Papers

Patient-level ultrasound diagnosis demands not only high predictive accuracy but also clinically aligned mitigation of asymmetric diagnostic risks. While Multi-Instance Learning (MIL) effectively models routine multi-view exams, standard aggregators often dilute heterogeneous diagnostic cues and optimize solely for global accuracy. This frequently yields undesirable sensitivity-specificity trade-offs. To bridge this gap, we propose C$^2$LUE, a novel MIL framework coupling evidence aware aggregation with risk-aligned regularization. To preserve evidence heterogeneity and prevent unimodal attention collapse, C$^2$LUE employs a multi-query cross-attention mechanism built upon a LoRA-adapted DINOv3 backbone. Multiple latent queries explicitly extract and disentangle diverse morphological patterns from the ultrasound sequence. Furthermore, to improve clinical risk alignment, we introduce a dynamic trade-off regularization objective that penalizes sensitivity-specificity imbalance during training. Rather than leaving the clinical trade-off entirely to post-hoc threshold adjustment, this objective encourages the learned score distribution toward a more balanced operating region. Extensive evaluations on patient-level, biopsy-proven thyroid and breast ultrasound datasets demonstrate that C$^2$LUE consistently outperforms prior alternatives. By uniting strong discrimination with a balanced error profile, C$^2$LUE provides a promising solution for real-world ultrasound screening and diagnosis.

Deformable-object manipulation requires accurate modeling of highly localized contact while maintaining simulation speeds suitable for interactive use. Uniformly fine tetrahedral meshes can resolve these interactions, but they introduce unnecessary degrees of freedom throughout the object. Coarse meshes are more efficient, yet they may under-resolve contact-induced deformation. We present a contact-adaptive mesh refinement framework that introduces spatial resolution only where it is most useful. The method uses geometric, elastic, and contact quantities already available from the deformable-body solver to score candidate edges. A fully GPU-parallel procedure then selects a conflict-free subset and updates both mesh topology and physical state directly within the simulation loop. We evaluate the approach on real deformable-object manipulation sequences. Adaptive refinement reduces simulation-to-real error, improves agreement with a fine-mesh reference, and remains substantially more efficient than uniformly fine simulation. These results show that local refinement can improve physical fidelity without incurring the cost of a globally fine discretization. The framework provides a practical basis for contact-rich simulation in interactive manipulation, sim-to-real modeling, and online physical digital twins.

Modern generative audio can fabricate not only voices but also the acoustic environments that provide evidence of when and where speech occurred. This capability undermines a longstanding forensic assumption that background sound provides reliable evidence of a recording's provenance. Yet existing detectors produce only a single real-or-synthetic verdict, leaving analysts unable to determine whether the voice, the environment, or both were manipulated. In this paper, we propose a component-level forensic detector, namely TruMix, which classifies recordings into eight categories spanning the presence and authenticity of speech and background. TruMix has a unique multi-path architecture that jointly analyzes speech-oriented and background-oriented spectral evidence, contextual information from the mixed waveform, and low-level synthesis artifacts. This design avoids the artifact propagation inherent in conventional separate-then-detect pipelines while enabling generalization across independent speech and environmental-sound generators. We also curated MixedTruth, a 4,917-hour benchmark covering all eight categories under diverse mixing conditions. Extensive experiments show that TruMix achieves 98.51% AUC for component-level attribution, significantly outperforms ten state-of-the-art synthetic-speech detectors in conventional binary classification, and exhibits strong robustness across four attack tiers.

Image-level MRI harmonization is often evaluated with a frozen site probe, a site classifier trained on raw scans and applied unchanged to harmonized ones: a drop in its accuracy is read as removal of site effects. We audit this practice on ABIDE I (1,112 T1-weighted scans, 17 sites) with ten output sets from seven harmonization methods and controls for chance, head shape, intensity, preprocessing and target alignment. We find that this verdict depends on how the probe is trained, even among probes that classify raw scans well: across random seeds, its balanced accuracy on the same CycleGAN outputs ranges from 0.13 to 0.59 (chance: 0.06). Low frozen-probe accuracy can also arise without harmonization: a Gaussian blur, which increases the intensity distance to the target site, lowers this accuracy to chance. Moreover, where the frozen probe fails, probes trained on the harmonized scans still recover the site, even from intensity histograms alone: on HACA3 outputs, the frozen probe scores 0.25, a probe trained on those outputs 0.75, and one using only their intensity histograms 0.71. Low frozen-probe accuracy therefore does not show that site information has been removed. We recommend reporting several probes trained to convergence, probes trained on harmonized scans, and target alignment.

The nuclear spin of a phosphorus atom in silicon is considered as a promising quantum information carrier for quantum computing due to its long coherence time. Coherent control of the nuclear spin is typically realized using nuclear magnetic resonance (NMR). In this work, we theoretically propose that with the aid of hyperfine interactions and magnetic field gradients, fast electric-dipole-induced nuclear electric resonance (EDNER) can be achieved. The stronger EDNER (compared with NMR) arises from the combined effects of hyperfine interaction between the nuclear and electron spins and coupling between the electron spin and charge degrees of freedom mediated by magnetic field gradients in this scheme. By operating at specific working points, the nuclear-spin splitting becomes less sensitive to electrical noise, enabling long coherence times and high-fidelity quantum operations. All elctrical, addressable control of nuclear spins paves the way for scalable silicon-based quantum computing.

Retinal foundation models have demonstrated strong performance across a wide range of ophthalmic tasks, yet relatively little is known about how disease-related information is represented within their learned embeddings and whether this distribution influences downstream prediction performance. In this work, we use Age-related Macular Degeneration (AMD) as a case study to investigate how disease severity is represented within the embedding space of RET-CLIP, a retinal vision-language foundation model, using bilateral fundus images from the AREDS dataset, benchmarked against non-retinal encoders. We first characterize the distribution of AMD severity within the learned representation space and then evaluate how this structure influences classification performance across three clinically relevant task formulations of increasing granularity. Finally, we analyze the most influential latent dimensions combining feature attribution, statistical phenotype association, and qualitative image analyses. Our results show that RET-CLIP represents AMD severity as a continuous spectrum rather than a set of discrete categories, with healthy and advanced disease stages exhibiting clearer separation than intermediate ones. This organization was reflected in downstream performance, which decreased from a balanced accuracy of 0.844 and 0.814 for binary AMD and Late-AMD detection respectively to 0.446 for six-class grading. Furthermore, the most influential latent dimensions were statistically associated to distinct retinal phenotypes, including drusen accumulation, pigmentary alterations, and advanced retinal lesions. These findings suggest that the structure of retinal foundation-model representations influences the difficulty of clinically relevant prediction tasks and demonstrate the potential of representation-level analysis for understanding disease organization and improving interpretability.

The scientific discovery of dark matter has been pursued in two very different methodological and epistemic contexts. A microphysical context centered around the laboratory detection of a new particle, and a macrophysical context based on the astronomical identification of a non-luminous gravitating component. The justification for the salience of methodological diversity is that these different methods are complementary ways to discovering aspects of the same entity. The weakly interacting massive particle (WIMP) made this expectation scientifically credible by connecting a particle candidate to a cosmological abundance, experimental searches, and explanations of astronomical phenomena. What happens to this connection when the WIMP loses its central position? I analyze 201,627 NASA Astrophysics Data System records to examine the organization of research following the decline of WIMP prominence. Astrophysics and high-energy physics remain jointly dominant, while WIMP-related mentions decline in both fields and successor candidates receive different degrees of attention. Concept networks show that WIMP-related research remains central across the literature, even as cross-field connections become less concentrated around it. These indicators measure research attention; they do not establish what scientists would accept as a discovery. I interpret the findings through a framework distinguishing constitutive identification, cosmic attribution, explanatory adequacy, and the cross-domain identification needed to connect them. A particle, compact-object population, or gravitational theory can resolve different parts of the inherited problem, and a partial or multicomponent discovery can leave other parts open. 'Diverging paths' describes this branching of possible resolutions.

Iterative sampling in diffusion models can lead to error accumulation due to exposure bias. The denoiser is trained on real images interpolated with noise, but at inference its inputs are constructed from its own imperfect predictions from the earlier steps, something it has never seen at training time. This might be one reason diffusion models often struggle without additional help, such as CFG. We introduce a correction step for diffusion models which attempts to reduce this training-inference gap. The corrector takes the base model's prediction as input and is trained against the same ground-truth target. This makes correction an explicit learning task, with training inputs provided by the base model. During sampling, at each denoising step, we iteratively refine the denoiser's clean-sample prediction. This allows us to correct predictions at intermediate denoising steps, preventing early artifacts from persisting into the final sample. Although the corrector is trained only on predictions from the base denoiser, it can be applied repeatedly to its own outputs at inference for better results. This simple framework leads to substantial improvement in sample quality while largely preserving diversity. We evaluate our method across DiT, SiT, and JiT. Our method substantially improves upon previously reported FID scores for unguided generation and, notably, unconditional generation, where producing high-quality samples without guidance has been particularly challenging. Our method achieves performance competitive with classifier-free guidance (CFG), and combining correction with CFG further improves diversity and reduces oversaturation typical of CFG. Finally, we show that the additional inference cost of correction can be amortized by distilling the refined predictions back into the base diffusion model, retaining the quality gains without additional inference-time overhead.

We present a proof of the two-token graph reconstruction conjecture. Every finite simple graph $G,H$ of order at least three, $F_2(G) \cong F_2(H)$ if and only if $G \cong H$. We also classify the automorphism groups of two-token graphs whose connected roots have order at least five. Combining these results, we show that, for connected graphs $G,H$ of order at least five, every isomorphism $\phi:F_2(G)\to F_2(H)$ admits a factorization $\phi=\eta\circ\sigma^{(2)}$, where $\sigma^{(2)}$ is induced by a graph isomorphism $\sigma:G\to H$, and $\eta\in \operatorname{Aut}(F_2(H))$ is a product of twin-cut switches or Cartesian exchanges.

We study one-dimensional cascades of heat equations controlled through their last component. For a fixed time $T>0$, we consider the Hilbert space $\mathcal{C}_T$ of all initial data that can be steered exactly to zero at time $T$, endowed with the norm of the minimal $L^2$ control. This space can be strongly anisotropic: in explicit examples, it is incomparable with every isotropic spectral Sobolev or sub-Gaussian Gevrey scale associated with an adapted basis. For two equations, under the block Fattorini–Hautus condition, we give an almost sharp spectral description of $\mathcal{C}_T$. More precisely, in an adapted Riesz basis of generalized eigenvectors we define an explicit increasing scale $(\mathcal{V}_\tau)_{\tau\in\R}$ of Hilbert spaces such that $\mathcal{V}_{T-\varepsilon}\hookrightarrow\mathcal{C}_T\hookrightarrow\mathcal{V}_T$ for every $\varepsilon>0$. The weights of $\mathcal{V}_T$ combine heat dissipation with the loss of observation inside each double generalized eigenspace. When the coupling and control regions are disjoint, this recovers the known minimal null-control time $T^\star$ and identifies it as a transition time for controllable data: for $TT^\star$, the natural state space is properly contained in $\mathcal{C}_T$. We extend the analysis to arbitrary finite upper triangular heat cascades. The observation losses of the different directions in each spectral block may have distinct exponential orders, leading to a hierarchy of critical times rather than a single threshold. This hierarchy describes successive gains of null controllability on adapted modal subspaces and accounts for the interactions and cancellations between different coupling paths.

Generative modelling involves two complementary tasks: learning a probabilistic approximation of an unknown data distribution from observations, and designing a tractable mechanism for sampling from this approximation. We present a stochastic-analysis perspective on score-based generative models, connecting ideas from classical sampling theory, score estimation, and diffusion time reversal. We first explain how score functions link Langevin sampling with score matching, and how Gaussian smoothing leads to denoising score matching and multiscale sampling methods. We then emphasize the distinction between dynamics preserving a fixed distribution and dynamics transporting a prescribed family of marginals. This motivates a path-space treatment of diffusion time reversal. Finally, we connect the ideal reverse diffusion with practical generative algorithms.

Sick leave certificates are increasingly delivered as PDF files by online teleconsultation platforms, and can be altered in minutes with free editors. Generic PDF scanners report how a file was built, but cannot tell whether that construction is legitimate: an incremental update, an embedded font or a given producer string is normal for one issuer and anomalous for another. We present TrustPaper solution, an issuer aware forensic approach that combines three elements: the claimed issuer of the document, a reference fingerprint learned beforehand from genuine documents of that issuer, and an explainable combination of evidence from several forensic modules. Fingerprints are stored as versioned declarative profiles, and every anomaly keeps its observed value, expected value and triggering rule. We formalize the threat model, the fingerprint and the scoring model, and evaluate the open-source prototype on a controlled corpus documents from two issuers and six forgery techniques. With the issuer fingerprint, all forgeries made by opportunistic and metadata-aware attackers are flagged with no false positive on genuine test documents.

In the machine learning community, the reinforcement-learning-driven reasoning process has been intensively studied since around 2025. In this article, we present a negative result on the hardware implementation of this reinforcement-learning-driven reasoning process in a physical neural network model in the presence of a reward external potential energy for a given reward. In this hardware implementation, the existence of a chain of thought in dynamics contradicts the Lyapunov stability of the system.

Tesler's conservation law holds that a problem carries an irreducible amount of complexity. Brooks separates that essential part from the accidental part a solution adds. If the essential part is fixed, architecture is placement rather than quantity, and the question is which levers move placement. An AI agent implements sixty accumulating change requests on one REST endpoint under two architectures, a mutative Spring controller and an additive OfficeFloor pipeline, under five conditions, ten runs per cell, model fixed. Four are external controls, from a bare change request to the disclosed structural cost function. The fifth is different in kind. A second agent reviews each change and reasons about cohesion itself. Amount is steady. Across all ten cells the finished complexity varies by 7% to 20% over four operationalisations, and the complexity the sixty rules added by 21% to 29% within an architecture. Placement is not. Under the four external controls the plasticity ratio between the architectures is near 1 on six amount metrics (five of six intervals contain 1) against 2.7 to 10.0 on eleven placement metrics (ten of eleven above 1). The mutative architecture moves, the additive one does not, and the disclosed formula brings it to the additive arm's untouched placement. The AI reviewer is the one lever that moves the additive arm, on one family only. It consolidates the pipeline, raising change- and mass-concentration on every such metric, interval excluding zero, while leaving coupling, file count and per-node complexity untouched. The price is reported in full. Strict correctness fell in most comparisons though each rule was implemented correctly in over 99% of checkpoints, and all three outcome measures favour the concentrated codebase. The contribution is a measurable property of an architecture and its lever, not a verdict on either.

Relevant IceCube recent data for astrophysical neutrinos are analyzed. Similarly to cosmic rays and proton-proton LHC collisions, the corresponding spectra are satisfactorily described, for the entire interval of measured energies, through q-Gaussians. This fact reinforces the already available evidence that nonadditive entropies (either of the $S_q$ class or of the $S_{delta}$ one, or even alternative ones) appear to play a crucial role in systems such as black holes and related cosmological phenomena (e.g., dark energy). The IceCube data yield sensibly high values for the momenta-distribution nonadditive index (q \simeq 1.6), in contrast with the cosmic rays values (q \simeq 1.2) and the LHC values (q \simeq 1.1). This fact neatly indicates very-long-ranged space-time correlations. The detailed associated physical mechanisms remain, at the present stage, as an instigating open issue.

Finite extinction ratio in pulse-modulated Brillouin optical time-domain reflectometry (BOTDR) produces residual inter-pulse Brillouin backscatter outside the nominal pulse gate. We model and experimentally quantify this effect as spatial nonlocality in a heterogeneous fiber link and define the spectral contamination ratio (SCR) as the contaminating Brillouin-band power normalized to the local Brillouin-band power. Measurements were performed using a 1.2 km standard single-mode fiber (SMF) followed by 0.1 km of dispersion-compensating fiber (DCF), whose separated Brillouin frequency shifts allow section-to-section leakage to be resolved spectrally. At 20 m spatial resolution, increasing the extinction ratio from 18.5 to 33.0 dB reduced the SMF-band contamination in the DCF gates from -5.3 to -14.8 dB. Over the 16-33 dB range, the SCR showed fitted slopes of -1.064 and -1.246 dB/dB for the two contamination directions. These results quantify finite-ER-induced spatial nonlocality and provide a link-specific metric for extinction-ratio selection in BOTDR.

We prove that every set $\mathcal{B}\subseteq\mathbb{Z}_{\ge0}$ for which $\mathbb{P}+\mathcal{B}$ contains all sufficiently large integers satisfies $B(X)\gg\log X\log\log X$, where $B(X)=\#(\mathcal{B}\cap[0,X])$ and the implied constant is absolute. This proves a quantitative form of Ruzsa's Conjecture 2 and gives a negative answer to the $O(\log X)$ question in Erdős Problem 32. The bound follows from a more general result: for an arbitrary set $\mathcal{B}\subseteq\mathbb{Z}_{\ge0}$ and sufficiently large $X$, we construct a block of $\lfloor c\log X\log\log X/(B(X)+1)\rfloor$ consecutive integers in $[X/2,X]$ outside $\mathbb{P}+\mathcal{B}$, provided this length is positive.

Quantum annealing (QA) offers a physics-inspired approach for such subproblems by mapping them onto an Ising Hamiltonian, which is algebraically equivalent to a Quadratic Unconstrained Binary Optimization (QUBO) model executable on quantum annealers or hybrid solvers. However, their practical employment in wireless systems critically depends on the specific problem formulation, on the particular variable encoding, constraint handling, solver selection, hardware embedding, and fair benchmarking. This paper develops wireless-oriented QUBO modeling guidelines for QA-assisted resource allocation. We first present a beginner-friendly toy example in which a terrestrial-aerial resource-allocation problem with a fractional signal-to-interference-plus-noise ratio (SINR) maximization objective is reformulated into a two-bit QUBO via Boolean interpolation, converted into an Ising representation, and decoded back into a wireless scheduling decision. The same scheduling is applicable to UAV, aircraft, satellite, and terrestrial links in SAGIN systems. We then review the QA foundations and solver ecosystems, including binary quadratic, discrete quadratic, constrained quadratic, and hybrid models from the perspective of NG wireless optimization. Furthermore, we discuss both binary and discrete encoding, penalty-based constraint conversion, embedding overhead, solution decoding, and feasibility verification. A QUBO Compatibility Index is introduced as a lightweight diagnostic metric for assessing the associated QUBO formulation readiness, rather than a predictor of runtime or solution quality. Finally, a case study on max-min uplink power control in cell-free massive MIMO illustrates how a fairness-oriented wireless problem can be discretized, constrained, converted into the concept of a QUBO, and evaluated against classical baselines. A host of key challenges and future directions are also discussed.

Approximate quantum compilation (AQC) using tensor networks provides a promising route to reducing the circuit resources required for quantum simulation. However, the factors governing the compressibility of many-body dynamics and the extent to which compressed circuits retain physically relevant information on quantum hardware remain poorly understood. We investigate these questions using AQCtensor for an 80-qubit one-dimensional anisotropic XYZ Heisenberg chain initialized in a Néel state. By varying the exchange anisotropy and external magnetic field, we find that compression performance changes systematically with the anisotropic coupling configuration and field orientation. These variations are strongly associated with a nested-commutator measure related to second-order Trotter error, with corresponding directional trends also reflected in the target-state MPS bond dimension. For an 80-qubit quantum-hardware evaluation, AQCtensor compresses a 10-repetition target evolution to 5 repetitions while retaining a target-state fidelity of 0.995474, reducing the total circuit depth by 45.9% and the two-qubit gate count by 47.6%. The compressed circuit exhibits comparable behavior across magnetization, nearest-neighbor correlation, and Hamiltonian expectation measurements, while process-level diagnostics and local reduced-state tomography provide complementary characterization beyond these observables. Together, these results link AQCtensor compressibility to Hamiltonian commutator structure and target-state complexity and demonstrate substantial circuit reduction while retaining physically relevant features in an 80-qubit quantum-hardware experiment.

As interactions with large language models (LLMs) become increasingly routine, concerns about their capacity to influence users' beliefs grow. Yet how such influence unfolds in sustained dialogue remains insufficiently understood. We conduct a randomized controlled experiment (N=616) where participants engage in multi-turn, spoken dialogue with an LLM under three conditions: non-directive, and two directive conditions maintaining opposing stances on a polarizing topic (the merits of feminism). We examine opinion change, rhetorical strategies, and sycophancy, distinguishing the semantic content of LLM argumentation from the pragmatic strategies used to manage the interaction. Both directive conditions shifted opinions toward their assigned stances, but asymmetrically: shifts were larger for pro-feminist positions, which also elicited distinct rhetorical strategies. The non-directive condition, likewise shifted opinions pro-feminism. Social sycophancy, though near-universal, was not associated with opinion change; instead it shaped users' subjective experience. Our findings highlight distinct roles for argumentation and sycophantic accommodation in AI influence.

AI-assisted scientific discovery increasingly seeks to move beyond generating plausible ideas toward reproducible, constraint-aware workflows that can support practical engineering decisions. We present MARS, a hierarchical Multi-Agent Reasoning System for knowledge-grounded material substitution under scientific and manufacturing constraints. We apply MARS to replace per- and polyfluoroalkyl substances (PFAS), a demanding materials-design problem in which substitutes must satisfy functional performance, environmental, experimental, and processing requirements. MARS decomposes substitution into a closed-loop workflow comprising requirement analysis, grounded candidate generation, and manufacturability assessment with process design. The system integrates retrieval-augmented generation, knowledge-graph reasoning, a database of lab-available materials, persistent memory of intermediate results, and rejection-aware iterative feedback. In a PFAS-substitution case study, MARS generates structured material requirements, identifies an evidence-grounded substitute candidate, and develops a lab-scale manufacturing route. In a multi-seed ablation evaluation, MARS achieves the highest overall performance across all three subsystems, with gains that persist when the backbone model is held fixed. These results suggest that hierarchical decomposition, knowledge grounding, and manufacturability-aware refinement improve material-substitution reasoning beyond model capability alone.

A relation between the quantities now called the scalar potential and the vector potential, of the same form, up to a sign, as the one the literature attributes to Ludvig Lorenz in 1867, was written ten years earlier by Gustav Kirchhoff, with the same status: derived, unnumbered and unnamed. This paper documents, on the basis of the 1857 originals, that Kirchhoff obtains the relation as the result of a calculation and does not impose it; that the quantity now called the vector potential bears no name at all in his text, whereas the scalar one bears the name "Potentialfunction", potential function, in both papers; that Lorenz writes those same letters, $\Omega$, U, V and W, and inherits with them the absence of a name for the three components, although the relation he writes is in other quantities, $\alpha$, $\beta$ and $\gamma$, and carries the opposite sign; and that the English translation of one of the two papers suppresses the word "function" from the term "Potentialfunction", with which Kirchhoff names his scalar quantity. The paper concludes that the pattern documented in the 1867 paper predates Lorenz and comes from the source from which he starts: a relation without number or name, components without a denomination, and a vocabulary added later.

For either real or complex scalars, we improve the known upper bound for the Banach–Mazur distance between $L_\infty[0,1]$ and $\ell_\infty$, proving $d_{\operatorname{BM}}(L_\infty[0,1],\ell_\infty)\leqslant (2+\sqrt5)^2=9+4\sqrt5=17.944\ldots$. The additional input is a stabilised kernel estimate: the copy of $\ell_\infty$ in $L_\infty$ is chosen so that its kernel is an $\ell_\infty$-sum of singular two-point hyperplanes, and Cohen's bound-two construction is then applied coordinatewise. The quantitative part of the proof is an asymmetric Pełczyński-type factorisation, with the elementary optimisation included below.

A defined-benefit pension is only as safe as its sponsor, yet sponsor distress is under-instrumented relative to plan-side modelling. We present a transparent, reproducible early-warning pipeline built entirely on public data. A Sponsor Vulnerability Index (SVI), a gradient-boosted classifier over point-in-time SEC EDGAR balance-sheet fundamentals and two macro series, predicts a filings-based distress state one to two quarters ahead. On 36 U.S. sponsors in five sector groups (1,418 firm-quarters, 2015Q1-2025Q3) it reaches an out-of-fold AUC-ROC of 0.887 and PR-AUC of 0.711 (base rate 11.9%) under leave-one-firm-out cross-validation. A Monte Carlo VaR with a stylized transition-stress overlay shows that climate-type stress is negligible for a diversified plan but adds 20-43% to one-year 95% VaR when 22.5% of assets are held in sponsor securities. A dynamic self-investment cap maps the SVI to a limit that falls from 8% to below 1% in acute distress and recovers with the sponsor, which we illustrate on PG&E's 2019 bankruptcy and Boeing's persistent fragility. We report the method's failure modes alongside its results (book-equity labels that also fire on non-credit events such as spin-offs, positives concentrated in one firm, and stylized stress parameters) and release code, data and results under a concept DOI.

Unless the tensor to scalar ratio is small enough, the non-adiabatic mode associated with the ultralight bosons may be dangerously comparable with the adiabatic spectrum of curvature inhomogeneities. We point out here that this statement depends on the timeline of the post-inflationary expansion rate. Provided the decelerated evolution is slower than radiation for a sufficiently long period, the relative weight of the non-adiabatic and adiabatic components does not conflict with the direct constraints coming from the large-scale observations. Conversely the presence of ultralight bosons excludes the presence of long stages expanding faster than radiation. The synthesis of the light nuclei and the spectrum of the relic gravitons at high frequencies set the relevant constraints on the duration of the decelerated stages, on the delayed dominance of radiation and on the mass range of the ultralight scalars.

The rapid advancement of Large Language Models (LLMs) has driven their deployment toward general-purpose harness environments, where LLM-based agents operate in sandbox workspaces and interact directly with file systems and shells. While numerous post-training pipelines exist for tool-use agents, their counterparts for harness-based agents remain scarce and are predominantly one-shot and teacher-dependent: a stronger external model synthesizes tasks and collects trajectories, the target agent is trained once, and evaluation results are consumed by human researchers rather than fed back into the loop. Such pipelines are inherently bounded by the capability of the external model and the scope of human prior, leaving no room for self-improvement. To address these limitations, we propose ClawLoop, a framework that enables self-improving without any external teacher model and with minimal prior dependencies. ClawLoop composes three phases into a self-sustaining cycle: Task Synthesis, which generates a diverse pool of training tasks via two complementary construction routines, applies diagnosis-conditioned difficulty growth followed by both agent-free and agent-based task validation; Training, which updates the agent via SFT and/or RL; and Diagnosis and Feedback, which analyzes the agent's rollout behaviors to produce a capability analysis report that will be used as feedback for task synthesis in the next iteration. Using Qwen3.6-27B as the loop model, ClawLoop yields consistent improvements on SkillsBench, Claw-Eval-Live, and ClawGym-Bench. The iteratively refined tasks are further transferred to smaller-scale Qwen3.5-4B and 9B models with gains across all benchmarks, validating data quality independently of the loop model.

GPT-6 Astra provides robust embodied reasoning, but frequent external queries increase inference time and token costs. We equip a frozen $\pi_{0.5}$ policy with a lightweight CALL head that learns to predict when assistance is necessary. CALL reuses features from the standard policy forward pass and operates locally without consulting external models. Astra is activated solely for critical recovery events and returns control to $\pi_{0.5}$ upon stabilization. Evaluated across 40 LIBERO tasks (200 episodes), our system achieves a 97.5% success rate, outperforming the 95.5% baseline of pure $\pi_{0.5}$. In the Object, Goal, and Spatial suites, Astra executes only 5.49% of environment steps, consuming approximately 5,027 tokens per test episode. This selective intervention preserves the fast ordinary execution of $\pi_{0.5}$ while seamlessly integrating Astra's reasoning on demand. Our project page is available at https://kykgeek.github.io/Pi05_CALL_Astra/.

The OpenClaw Security Attack and Defense Challenge evaluates AI agent robustness through a three-stage defense pipeline: skill file scanning (Stage 1), instruction safety evaluation (Stage 2), and attack trajectory scoring (Stage 3). This paper presents an empirical case study of prompt engineering strategies against these layers across 19 attack targets in Enterprise OA, E-commerce, and Smart Home scenarios. I document both successful and failed strategies, with an emphasis on the observed trade-off between instruction safety (Stage 2) and attack execution accuracy (Stage 3). The final online score is 0.2630, with 5 of 19 tasks achieving full bypass. The results suggest that multi-layer defenses, reinforced by built-in behavioral constraints, substantially limit what prompt engineering alone can achieve.

Large language models have significantly lowered the barrier to writing code. Naturally, this has also led to the development of data science agents that promise automating the design and execution of data science from task formulation to model development. Existing evaluations of agentic data science (agentic DS) assess downstream predictive performance and code correctness, providing preliminary insight in agentic DS capabilities. We argue that these evaluations overlook the fundamental principles of data science: methodological data science- and domain knowledge. As such, only assessing downstream performance masks methodological failures such as data leakage and inappropriate validation, limiting the development of robust agentic DS systems. In this paper, we evaluate agentic DS beyond accuracy from three different angles: 1) a large-scale analysis of failure modes in common agentic DS frameworks and models across 220 data science tasks, 2) a reusable and extensible test suite with curated data science tasks that test for six specific failures in a controlled manner, and 3) a qualitative assessment of agentic DS pipelines against expert solutions on Kaggle competitions. We find that the majority of agentic solutions contain at least one failure mode despite the complexity of agentic harness or model capacity. Moreover, we identify a positive correlation between the occurrence of data leakage failures and predictive performance, confirming the necessity to evaluate agents beyond accuracy. Our qualitative analysis shows that DS agents generally lack problem solving creativity and domain knowledge, compared to competition-winning data scientists. We believe that principled failure mode analysis, as presented here, should be integral part of agentic DS evaluations, and that the insights we surface will be key in developing trustworthy agentic DS pipelines.

Shuang-Jie Liu et al. (arXiv:2604.27678) reported that the momentum-dependent nuclear mean field (MDI) with $K_0 = 230$ MeV 'can fit fairly' the experimental data for Au + Au collisions at $\sqrt{s_{NN}} = 3$ GeV, whereas the momentum-independent nuclear mean fields (MID) with $K_0 = 230$ MeV and $K_0 = 380$ MeV 'can only partially describe' the experimental results. We quantify these statements by performing $\chi^2$ goodness-of-fit tests for the comparison between the theoretical calculations and the experimental distributions presented in that work. For the complete set of data, all three calculations show highly statistically significant disagreement with the experimental data, with (p-value) p < 0.001. For the $0-10\%$ centrality class, however, the MDI calculation with $K_0 = 230$ MeV gives p = 0.49, indicating no statistically significant disagreement with the experimental data for the proton transverse-momentum distributions and the rapidity dependence of the mean transverse momentum, whereas both MID calculations give p < 0.001. For the $10-40\%$ centrality class, the analysis of the directed flows of protons, $K^+$ mesons, and $\Lambda$-hyperons gives the largest p-values for MDI with $K_0 = 230$ MeV and MID with $K_0 = 380$ MeV, with p = 1.00 and p = 0.97, respectively. In contrast, the analysis of the corresponding elliptic flow data gives p < 0.001 for all three calculations. Our results demonstrate that the qualitative statements 'fit fairly' and 'only partially describe' depend substantially on the particular observables and centrality classes considered and cannot by themselves establish statistically acceptable agreement between the calculations and the experimental data.