we define the torso $\mathrm{p}$ of $(G,S)$ by \[\mathrm{p}(G,S) = \min\{\mathrm{p}(\mathrm{torso}_G(S')) \mid S \subset S' \subset V(G)\}.\] For a minor-closed class of annotated graphs $\mathcal{C}$, we discuss equivalent conditions for $\mathrm{p}(G,S)$ bounded for all $(G,S)\in\mathcal{C}$ in terms of annotated graphs excluded in $\mathcal{C}$. In many cases, we collect known results from the literature. We include two new arguments for $\mathrm{p}$ being the $2$-treedepth and the feedback vertex number. Among other statements, we also include a variant of the Erdős-Pósa property for annotated cycles. We propose some new research directions.

Large reasoning models (LRMs) have achieved strong performance on challenging mathematical and scientific tasks through long chain-of-thought reasoning, but they often suffer from overthinking with unnecessary reasoning steps. Existing truncation-based methods typically detect consecutive trial-answer agreement and force early exit, yet such local stopping signals do not necessarily indicate final-answer stability and can disrupt useful subsequent reflection or verification, potentially leading to performance degradation. We propose consistency-aware logit masking (CALM), a training-free decoding intervention for mitigating overthinking in LRMs. CALM first monitors trial answers at step boundaries and then applies logit masking to action transition point (ATP) tokens such as "Wait", "Alternatively", and "Hmm", suppressing redundant transitional reasoning while allowing the model to continue its thought process more naturally. Experiments across four advanced LRMs and five benchmarks show that CALM reduces token consumption by 18.6%-25.0% compared with full CoT reasoning while maintaining performance, and can even improve accuracy by 0.07%-1.48% by reducing redundant reasoning. LLM-as-a-Judge results also demonstrate that CALM provides more coherent and concise results.

Mixture-of-Experts (MoE) have become a mainstream way to scale Large Language Models (LLMs), because each token activates only a small subset of experts, reducing inference cost. However, MoE decode remains highly memory-bound and exhibits strongly skewed, time-varying expert loads. 3D-stacked DRAM near-memory processing (NMP) provides the bandwidth that decode demands, yet its tiled organization makes load balancing expensive: expert parallelism (EP) preserves locality but suffers from hotspot-induced imbalance, while more distributed strategies weaken locality and increase NoC traffic. To address this locality-balance conflict, we observe that domain-specialized token routing creates offsetting expert-load fluctuations across decode steps, a property we call expert complementarity. Using this insight, we propose 3D-MoE, a hardware--software co-designed system for MoE decode on 3D-stacked NMP. 3D-MoE uses offline, one-time complementarity-guided grouping and locality-aware mapping to establish coarse-grained inter-group balance, handles residual intra-group hotspots through a lightweight Collaborative Memory-Network (CMN) that streams each hot expert's weight tiles once from its host DRAM channel to helper cores over a reconfigured local broadcast path, and uses cost-aware online remapping to remain robust to workload drift. Across MoE models, 3D-MoE achieves an average 4.63$\times$ MoE-layer speedup over EP and improves end-to-end throughput by 1.58$\times$--1.86$\times$ over the strongest NMP baselines, with 2.11\% additional logic-die area.

Coherent distributed radar architectures enhance the detectability of low-reflectivity targets. However, conventional coherent processing generally entails prohibitively large volumes of inter-station data exchange, thereby motivating the adoption of one-bit quantization strategies. In this work, we investigate a clutter-dominated regime in which the interference is driven by a dense ensemble of weak scatterers. We demonstrate that, under these conditions, one-bit quantization can exhibit a vanishing performance loss with respect to ideal high-resolution sampling in terms of detection capability, in contrast to the classical 2 dB performance penalty associated with one-bit detection in white Gaussian noise environments. For unknown target velocity, we reformulate joint velocity-search processing under the effective one-bit model. Maximizing over the velocity plane makes false-alarm calibration an extreme-value problem for a correlated two-dimensional random field, whose maximum is not characterized by its pointwise null distribution alone. We therefore use an expected Euler characteristic approximation to calibrate the search threshold. We obtain analytic expressions for a separable two-station geometry and use constrained coefficient fitting for larger networks. Monte Carlo simulations demonstrate negligible quantization loss in the clutter-dominated setting and conservative false-alarm calibration under the tested Gaussian slow-time model.

The structural behaviour of cubic (Fm-3m) K2OsCl6 under compression was investigated to 35.4 GPa using angle-dispersive synchrotron powder X-ray diffraction. Upon increasing pressure, K2OsCl6 undergoes a sequence of structural transformations from the ambient-pressure cubic phase to a tetragonal (P4/mnc) phase above 9.5(3) GPa, followed by a transition to a monoclinic (P21/n) phase beginning at approximately 14.2 GPa. The tetragonal and monoclinic phases coexist over the pressure range 14.2-17.2 GPa, while the monoclinic phase is observed as the sole phase at 17.9 GPa and remains stable up to the maximum pressure investigated. Analysis of the diffraction profiles and unit-cell parameters indicates that the intermediate tetragonal structure is likely associated with stress-induced symmetry lowering rather than a distinct thermodynamic phase. This distortion, which is adequately described by the tetragonal P4/mnc space group, is attributed to non-hydrostatic deviatoric stresses arising from the pressure-transmitting medium. Equation-of-state analysis reveals a pronounced reduction in compressibility across the structural transitions, with the bulk modulus increasing from 26(2) GPa for the cubic phase to 71(3) GPa for the monoclinic phase. The pressure-induced structural sequence is analogous to that reported for related compounds such as K2IrCl6, although the transitions in K2OsCl6 occur at lower pressures. All structural transformations are reversible upon decompression.