Science & Mathematics

Deception detection typically considers messages in isolation. Previous works in claim verification and role detection have displayed that additional information such as gameplay history is crucial to successful predictions. In this work, we demonstrate the value of preceding gameplay dialogue when applied to deceptive message detection within the hidden role game of Avalon. The findings show that LLMs can see improvement across a number of metrics, including accuracy by up to 31\% with the addition of gameplay context for the target message.

Although deep learning-based point cloud denoising methods have demonstrated impressive results in recent years, their effectiveness usually degrades substantially in real-world scenarios where the noise distribution of test point clouds deviates from that of the training set. To address this challenge, we propose a test-time adaptive denoising (TTAD) framework that enables existing denoising approaches to adapt to unknown noise patterns at test time without relying on additional supervision or retraining from source data. Taking a pre-trained denoising model and a test noisy point cloud as input, TTAD first employs a residual-driven noise modeling (RDNM) module to estimate the sample-specific noise distribution through analyzing the deviation between the noisy input and its pseudo-clean version denoised by the initial model. Then, a noise-aware contrastive adaptation (NACA) module is designed to build sample-specific noisy-noisier test-time training pair from the denoised point cloud, and update the model parameters over a few adaptation steps per sample at test time through contrastive learning. Furthermore, TTAD naturally extends to a continual TTAD (CTTAD) setting, facilitating efficient progressive refinement across sequential test samples with consistent noise distributions, while reducing the computational cost per sample. Extensive experiments on both synthetic and real-world datasets demonstrate that our approach significantly improves the generalization ability of existing denoising models under unseen noise, achieving superior performance across multiple evaluation metrics.

Constructing the frozen manifold for entangled bands requires balancing exact reproduction of the target bands with sufficient variational freedom to build a well-localized subspace from the broader Bloch pool. We present a workflow that systematically determines the energy ceiling for a target number of selected Wannier projections, freezing every Bloch state below this threshold across the sampled k-mesh. By measuring projectability, the spectral weight of each Bloch state projected onto the target localized basis, the workflow detects missing conduction channels and quantifies discrepancies between orbital character and energy-based freezing. In comparisons where we controlled the extent of the energy ceiling, freezing all sub-ceiling states is generally favored over projectability-based frozen states. Of the seven materials we tested, freezing all sub-ceiling states lowers the band error on four from between 11 meV and 1.2 eV to 18 meV or less, and matches the projectability-based choice to within about 2 meV on the remaining three. In each of the four, projectability-based frozen states give the lower spread, showing that spread alone cannot identify the better model. With the frozen states held fixed, the workflow minimizes the gauge-invariant spread, constructs the gauge from density-matrix columns on atomic orbitals, and passes an isolated set of bands to Wannier90 for localization. The workflow is implemented in two programs for seamless high-throughput Wannierization with WIEN2k (wien2wannier) and CRYSTAL (LCAO2Wannier). Using a single set of calibrated defaults, the CRYSTAL implementation with analytic overlaps produces models that reproduce the reference bands of five cells, from bulk h-BN in a joint valence-and-conduction model to a magnetic framework with 1424 orbitals. The WIEN2k implementation produces eight such models from seven calculations.

Neural networks trained to predict materials properties learn representations whose physical content can be difficult to extract. We use the neural tangent kernel (NTK), evaluated at trained parameters, to recover the geometry learned by a network that predicts superconducting transition temperatures from elemental composition alone. Kernel ridge regression with the trained NTK closely reproduces the network's predictive performance and substantially outperforms regression with the initial kernel. Spectral analysis and multidimensional scaling reveal a geometry dominated by a small number of directions. Clustering and local neighborhood analysis organize compounds into recognizable superconducting families and connect materials with related chemical and correlated-electron properties despite substantial differences in $T_c$. These results establish trained tangent kernels as practical probes of learned materials representations, connecting property prediction with the exploration of materials space.

Reliable quadrotor flight in disturbed environments requires predictive models that remain accurate as unmodeled aerodynamic effects change during deployment. We introduce a disturbance-aware Koopman Control Family (DA-KCF) framework that preserves an offline-identified nominal model while learning only a disturbance-induced correction online. A reference-dependent KCF is first learned from simulation data and calibrated using limited hardware data. During deployment, the nominal KCF remains fixed while a lightweight residual ensemble learns a low-dimensional disturbance correction from causal state-command-state transitions. The Adaptive DA-KCF model is incorporated into MPC, where disturbance corrections are evaluated along the prediction horizon to optimize position and yaw references sent to the onboard stabilizing controller. Hardware experiments on a nano-quadrotor under uniform wind and a spatially varying wake demonstrate that online adaptation improves multi-step prediction and trajectory tracking as disturbance-induced model mismatch increases, while satisfying real-time control requirements.

A failed gaming trajectory reveals that an agent needs intervention, but not necessarily how its execution harness should change. Effective harness adaptation therefore requires gathering diagnostic evidence, rather than merely proposing alternative recovery rules. We present KPHarness, a framework that adapts the external execution harness of a frozen vision-language-action model through knowledge-probing co-evolution. Recovery knowledge supplies context-specific intervention hypotheses that guide active probing; probe outcomes reveal when these hypotheses apply and provide evidence for refining both recovery knowledge and later probe selection. This reciprocal process allows the harness to improve not only which interventions it applies, but also how it investigates failures. KPHarness separates fast policy execution from a slow research loop that translates accumulated knowledge and probing evidence into bounded harness revisions. Each revision is compared with the incumbent under matched sandbox conditions and accepted only when the paired evaluation shows improvement. Recoveries supported by further validation are retained for cross-task reuse, while immediate interaction feedback and delayed episode-success credit update action- and knowledge-probing policies during adaptation. Experiments on Minecraft tasks spanning mining, combat, and crafting show improved task coverage over bare policy execution, a manually specified fixed harness, and knowledge-only harness evolution with non-adaptive probing. Evaluation on previously unseen tasks further shows improved task completion with the evolved harness, knowledge bank, and probing policies held fixed,

3D point clouds generation requires capturing both part-level layout and surface detail while preserving permutation equivariance. Existing latent generative models often encode an entire shape into a single vector, which limits their ability to retain part-level structure. We propose TokenFlow, a hierarchical flow matching framework that represents each shape as a set of latent tokens and couples a rectified flow over these tokens with a conditional flow over points. A bottleneck of $M \ll N$ learnable tokens aggregates global context through cross-attention at $O(NM)$ cost. On ShapeNet, TokenFlow achieves the best MMD and coverage across airplane, car, and chair, while obtaining 1-NNA comparable to the strongest baselines. On ModelNet, its learned latent set reaches the best reported linear-evaluation accuracy on ModelNet10. These results show that representing a shape as a set of latent tokens, aggregated at $O(NM)$ cost, provides an effective representation for 3D point cloud generation.

The adoption of social companion AI (SCAI) is often framed in terms of the positive features it offers users, including its nonjudgemental nature, agreeableness, and availability. Here, we argue the appeal of these features primarily arises from the difficulties inherent in human sociality. SCAI reduces what we term relational complexity, or the demands and risks inherent in human relationships. Drawing insights from studies of human cognition, communication, and human-computer interaction, we identify six distinct sites of relational complexity that we posit are integral to understanding SCAI adoption. These include: (i) relationship formation and dissolution; (ii) achieving mutual understanding; (iii) interpersonal motives and strategic uncertainty; (iv) social partner evaluation and self-presentation; (v) social embeddedness and information flow; and (vi) reciprocity and relationship maintenance. By attenuating relational complexity, SCAI offers users some of the experiences associated with companionship while reducing the demands intrinsic to obtaining such experiences from another person. However, the lower interpersonal risk and diminished cognitive load that attracts users to SCAI may contribute to myriad negative impacts, including relationship displacement, social demotivation, social deskilling, and risks to privacy. The concept of relational complexity thus provides a framework for studying both the appeal of artificial companionship and its implications for human wellbeing.

Mixture-of-experts (MoE) models already make a sparse choice for every token: which experts should process it. We ask whether a sparse dictionary should reuse that choice. NR-SAE gives each routed expert a small dictionary and opens only those the frozen router selects. At block 15 of Gemma 4 26B-A4B, after ten million C4 training tokens and across three seeds, the expert-aligned autoencoder explains 0.518 of the variance in router-weighted expert outputs on a fixed evaluation capture. A rank-addressed control with the same parameters but sixteen times as many candidates reaches 0.481, a candidate-matched control 0.458, and an architecture-matched control with randomized expert identity 0.387. A transcoder variant also leads all three controls: 0.432, against 0.347, 0.371, and 0.118. Part of this advantage is architectural: each expert writes into a subspace of at most a quarter of the model width, so naming the expert names the subspace. How much it explains, we have not measured. At this layer and budget, the router already knows something a dictionary would otherwise have to learn. Our implementation did not train faster, and semantic interpretability, downstream loss, and causal faithfulness remain untested.

In the telephone model a message starts at a source vertex, and in every round each informed vertex may call one neighbour. Telephone multicast asks for a schedule that informs a given set of $k$ terminals in the fewest rounds. On directed graphs, the best known ratio for this problem was $\tilde O(\sqrt k)$, due to Elkin and Kortsarz and to Hathcock, Kortsarz and Ravi. For every integer $\ell\ge2$, we give a deterministic combinatorial algorithm with running time $n^{O(\ell)}$ that computes a schedule of length $O(\ell^3\log k)\cdot(\mathrm{OPT}+k^{1/\ell})$, where $n$ is the number of vertices and $\mathrm{OPT}$ the optimum. Since $\mathrm{OPT}\ge\log_2(k+1)$, this gives ratio $O(k^ε)$ in time $n^{O(1/ε)}$ for every fixed $ε>0$, and ratio $O(\log^4k)$ in quasi-polynomial time. The method rests on jump trees with additive vertex loads, a height reduction that bounds the load of every vertex, and a recursive greedy algorithm with an exponential potential. We also show that directed multicast is a special case of the degree-bounded directed Steiner tree problem on a time-expanded digraph. With known algorithms for that problem this gives randomized ratios $O(\log^2n)$ in time $n^{O(\log n)}$ and $O(n^ε)$ in time $n^{O(1/ε)}$, which, unlike ours, grow with $n$. On the hardness side, directed multicast stays $(1-ε)\ln k$-hard even when $\mathrm{OPT}=O(\log k)$, the regime in which the previous algorithms were weakest.

Atomically flat and chemically controlled oxide substrate surfaces are essential for high-quality epitaxial growth of complex materials and heterostructures. Previous studies have demonstrated in situ thermal preparation of selected oxide substrates, primarily focusing on conventional low-index orientations such as SrTiO$_3$ (100) and LaAlO$_3$ (100). Here, we extend these methodologies to a broader range of oxide substrates and crystallographic orientations, including challenging polar surfaces. We present a systematic study of the in situ thermal preparation of SrTiO$_3$ (100), SrTiO$_3$ (111), GdScO$_3$ (110), LaAlO$_3$ (111), (LaAlO$_3$)$_{0.3}$(Sr$_2$AlTaO$_6$)$_{0.7}$ (001), and NdGaO$_3$ (110) single crystal substrates. The substrates are annealed in a vacuum chamber under a controlled oxygen background environment using a CO$_2$ laser heating system. The optimized annealing conditions produce atomically flat step-and-terrace structures, long-range surface crystallinity, and effectively clean surfaces across all the investigated substrates. Our results establish a versatile, reproducible, and broadly applicable platform for the in situ preparation of oxide substrates immediately prior to epitaxial growth. These results also provide practical guidelines for interface engineering and surface-sensitive investigations of complex materials.

Privacy-preserving machine learning (PPML) offers a promising path toward bringing powerful AI services to the edge without exposing sensitive user data. However, its practical deployment remains constrained by the high communication, computation, and data-movement costs of secure multi-party computation (MPC), particularly during nonlinear inference. Existing MPC systems typically decouple input-independent correlated randomness generation from online execution, creating substantial memory pressure to buffer single-use cryptographic data. Moreover, limited storage can force a stop-and-refill execution pattern that repeatedly stalls the pipeline to replenish these buffers, bringing randomness generation back onto the critical path. We propose Interflow, an algorithm-hardware co-design that rethinks this separation by batch-wise processing and pipelining correlated-randomness generation with online execution: while the current batch consumes pre-generated randomness to execute secure nonlinear operations, correlation generation proceeds concurrently for the next batch. By leveraging interleaved execution, efficient data movement, and hardware acceleration, the design transforms preprocessing from a serialized, storage-intensive offline phase into a continuously overlapped pipeline that generates correlated randomness closer to the point of consumption, reducing memory footprint and data movement overhead. Implemented on the Zynq UltraScale+ ZCU102, Interflow improves nonlinear evaluation throughput by up to 1.55x for CNN operators and 1.48x for transformer operators. It achieves up to 1.26x speedup for ResNet-50 and up to 1.21x speedup for BERT-base, and reduces memory footprint by up to 37%, compared with baseline frameworks.

Cytokine-mediated inflammation disorders such as sepsis and cytokine release syndrome demand bedside multiplex panels that resolve pro- and anti-inflammatory dynamics in minutes, yet label-free FET cytokine sensors remain limited to few devices, single analytes, and PDMS packaging that sorbs small molecules and leaches oligomers. We present a CMOS platform pairing a 128 x 128 ion-sensitive field-effect transistor (ISFET) array (SkyWater 130 nm, SiN gate) with a FlexDym 4 x 1 micro-well overlay that partitions the membrane into four chemically independent regions (approximately 2,176 pixels each, square-root-of-N averaging). Two wells carry capture antibodies (anti-IL-6, anti-IL-10), one is a non-specific reference, and one tests specificity; the transduced signal is defined as the differential current relative to the reference signal (Delta I_D(W_n) - Delta I_D(W_3)). Single-analyte dose-response measurements spanning 100 fM to 10 nM successfully resolve both IL-6 and IL-10 above the non-specific detection floor, including at healthy-serum baseline concentrations. The array shows a per-pixel pH sensitivity of 1.13 microamperes per pH unit (CV approximately 14%). Across two chips, device-to-device variation is dominated by transconductance spread (g_m CV approximately 62%); a one-time per-device g_m calibration reduces analyte-measurement variance to approximately 16%. We further validate the surface functionalization, characterize device-to-device variation, and demonstrate the FlexDym integration workflow. Simultaneous multiplexed cytokine measurements resolve IL-6 (W1) and IL-10 (W2) against the chemistry-only reference (W3). At the same time, cross-specificity characterization is assessed by applying IL-6 to an anti-IL-10 well (W4), establishing a scalable label-free ISFET architecture for multiplexed cytokine sensing.

Long-horizon terminal agents may solve a task in one run yet fail across repeated runs. Stochastic errors can propagate through actions and context, creating a gap between capability and reliable execution. We present RiseHarness, a framework for iteratively improving agent harnesses with model weights fixed. It focuses on flipping tasks with both successful and failed runs under the same configuration, using successful runs as references for failure analysis. RiseHarness combines agent trajectories with internal harness traces from a runtime monitor to examine how model actions interact with command execution, feedback, context management, and completion decisions. A Harness Handbook links this evidence to harness behaviors and source code, helping identify recurring problems in mechanisms shared across tasks. RiseHarness then searches other harness frameworks for relevant implementations and adapts them as design references to the current model and execution protocol. Candidate changes undergo behavioral checks and repeated task evaluation; versions with measured gains seed the next iteration. This loop integrates suitable mechanisms from different frameworks into a single harness and uses new execution evidence to guide further changes. We evaluate task success and reliability across models, ablate harness execution evidence and external research, and test transfer to unseen benchmarks with the same model. RiseHarness treats a specific model as the unit of harness adaptation, studying how changes to shared interaction mechanisms can make existing capabilities more reliable across tasks.

Sparse attention limits the keys used by each query, but finding those keys can still require a full-prefix scan at every layer. We introduce $MLIPS:$ $Multi$-$Layer$ $Indexer$ $Pool$ $Sharing$, a training-free method that shares candidate search while retaining each consumer's own scoring and selection. A producer publishes a top-$M$ pool larger than the attention budget $K$; subsequent layers select their own top-$K$ from that pool. Measurements across three sparse-attention model families show substantially higher coverage from broad pools than overlap between final selections. We implement layerwise sharing, persistent decode reuse, and a coarse producer in vLLM for DeepSeek-V4-Flash. Across four long-context suites, three consumer layers per producer give $1.14$-$1.40\times$ speedups and $14.8$-$27.4\%$ lower GPU energy, with observed accuracy changes of $-1.49$ to $+0.29$ percentage points in single runs. More extensive sharing exposes task-dependent quality losses. Sampled prefill traces also reveal recurring selected KV positions across query groups and layers. We characterize this shared working-set structure and use it to motivate Top-$G$, an exploratory bounded pool construction for prefill. Its deployment, task accuracy, and serving benefits remain open.