Papers

Covert channels exploiting network protocols for data exfiltration and command-and-control (C2) are integral parts of modern cyberattacks. In search of a significant covert channel within the fabric of the Internet, we targeted the combinatorial properties of the Client Hello (CHLO) packets in the ubiquitous Transport Layer Security (TLS) protocol. The proposed Helol tunnel is a novel covert approach to embedding information in TLS Client Hello packets, which involves the strategic rearrangement of their cryptographic information elements. To sustain TLS protocol extensibility, the recent anti-ossification TLS compliance measures encourage the interactive middleboxes and next-generation firewalls (NGFWs) to preserve the parameter configuration in the Client Hello packets. Furthermore, to improve user privacy, popular Internet applications are varying their TLS CHLO parameter configurations to resist TLS fingerprinting by third-party network entities. We demonstrate the strength of the Helol tunnel to exploit these recent developments to evade NGFWs with interactive proxy and comprehensive threat protection. We also numerically show the efficacy of Helol tunneling over state-of-the-art covert channels that exploit TLS through the use of real traffic captures and public TLS fingerprinting data.

Machine learning predicts outcomes well, but predictive accuracy does not ensure reliable counterfactual responses. We examine how to combine machine learning and theory for measurement and counterfactual analysis, using maritime CO2 emissions where physics provides a benchmark speed response. Matching hourly tracking data for dry bulk and container ships to annual fuel consumption reported under EU regulations, we compare engineering calculations, structural regressions, hybrid models, and machine learning. Out of sample, all estimated models predict within a few percent of reported totals, outperforming standard engineering calculations. Yet pure machine learning and unrestricted structural regression imply attenuated speed responses. Hybrids preserve the structural component's speed response by excluding speed-related inputs from the machine learning component. A cost-benefit analysis of speed reductions illustrates the policy stakes: an attenuated speed response can flip the sign of net benefits. Accurate aggregate predictions therefore cannot substitute for scrutiny of the restrictions determining counterfactual responses.

In the streaming set cover problem, $m$ sets from a universe of size $n$ are arriving one by one in a stream, and the algorithm is allowed to process the stream using one or a few passes and a space of $o(mn)$, which is sublinear in the input size. The goal is to determine the minimal (or approximately minimal) number of sets that cover the universe at the end of the last pass. This problem has been studied extensively over the years with rapid progress that led to several $O(\log{n})$-approximation algorithms in $\tilde{O}(mn^{1/p})$ space and $O(p)$ passes. However, progress on this front has largely stagnated over the past decade, despite the absence of any lower bounds that rule out even an $O(\log{n})$-approximation in $O(m)$ space and just two passes. We provide a simple explanation for this lack of progress by establishing an optimal three-way space-pass-approximation tradeoff for this problem: any $\alpha$-approximation algorithm for streaming set cover requires $$ \widetilde{\Omega}\Big(\frac{m}{\alpha} \cdot \big(\frac{n}{\alpha}\big)^{1/p}\Big) $$ space in $p$ passes whenever $\alpha \ll n^{1/(p+1)}$. In light of prior work, this result is optimal up to constant factors in $p$ and logarithmic factors in $n,m$ for any $\alpha\geq p$. Our bound is optimal with respect to the range of $\alpha$ also, and fully settles the complexity of this fundamental problem in the streaming model. The proof of this result is (surprisingly) simple and non-technical and relies on a randomized reduction from a variant of the standard pointer chasing problem in communication complexity, using elementary properties of random sets.

Decision models return probabilities intended for routing, abstention and automated action. Calibration makes those probabilities useful on average, but does not establish whether low confidence reflects chance or missing knowledge, nor whether confidence falls when a model moves beyond what it knows. We audit this distinction in Jev, a decision model, with over 15 public datasets and 6 generated task families, with paired interventions that vary the information supplied for a fixed item. Jev's confidence is calibrated on familiar closed-choice tasks but fails as an indicator of missing knowledge: with no answer-relevant information it assigns up to 0.80 to a salient option, and on news beyond an observed knowledge boundary it exceeds accuracy by 0.21–0.33, a gap that recalibration on earlier months does not close. Targeted yes/no questions give sharper readouts of the case: whether an outcome is settled (AUROC 1.00) and whether the evidence suffices (0.95, against 0.85 for confidence on the same items). Asking whether Jev knows the answer appears to flag fabricated entities and post-boundary news (0.91), but with realistic names or with dates removed it shows no advantage over answer uncertainty. Black-box knowledge audits therefore need explicit controls for surface cues. Code: https://github.com/Syntheme/beyond-answer-confidence.

As artificial intelligence is increasingly deployed, algorithmic unfairness has raised growing concerns and intensified demands for transparent fairness auditing. In practice, the tolerable degree of algorithmic unfairness depends on the specific legal, ethical, or application context. Given a prespecified tolerance threshold, an important statistical question is how to determine whether a group disparity exceeds the allowable tolerance across different auditing objectives. To address this problem, we develop a unified tolerance-based fairness auditing framework for two complementary auditing objectives: violation certification, which prioritizes control of false violation declarations, and sensitivity screening, which prioritizes reducing missed violations. For the first objective, we develop a constrained empirical likelihood test for formal settings that uses least-favorable-point calibration and can be combined with false flagging rate control for simultaneous subgroup auditing. For the second objective, we develop split empirical likelihood and adjusted split empirical likelihood tests using an adaptive boundary-proxy principle for early-warning settings. Numerical experiments show the distinct error-control–sensitivity trade-offs of these procedures. A COMPAS analysis illustrates the framework in predictive fairness auditing.

In this paper, we explore the impulsive dynamics common to single-joint, two-link models of walking and brachiating gaits with respect to slope and switching time. In particular, we investigate how the stability of a gait and bifurcations encountered within a family of gaits change under time-based and state-based switching of the impulsive dynamics.

In this work, we systematically investigate homoepitaxial $\beta$-Ga$_2$O$_3$ growth by halide vapor phase epitaxy (HVPE) on (010), (001), and ($\bar{2}$01) native substrates over a broad growth parameter space, extending the growth temperature as high as 1150 $^\circ$C. The growth kinetics exhibit strong dependence on precursor supply, temperature, reactor pressure, and Ga-source-to-substrate distance. Increasing the HCl flow from 80 to 300 sccm enhances the growth rate by up to seven-fold, while increasing the temperature from 1050 to 1150 $^\circ$C nearly doubles the deposition rate. At 1150 $^\circ$C, atmospheric-pressure growth yields rates more than 2.5 times higher than those at 100 Torr, while reducing the O$_2$ flow to 5 sccm further increases the growth rate to 14.3 $\mu$m/h, producing films up to 42.9 $\mu$m thick in 3 h. Notably, phase-pure homoepitaxial layers with good crystalline quality are obtained across all three orientations, with low rocking-curve full width at half maximum (FWHM) values of 63.4, 66.6, and 42.8 arcsec for the (010), (001), and ($\bar{2}$01) epilayers, respectively. The surface morphology exhibits a pronounced crystallographic dependence and evolves strongly with the HVPE growth conditions. In particular, the ($\bar{2}$01) surface develops well-defined step-and-terrace morphology with local RMS roughness as low as 0.17 nm. These results extend $\beta$-Ga$_2$O$_3$ HVPE homoepitaxy into a higher-temperature regime and demonstrate a broad process window for high-rate growth of thick, high-quality epilayers with strongly orientation-dependent surface evolution.

Who or what is conscious? Because subjective experience is directly accessible only in the first person, judgments about consciousness in other entities depend partly on analogy. Historically, such inferences have focused on nonhuman animals, but advances in artificial intelligence have raised the possibility of conscious AI. Here we develop a causal framework for evaluating such evidential analogies. The key distinction is between similarities in factors plausibly involved in generating consciousness and similarities in downstream behavioural or cognitive effects. Our framework weights source-target similarity by causal relevance while allowing for unknown causes, disabling differences and alternative routes to consciousness. Applied to biological systems, it explains why analogical support generally weakens with increasing causal distance from humans. Applied to contemporary AI, it suggests that behavioural similarity provides only limited evidence for consciousness because relevant causal correspondences remain poorly established. The framework also clarifies what evidence would strengthen claims of artificial consciousness.

Model-based reinforcement learning (MBRL) can exploit simulated experience, but a simulator-to-target shift creates a model-selection problem: correcting the simulator and fitting the target directly can each fail under limited target data. We introduce the Model-Corrected World Model (MC-WM), which separates initial target data into disjoint fit, selection, and calibration partitions and deploys the family with lower standardized calibration risk. A learned confidence signal and deterministic validity predicates weight one-step imagined policy updates without rewriting physical rewards. We evaluate 540 unique reported run cells across three controlled Multi-Joint dynamics with Contact (MuJoCo) shifts; one exact-routing cell was repeated after a pre-deployment artifact gate, giving 541 completed executions.

Large Language Models (LLMs) are increasingly used as probabilistic generators for simulation, synthetic data generation, and decision support in settings where real-world data are unavailable. Yet, the structure and reliability of the distributions they produce remain understudied. Here, we systematically analyze LLM-generated distributions of preferences for air travel, restaurants, and consumer products. Encouragingly, all models considered in our analysis exhibit self-coherence, with the most probable outcomes stabilizing rapidly under repeated sampling. At the same time, we observe substantial discordance across both model families and scales, with little consensus even among their most probable outcomes. These patterns hold across nine open-weight models, three choice domains, and show robustness under temperature changes, greedy decoding, and perturbations of prompt and ordering. Our findings indicate that outcomes are influenced more by the choice of model than by the wording of the prompt, challenging the common assumption that sufficiently capable LLMs produce similar preference distributions when used as stand-ins for survey respondents.

Dividing source galaxies into tomographic redshift bins is a cornerstone of modern weak gravitational lensing analyses, enabling measurements of the growth of cosmic structure and the nature of dark energy. In practice, these tomographic bins are defined using photometric redshift (photo-$z$) estimates. However, correlations between photo-$z$ estimates and weak lensing shear can introduce redshift-dependent selection biases in the measured shear signal; if left uncorrected, these biases distort the inferred amplitude and redshift evolution of the lensing signal, and in turn bias the measurement of the growth of cosmic structure across cosmic time. In this paper, we extend the analytical self-calibration for shear measurement (AnaCal) framework to account for photo-$z$-based selection biases in tomographic weak lensing analyses by propagating shear responses through the selection process. This approach eliminates the need for external image simulations to calibrate this correction. As a first sanity check on real data, we validate the photo-$z$ estimates derived from AnaCal fluxes on the Rubin Observatory Data Preview 1 dataset, and find that they reach photo-$z$ quality comparable to, and at high redshift slightly better than, the standard LSST estimates. We then validate the shear calibration on LSST-like image simulations with blending at the expected LSST Y10 depth, with two representative photo-$z$ algorithms – a template-fitting method and a machine-learning method – and show that the multiplicative shear bias induced by photo-$z$ selection remains within the LSST ten-year requirement $|m| < 3\times 10^{-3}$ across all five tomographic bins for both algorithms. These results establish AnaCal as a self-consistent pipeline for tomographic weak lensing science in upcoming LSST analyses.

We study finite-horizon portfolio optimization with proportional transaction costs and trading opportunities arriving at the jump times of a Cox process. Borrowing and short-selling are prohibited, while utility functions need not be concave, increasing, or differentiable. The admissible class includes differentiable utilities with asymptotic elasticity greater than or equal to one. The associated Hamilton–Jacobi–Bellman equation is semi-linear and nonlocal, with control entering only through a zeroth-order term. We show that the normalized equation admits a unique bounded solution that is continuous in all state variables and classical in time and log-price. The proof combines a contraction argument with interior Schauder estimates, and verification identifies the value function and yields an optimal Markovian feedback strategy. Numerical examples with non-concave utilities exhibit, at fixed time and pre-trade wealth, multiple intervals associated with the same trading action, departures from the usual buy–no-trade–sell ordering, and abrupt switches in the optimal post-trade risky weight.

Knowledge distillation aims to transfer the factual knowledge of large language models to smaller models for efficient deployment. Yet a teacher may recall a relation in one direction while failing to generate the answer in the reverse direction. Distillation from its generated answers can therefore propagate this directional limitation to the student. The same teacher can nevertheless recognize such an answer by scoring the relation in the direction it knows. We introduce directional label distillation, in which frozen teachers score candidate answers in that known direction and the best-scoring candidate becomes the student's training target. On facts about parents and their children, known-direction scoring yields more accurate labels than scoring the requested direction, even after tuned corrections for name priors. With prior-corrected scores, the better direction depends on the facts rather than the template, and reverses on mined facts whose notable entity is the parent rather than the child. With the evaluated children's forward facts withheld, students trained on known-direction labels improve open-ended accuracy on their trained queries by 13 to 15 points over students trained on prior-corrected reverse labels. After generated answers are matched to a fixed name list by lexical similarity, students reproduce nearly all selected labels. Their accuracy largely follows label quality. The label advantage holds on unscreened queries and when candidates are retrieved without inserting correct answers. Our findings show that directional verification mitigates the transfer of errors from teacher-generated answers to students by providing more accurate training targets. Code is available at https://github.com/js-lee-AI/directional-verification.

Reliable roll prediction of unmanned surface vehicles (USVs) is essential for ensuring navi?gational safety and enhancing autonomous decision-making. While existing studies primarily focus on improving prediction accuracy, the quantification of prediction reliability remains insufficiently addressed. To bridge this gap, this paper proposes a reliability-aware prediction paradigm that integrates confidence assessment into the predictive pipeline. The architecture utilizes a multi-task learning structure where a shared feature extraction backbone feeds into dual heads: a regression head for precise roll prediction and a quantification head for confidence scoring. This configuration provides accurate prediction and corresponding confidence for risk?sensitive downstream tasks. In addition, an adaptive centralization strategy tailored for short?term real-time roll prediction is introduced to improve model generalization under varying operational conditions. Experiments conducted on a real-sea dataset demonstrate that the proposed method effectively quantifies the reliability of prediction results and maintains superior generalization under varying conditions, offering significant potential for practical engineering applications.

A search for dark matter particles is presented using events with large missing transverse momentum and a single energetic jet containing few constituents. The analysis uses proton-proton collision data recorded by the CMS experiment at a center-of-mass energy of 13 TeV. This corresponds to an integrated luminosity of 138 fb$^{-1}$, recorded during the 2016$-$2018 LHC data-taking period. This is the first search at the LHC to employ the low-multiplicity jet signature. Supervised machine learning is used to maximize sensitivity to potential signals in this challenging final state. No excess of events over the standard model background prediction is observed. The results are used to set upper limits on the dark matter production cross section at 95% confidence level (CL) within simplified models featuring vector and axial-vector mediators. For a dark matter particle mass of approximately 150 (550) GeV, mediator masses up to 4250 (3500) GeV are excluded at 95% CL.

Existing synthetic image evaluators typically provide only a scalar quality score and do not identify the image regions that support it. We introduce VIEScore2, a unified evaluator for image generation and editing tasks with optional conditioning images. VIEScore2 represents an image as an N x N grid and jointly predicts quality scores and defect locations in a single model pass. Its text-native grid representation provides a common interface for heterogeneous spatial supervision and enables directly verifiable post-training objectives. We train on 38K examples spanning score-only, localization-only, and joint supervision across generation and editing tasks. Starting from supervised fine-tuning, we further apply GRPO to improve defect localization using rewards that combine cell-level Dice overlap, score accuracy, and output-format validity. A parameter-free parser converts the structured predictions into readable explanations. On the primary suite, VIEScore2 achieves an overall-score SRCC of 0.601, compared with 0.491 for Gemini-3-Flash, the strongest zero-shot general-purpose VLM baseline under matched inputs. For defect localization, VIEScore2 outperforms both general-purpose VLMs and specialized spatial evaluators on three of six benchmarks in per-image grid IoU and ranks among the top three on five, including datasets beyond its training sources.

Before a machine learning model ships, it often has to pass a suite of automated tests. Requiring every test to pass looks safe, yet it can reject many models that would have served users well, and it does not say how trustworthy a passing model actually is. We treat the release gate as a design problem: choose how many tests a model must pass so that cleared models meet a stated reliability target, while keeping as many good models as possible. A two-class latent-factor model makes both costs explicit and reduces each calculation to a one-dimensional integral. We prove that when both classes share the same latent correlation, a stricter gate always raises reliability, so the gate that keeps the most good models is the most lenient one that still meets the target. Under pass-all gating, any reliability target short of perfection is attainable within the model, but the share of good models kept tends to zero as the suite grows. Correlation between tests sets the price. In one configuration, a 99 percent target needs 8 independent tests, but 74 tests at a latent correlation of 0.3 and 5,182 at 0.5, where the gate keeps fewer than one good model in ten. We also give a validation procedure, built on exact binomial bounds, that certifies a gate from labelled data even when the gate is chosen from a fixed shortlist.

Learning-based driving planners are usually trained and evaluated in open loop against logged trajectories. In closed loop, a trajectory with small displacement error can still stall the vehicle, steer it into a conflict with surrounding agents, or be executed with abrupt braking. We introduce Closed-Loop Refinement and Execution (CLRE), a hierarchical receding-horizon control framework designed to mitigate these failure modes while leaving the upstream planner frozen and adding no new learned model. The upper layer treats the nominal trajectory as a reference and solves a finite-horizon optimal control problem that trades route progress against interaction with predicted agents. Solving it from several initializations gives a candidate set, and a prediction-conditioned oriented-bounding-box (OBB) feasibility test retains only candidates whose minimum predicted OBB clearance over the horizon meets a threshold. The lower layer executes the lowest-cost survivor, or a route-centerline backup when none remains, through the tracking controller supplied with the planner, augmented by a range-based speed bound and a saturated proportional braking law. In closed-loop simulation on the 220-route Bench2Drive validation set with VAD as the upstream planner, CLRE raises the driving score from 42.26 to 55.89 and route completion from 55.69 to 71.51, and reduces collision events from 117 to 97.

Majority voting over sampled completions is the workhorse of test-time scaling, and reinforcement learning with verifiable rewards (RLVR) is the workhorse for making each completion better. The standard pipeline composes the two: train one policy with RLVR, then sample it many times and vote. We show that this composition is lossy. A vote can only overturn mistakes that its voters do not share, and RLVR sharpens a policy so that its samples increasingly make the same mistakes. With every method drawing exactly $160$ completions per problem, training a single LoRA adapter on the full RLVR budget raises single-sample accuracy on every model we test ($1.5$B-$8$B). Yet on three of four models it leaves the majority vote below that of the untrained base model, by up to $4.8$ points. The damage builds during training: voter errors grow steadily more correlated, and the majority vote accuracy peaks early before falling by up to $7.0$ points. The cause is concentration, not RLVR itself. We split the same data and training budget across $K$ LoRA adapters, each trained on its own random disjoint shard, and call the result an adapter thicket. Thickets out-vote the fully trained adapter in all $16$ (model, $K$) settings, and for $K{\geq}4$ they stay within $0.8$ points of the base model or above it. A single adapter stopped early, at a thicket member's step count, is a strong control that matches thickets for small $K$. For $K{\geq}8$, thickets keep more of RLVR's single-sample gain and out-vote this control in six of eight settings. The cost of concentration also grows with the number of votes: from $16$ to $160$ votes, the thicket's lead over the fully trained adapter widens from $1.3$ to $3.3$ points. When the plan is to sample and vote, an RLVR budget is better spent broad than deep.

Spectroscopy reads structure by sending a structured probe into a system and measuring what comes back. Here we apply this principle to animal decision-making. Harmonic Theory casts choice as motion on an angular landscape over heading, whose Fourier components form an animal's decision spectrum. We show that the arrangement of cues enters that landscape as a structure factor, so it factors like a diffraction amplitude, and symmetry imposes selection rules: a p-fold cue array annihilates every harmonic that is not a multiple of p, and sweeping two cues apart draws a notch through each harmonic. Cue geometry becomes a tunable probe: the experimenter sets which harmonics an experiment can see. We read these spectra from public recordings of flies, schooling fish, larval zebrafish, mice and a swimming alga, and from the fly's neural compass, where symmetric scenes suppress the forbidden harmonics and breaking that symmetry restores them. Measuring the compass form factor favours the theory's smooth bump over its square one, and raising cue contrast narrows that bump in every fly. A width fitted to one harmonic then predicts the next. Cue geometry is a spectrometer for decisions, and symmetry sets what it can see.

A sudden transition from an early matter-dominated era to a radiation-dominated era provides a representative mechanism for enhancing induced gravitational waves (GWs). In the primordial black hole (PBH) evaporation scenario, discrete PBH number-density fluctuations naturally generate Poisson isocurvature perturbations. By evaluating the GWs induced by these perturbations through the sudden evaporation transition, we show how the full hierarchy of intrinsic non-Gaussian statistics of a Poisson isocurvature source enters the induced GW spectra. We find that the tensor power spectrum retains only the two-point correlations of the scalar source, while higher-order correlations first appear in the tensor bispectrum. We also find a nontrivial scale dependence in the relative amplitudes of different bispectrum configurations, which may provide additional information for distinguishing the discrete PBH scenario from other induced-GW mechanisms.

Cryptic ligand-binding pockets are not apparent in experimentally determined apo structures, making them difficult to identify from unbound receptor geometry. A complementary challenge is to make the structural measurements and learned evidence behind each prediction directly inspectable. We introduce a supervised algebraic counting field (ACF) for predicting cryptic-pocket residues from apo structures. ACF compiles explicit geometric, physicochemical, and topological features into compact, integer-weighted lookup tables. Each prediction score can be reconstructed from feature values, training counts, table weights, and spatial aggregation, without sequence search, structural-template transfer, or a protein language model at inference. We evaluate ACF on CryptoBench and two locked external collections, separating ranking performance from the effects of residue-calling budgets. On an external set of 57 post-CryptoBench apo-holo units, ACF exceeded P2Rank by +0.044 in mean paired ROC-AUC (multiplicity-adjusted 95% CI [+0.010, +0.079]). The advantage was dataset-dependent: official-fold ROC-AUC and matched-budget F1 differences against P2Rank remained unresolved, and a second external evaluation did not confirm gains from added structural features. ACF thus provides a compact predictor with externally validated signal and an inspectable path from structural measurements and training counts to residue scores.

One-dimensional transition metal dichalcogenides (TMD) nanoribbons (NRs) offer a promising route to aggressive channel-width scaling in nanoscale transistors. However, controlling their layer number, a key determinant of device performance, has remained elusive. Here, we demonstrate a chemical vapor deposition (CVD) approach for layer-by-layer growth of MoS$_2$ NRs on a low-symmetry sapphire surface. This approach enables the oriented growth of single-crystalline bilayer NRs with a 2H stacking configuration, while maintaining widths below 20 nm. Increasing the MoO$_3$ precursor supply was found to be crucial for promoting the growth of the second and subsequent layers. By controlling the growth time and precursor supply, the layer number of MoS$_2$ NRs can be increased up to four layers. This method can also be extended to the growth of vertical MoS$_2$/WS$_2$ hetero-NRs. The average carrier mobility and current density of bilayer NR field-effect transistors (FETs) are two- and three-times higher than those of monolayer counterparts, respectively, and a maximum mobility of 86.7 cm$^2$V$^{-1}$s$^{-1}$ was attained with a bilayer NR-FET. These results highlight layer number as a key parameter for optimizing NR device performance. Our thickness-controlled growth strategy provides a general route towards TMD NR-based transistors with enhanced scalability and performance.

High-dimensional, low-sample-size bulk gene expression data poses a fundamental challenge in transcriptomics, which typically includes tens of thousands of genes but relatively few samples, leading to overfitting and unstable selection for key features as biomarkers. We propose a regression by community centers (RCC) framework tailored for such data. RCC converts gene expression data into a feature proximity network and leverages the phase transition of the giant connected component to identify a critical distance threshold that yields a sparse yet maximally informative network representation, indicated by a minimum normalized shortest compression length. Intuitively, at the criticality, the network balances fragmentation and over-connectivity, allowing meaningful gene-gene associations communities to emerge while filtering noises, such that community-centers capture the most informative and non-redundant signals. These representative genes are then fed to a simple ordinary least squares (OLS) model for downstream predictions. Across five bulk gene expression datasets, RCC displays exceptional prediction performance, clearly outperforming benchmark methods, remains robust to missing data and noises, and does not rely on external biological knowledge. These results suggest that network-based representations provide an effective and general framework for high-dimensional, low-sample-size prediction tasks beyond transcriptomics and illustrate how network-science ideas can support robust learning in data-scarce regimes.

Kolmogorov-Arnold Networks (KANs) represent a compelling alternative to traditional Multi-Layer Perceptron (MLP)-based neural networks. By employing activation functions as learnable elements, KANs offer superior interpretability, making them suited for scientific domains. In this work, we investigate the neural scaling laws of KANs and the structural evolution of their learnable activation functions under dataset expansion. Specifically, we evaluate the scaling behavior of three KAN variants—BSRBF-KAN, Gottlieb-KAN, and Faster-KAN—across standard image classification benchmarks (MNIST and Fashion-MNIST) and a specialized scientific regression task (magnetic parameter estimation from domain images of moiré magnetic textures). Our results demonstrate that the test loss ${\cal L}$ exhibits a broken neural scaling law (BNSL) behavior as a function of the dataset size $N_D$. After passing through a random-guess regime, the loss follows architecture- and task-dependent scaling behavior. The loss crosses from a faster- to a slower-scaling branch, ${\cal L}\propto N_D^{-\alpha}$ and ${\cal L}\propto N_D^{-\beta}$ with $\alpha>\beta$ for image classification tasks. The exponents $\alpha$ and $\beta$ depend strongly on both the specific network architecture and the dataset-size regime, ranging from 0.4 to 1.5 and from 0.06 to 0.6, respectively. For the magnetic parameter-regression task, the loss follows a single scaling law with its exponent ranging from 1.28 to 2.59. Additionally, we provide a structural analysis of how activation functions refine their complexity as data volume increases, finding that dataset expansion drives a transition from simple linear-like approximations toward stable, interpretable symbolic forms. These findings provide a quantitative roadmap for the efficient application of KANs while managing the trade-off between model expressivity and computational overhead.

Accurate medication recommendation is central to clinical decision-making, directly determining therapeutic efficacy and patient safety. However, existing methods suffer from two key limitations: drugs are often abstracted as discrete tokens, ignoring their molecular structures and pharmacological mechanisms, and the commonly used "flat" recommendation paradigm fails to leverage the hierarchical logic of the internationally standardized Anatomical Therapeutic Chemical (ATC) classification system. To address these issues, we propose HADRec, a Hierarchy-Aware Drug Recommendation framework that integrates molecular knowledge with electronic health records (EHRs). HADRec employs LLaMA-7B to encode clinical notes for rich patient representations and ChemBERTa to encode drug Simplified Molecular Input Line Entry System strings, building a global molecular knowledge base. A cross-attention mechanism then performs deep multimodal fusion between patient states and drug features. The framework further incorporates a hierarchical predictor and a novel consistency constraint loss to enforce strict adherence to ATC logical dependencies. Extensive experiments on MIMIC-III demonstrate that HADRec achieves state-of-the-art performance across Jaccard, F1, and PR-AUC. External validation on MIMIC-IV confirms strong generalization under distribution shifts, and calibration analysis shows well-calibrated predictive confidence on MIMIC-IV with ECE = 0.04, and Brier = 0.06. Counterfactual evaluation reveals clinically aligned reasoning, disentangling disease-specific treatments from general care. Together, these results establish HADRec as a high-performance, interpretable, and clinically grounded pathway toward safe and reliable AI-driven medication recommendation.

Post-training quantization (PTQ) methods typically use sequential quantization that partitions a pre-trained LLM into a series of units (e.g., transformer blocks), with one unit quantized at each stage. State-of-the-art PTQ methods are predominantly learning-based, optimizing auxiliary quantization parameters (e.g., scaling factors, rotation matrices, clipping thresholds, and adapters) via gradient descent to minimize a reconstruction loss. A common practice is to use mean squared error (MSE) as the reconstruction loss function, yet its induced optimization behavior remains largely unexplored. In this work, we take a holistic view of sequential quantization and systematically investigate how optimization evolves from the first quantization stage to the last, aiming for a deep understanding of optimization in learning-based PTQ schemes. Through extensive empirical studies spanning representative learning-based PTQ methods, LLM families, model scales, architectures, quantization settings and various tasks, we consistently uncover Optimization Imbalance: reconstruction loss magnitudes vary dramatically across stages, accompanied by highly uneven gradient magnitudes and parameter updates under MSE. We term the cross-stage range of loss magnitudes the reconstruction loss scale, and reveal that MSE translates the unexpectedly large reconstruction loss scale into highly uneven gradient magnitudes, which in turn lead to uneven optimization strength across quantization stages. This finding suggests a general principle for improving learning-based PTQ: optimization strength across stages should be decoupled from the reconstruction loss scale. Theoretically, we show that root mean squared error (RMSE) variants defined at the sample, channel, token, and element levels naturally realize this principle through implicit gradient normalization, outperforming MSE significantly as a drop-in replacement.

Vision-language-action (VLA) models struggle on history-dependent manipulation tasks, where the current observation alone does not determine the action, and the policy needs a memory of the history. Existing memory methods decide what to remember by design, for example, keeping frames with large pixel changes, and show inconsistent gains across tasks. We view what to remember as an optimisation problem. From the POMDP formulation of imitation learning, we show that the optimal memory maximises the conditional mutual information $I(a_t; m_t \mid o_t)$ between the action and the memory given the current observation. Intuitively, this means preserving the action-relevant information in the history that is not already contained in the current observation. Based on our analysis, we propose Divide-and-Remember (D&R), a recursive memory method that learns a memory function $m_t = M(h_t)$ and scales to long contexts while staying compute-light. It involves two strategies: (1) the selection over the full history is divided recursively into subproblems of top-$K$ selection over $2K$ tokens, so that fixed-size, lightweight selectors learned end-to-end support an unbounded history; (2) all recursion blocks share one selector, which captures the selection rule common to every block and keeps the method efficient. On RoboMME, a benchmark of 16 long-horizon manipulation tasks that require remembering when, where, what, and how to act, D&R achieves a state-of-the-art average success rate with consistent gains across all four suites under a budget of only 64 tokens; real-robot experiments show the same gain. Code, checkpoints and more results are at https://dnr-memory.github.io/

We focus on language-conditioned flow-based manipulation, where robot flows (robot velocity fields) serve as embodiment-agnostic, motion-centric representations for leveraging data collected from multiple robot platforms. This task is crucial because language-conditioned manipulation is essential for practical robotic systems, yet scaling robot foundation models remains limited by the labor-intensive collection of embodiment-specific data. Existing methods either coarsely approximate robot flows with sparse keypoint displacements, or cannot handle language-conditioned manipulation. To address this limitation, we propose NarrativeFlow, which models robot flows as continuous velocity fields using a flow-matching formulation conditioned on language. Accordingly, NarrativeFlow generates robot flows that are physically consistent with real-world manipulation. To validate NarrativeFlow, we have conducted experiments on standard datasets for language-conditioned manipulation. The experimental results show that NarrativeFlow outperforms representative baseline methods on standard evaluation metrics. Furthermore, through real-world experiments, we show that NarrativeFlow achieves higher success rates than baseline methods across multiple manipulation tasks. The project page is available at https://shota0520.github.io/NarrativeFlow-project-page/

Multi-agent AI scientists have shown improving performance across a diverse range of tasks. Yet a common approach is design-time agentic orchestration, which typically relies on fixed workflows. In contrast, human scientists coordinate and adjust their division of labor at runtime. We therefore ask: can AI scientists also coordinate at runtime? To this end, we introduce Runtime Agent Coordination (RAC), which selects agents from existing AI-scientist hosts during execution, assigns scoped work contracts, and provides artifact-grounded verification. Verification informs subsequent agents without blocking transitions or discarding artifacts. We conduct a single-seed exploratory evaluation across Agent Laboratory, EvoScientist, and ARK on ResearchClawBench, preserving host models, tools, and permissions under host-calibrated budgets. Four cumulative conditions separate native execution, runtime communication, runtime selection, and the combined addition of contracts and verification. Runtime selection yields the highest observed mean score for each host; adding contracts and verification reduces these means, with host-dependent outcomes relative to native execution. These results motivate runtime coordination while exposing the limits of additional coordination mechanisms under constrained budgets. Code is available at https://github.com/systemind-team/Runtime-AI-Scientist.