Papers

Vector-boson polarization provides a sensitive probe of new physics in the electroweak sector. Rare processes and suppressed longitudinal contributions make optimal polarization separation essential to exploit the available data. We show that parton-level differential polarization fractions define optimal observables and a fundamental sensitivity limit for polarized cross-section measurements. Their conditional expectations retain all experimentally accessible polarization information, yielding the best achievable sensitivity within the assumed model, and can be learned by neural networks for unbinned or binned inference. Measuring their distributions combines polarization sensitivity with the model independence of fiducial cross sections, probing deviations beyond constant rescalings of polarized contributions. We demonstrate the approach in inclusive ZZ production.

High-fidelity finite-element simulations provide accurate crashworthiness predictions, but their cost limits iterative design exploration. Deep learning surrogates can reduce this cost, but many component-level models are developed under a single prescribed boundary condition, limiting generalisation to boundary variations. This work proposes a Boundary-Condition-Aware Transformer Neural Operator (BAT-NO) for autoregressive prediction of transient displacement fields and scalar crashworthiness responses under variations in geometry and boundary conditions. A B-pillar simulation framework evaluates generalisation across variations in geometry, impact position and velocity, and support stiffness. BAT-NO combines recurrent mesh processing with latent-grid Fourier operator processing. Boundary-condition information is transferred to the latent grid through a hybrid local–global mechanism. Slice-based attention models interactions among physically related regions, while direct boundary-to-grid projection preserves local spatial structure. Across the validation sets for the shape-only, shape-and-loading, and shape-loading-boundary cases, BAT-NO achieves the lowest mean final-step mean nodal Euclidean displacement error among the evaluated baselines. In the most challenging case, it reduces the mean error by 32.6% relative to the second-best model. Hyperparameter tuning reduces the validation error from 0.451 to 0.269 mm, with a comparable error of 0.267 mm on 300 unseen test simulations sampled within the investigated design space. An attention-based scalar decoder jointly predicts six response trajectories with a mean relative error of 2.46%. Most derived crashworthiness indicators have median errors below 3%. These results show that explicit local and global boundary-condition representations improve crashworthiness prediction over expanded component-level design spaces.

Computational tools for biological research (biological tools, or BTs) are advancing rapidly in both quantity and demonstrated capability. While these technologies offer many benefits for scientific advancement, malicious actors may seek to exploit them to design and develop biological weapons (BWs). Generally, BTs require specialized expertise to operate successfully, presenting a barrier to nefarious use by nonexperts. Large language model (LLM) Agents may erode this barrier by enabling actors with limited domain expertise to successfully use BTs. Assessments of how AI could change bioweapons development pathways should examine how LLM Agents interact with BTs on relevant biological design tasks, but such assessments are limited. In this report, we present an assessment of the abilities of LLM Agents in the early stages of interaction with BTs. First, we evaluate seven frontier LLM Agents on their ability to select appropriate BTs for a given task. Second, we assess how these LLM Agents engage with two BTs, EVEscape and ESM3, through targeted case studies. These case studies probe the enabling capabilities required for autonomous BT operation, a potential entry point to a wide range of risk pathways. We identify and assess technical barriers at the intersection of LLM Agents and BTs, offering the biosecurity community and AI developers a foundation for further risk and capability assessments as technologies progress.

This report is a continuation of previous efforts to test the ability of large language model (LLM)-driven artificial intelligence (AI) agents to interface with AI-enabled biological tools (BTs). While rapid advancements in BTs in recent years have brought promise to accelerate scientific discovery, they also raise significant biosecurity concerns about potential misuse. The biosecurity community is particularly interested in the extent to which LLMs can lower technical barriers and assist non-expert users in accessing and operating BTs. Despite this interest, few evaluations have focused on LLM-BT interactions in the context of a defined threat model. To address this gap, this report describes a test of frontier LLM-driven AI Agents on their ability to use BTs to redesign peptides and proteins to evade nucleic acid synthesis screening measures. Highly relevant to biorisk, this task assesses a potential capability of AI agents that could enable a breach of a critical early defensive layer designed to prevent a multitude of biological misuse scenarios. The findings presented here intend to offer a foundation for biosecurity researchers and AI developers to conduct or further risk and capability assessments as these technologies progress.

We prove that every recovering pair $(\mathcal{A},\mathcal{B})$ of nonempty families of subsets of an $n$-element set satisfies $$|\mathcal{A}||\mathcal{B}|\le\left(\frac{223949}{100000}\right)^n=(2.23949)^n.$$ A result of Mond, Souza and Versteegen, combined with the sharp $(9/4)^n$ bound of Fang and Huang for cancellative pairs, gives $(2.2499)^n$ for recovering pairs. We improve this to $(2.23949)^n$. The proof combines two entropy inequalities. One is derived directly from the recovering property, while the other applies the Fang-Huang bound to families obtained from the original pair by a construction of Mond, Souza and Versteegen. The two inequalities are then combined by a convex combination.

We present SONDE (SOrted Neighbor Distance Embedding): a scalable manifold learning approach to create a nuanced, continuous, low-dimensional representation of the geometry of galaxies' local environments. SONDE is a variation on Isomap that uses similarity distances between galaxy neighborhoods to create an embedding space, within which galaxies cluster according to their neighborhood geometries. We additionally present a novel method to project different datasets into the same embedding space, enabling comparisons of environmental embeddings across redshifts. We demonstrate this method on the TNG100 simulations at $z = 0 - 5$ and show that the embedding values evolve with redshift and correlate with galaxies' physical properties. We train a neural network to predict physical properties based on environmental features and show that our first embedding dimension outperforms a standard environmental quantifier, the number of neighbors, while the top three dimensions encode as much environmental information as a ten-bin radial profile. Our results are consistent with known density correlations, including galaxy halo mass, central vs. satellite status, and quenching. SONDE provides a straightforward way to encode subtleties of galaxy environments that could have significant implications for future analysis of galaxy-environment interactions.

Implied stock-market volatility is widely used to identify uncertainty shocks and discipline models of their macroeconomic effects. This paper gives distributional content to the option-price variation underlying these exercises. A normalized at-the-money forward put price, the Downside Protection Premium (DPP), closely tracks the VIX and preserves the principal macroeconomic responses in the specifications examined. Its price and local strike slope separate risk-neutral shortfall probability from conditional severity. Severity accounts for almost all monthly premium comovement and 91.5 percent of the recession log premium. Conditional gains also rise substantially, while conditional losses remain larger in phase averages. Despite this severity dominance, quotations across strikes frequently exclude a common-scale explanation, conditional on the maintained forward and premium construction. Realized-volatility and future-volatility-news innovations both raise protection costs primarily through severity, yet produce different estimated employment responses in common impact-severity units; simultaneous significance depends on the inference formulation. Together, these findings supply joint empirical restrictions on the connection between uncertainty, option prices, and economic activity.

In network speed scaling, jobs arrive over time at a network of servers whose speeds can be tuned, every job must be processed by the servers along one of its allowed routes, and the goal is to minimize the total flow time plus the total energy. For stochastic arrivals, the competitive ratio of Vaze and Nair depends on the network, through the lengths of the routes it uses. We show that this dependence can be removed: we give an algorithm, which routes the jobs by solving a convex program and runs every server at a fixed speed, whose competitive ratio depends only on the power functions; for $P(s)=s^2$, it is at most the golden ratio $\varphi\approx1.618$. The key idea is a lower bound on the optimal cost which, like the cost of our algorithm, is a sum over the servers of a function of each server's load, so the analysis reduces to a single server. We also show that the routing rule of Vaze and Nair can be a factor $\Omega(L)$ away from optimal, where $L$ is the length of the longest route.

Public Exams are an assessment system designed in the context of Biology Education Research to increase student engagement in the exam design process. Students are shown incomplete versions of test questions ahead of time, with the format of the question revealed but some details withheld. Students are also given the chance to provide feedback on the questions, refining them for clarity before the actual exam. In this study, we examined the impact of Public Exams on the experiences and outcomes of N = 42 students in an introductory physics course at New Mexico State University. Using a combination of online numerical surveys, open-ended written responses, and student grade data, we conducted a mixed-methods analysis of the impact of Public Exams on students in this class. We also applied our method to re-analyze the publicly available data of Wiggins et al. Students reported various benefits of Public Exams, including reduced test anxiety and directing to core concepts, and we observed a significant positive correlation between the participation rate of students in editing public exam questions and their average exam grade. But students also expressed dissatisfaction with the incomplete information provided on problems, and they had a less positive response rate than the control sample. In other qualitative responses, students also consistently highlighted the applications of procedural knowledge as most helpful to their learning and a recommendation for the future. Our findings suggest that physics students at NMSU do benefit from Public Exams through reduced anxiety and directing to core concepts, but that the Public Exam system, as previously described, is not optimized for use in physics. We attribute this to the particular emphasis in physics on applying procedural knowledge and suggest modifications of the Public Exam system to better tailor it to a physics context.

Probability-flow ordinary differential equations (PF-ODEs) are widely used as deterministic samplers for score-based diffusion models. Their usual justification is that the Fokker–Planck equation of a diffusion can be rewritten as a continuity equation driven by the score function of the forward diffusion. This identity does not, however, guarantee that the resulting velocity field generates a well-posed flow. We provide theoretical insights into the design of such deterministic samplers for generative models based on diffusions and reflected diffusions. We identify sufficient conditions for a regular Lagrangian PF-ODE flow to exist; reverse sampling and invertibility require two-sided divergence control. For learned scores, sampler stability is controlled by an unweighted velocity error, exposing a mismatch with density-weighted score matching and motivating architectural control of Jacobians, divergence, growth, and compression. Under the manifold hypothesis, positive-time regularization justifies an early-stopped PF-ODE while constants deteriorate near the data endpoint; an explicit sphere example shows that the exact deterministic flow becomes singular as the noise level vanishes even though the diffusion marginals remain well defined. These theoretical insights translate into concrete design principles for stable, invertible, and constraint-preserving diffusion samplers. We illustrate the practical relevance of these design principles using controlled numerical experiments.

We prove that NIP nilrings of bounded nilexponent are nilpotent. We also deduce that NIP $\omega$-categorical $n$-Engel Lie algebras of characteristic $p>n$ are nilpotent, yielding a case of the Lie algebra analogous of Wilson conjecture for $\omega$-categorical groups.

For decades, university mathematics assessed executable procedure, computing and reproducing standard proofs, as a proxy for the understanding it cared about, because procedure was cheap to mark and understanding was not. Generative AI now executes much of this procedure. It does not merely let students bypass the proxy; it exposes how much of what was certified was a proxy. The crisis is not simply cheating but a proxy problem, and resolving it forces a question the discipline has long avoided: what is the irreducible human core of mathematical competence, and how can it be built when some of the executable work can be offloaded? We set out two futures: displacement-and-re-centering, in which AI takes over some execution and education moves toward judgment, formulation and verification; and atrophy, in which premature offloading may erode capacities that depend on prior construction. At the heart of the optimistic future is a verification instability: some forms of oversight are cheap and bounded, while others approach the competence required to construct the work. The outcome is a design choice, not a technological destiny. Rather than add a model, we state the response as two principles: construct before you delegate, and let verification set the ceiling: how far a student can independently verify AI output limits what they may safely delegate. We illustrate with a sequence and two worked problems, and close on the question: whether the construct-before-delegate dependency is a hard constraint or an artefact of how mathematics has been taught.

We demonstrate that Maxwell electrodynamics interacting with a planar Chern-Simons interface does not have plane wave solutions localized on the interface. We re-examine the derivation of such waves presented by N. Kaloper in [Phys.Rev.D114(2026)L051903] and show that at an intermediate step certain terms were missed in an equation therein. Taking these terms into account changes the final conclusion of [Phys.Rev.D114(2026)L051903]. We also show that the method used in [arXiv:2606.12513 [hep-th]] to reproduce the same result is not consistent.

Knowledge distillation is crucial for deploying vision-language models on resource-constrained devices. However, existing methods typically impose uniform supervision across visual tokens or rely on static token selection, which confuses task-relevant cues with background noise and degrades cross-modal reasoning. To address this limitation, we propose Key Visual Evidence-guided Knowledge Distillation (KVE-KD), a framework that dynamically focuses feature distillation on task-relevant visual tokens identified by the teacher model. Specifically, KVE-KD appoints the final pre-generation textual token as a unified semantic anchor and identifies the target cross-modal fusion layer by analyzing changes in the anchor representation through iterative visual-token contribution removal. Within this layer, KVE-KD ranks visual tokens via the anchor-conditioned attention distribution and selects the most informative visual tokens as key visual evidence with normalized entropy. The key visual evidence subsequently guides focused visual feature distillation, making the student align closely with the teacher's task-relevant visual representations while suppressing irrelevant background information. Extensive experiments on six benchmarks demonstrate that KVE-KD outperforms state-of-the-art cross-modal distillation methods, with particularly pronounced gains on tasks requiring complex reasoning and fine-grained visual understanding. Importantly, these improvements are achieved without introducing any inference-time overhead. The source code is available at https://github.com/zhangjianbin07/KVE-KD.

We consider a new algorithmic stablecoin called Dexy consisting of three components: a central bank, a secondary market in the form of a customized CP-AMM liquidity pool, and a trusted oracle. The price of the stablecoin in the liquidity pool is stabilized, relative to the oracle price, via central bank interventions. Users mint Dexy stablecoins from the bank using basecoins, the base cryptocurrency asset.

Real telescope beams are asymmetric, breaking the simple $B_{\ell}$ factorisation in harmonic space and coupling modes across $\ell$ and $m$. We develop a map-domain treatment in which beam convolution is a local, scan-dependent operation applied directly to the time-ordered data (TOD), accommodating the full beam shape and giving a direct handle on intensity-to-polarisation ($I \rightarrow P$) leakage. A noiseless validation against spherical-harmonic convolution reproduces the input temperature and $B$-mode power spectra up to $\ell \simeq 1000$, well beyond the range of primordial $B$-mode science. We outline how the operation maps onto a sparse linear BeamOperator for modular map-making.

Long chain-of-thought (CoT) traces impose substantial output-token costs. Under constrained budgets, compression must preserve answer-critical information, making boundary placement central. Token-level and fixed-length boundaries can fragment coherent spans such as phrases, formulas, and local derivations, whereas step-level boundaries can bind content requiring different compression actions. We introduce SynLat, a text-latent CoT framework that aligns compression boundaries with syntactic structure through non-overlapping Syntax-Aligned Units (SAUs). An answer-conditioned Teacher constructs progressive KEEP/LATENT targets for a single compression-conditioned Student, which generates mixed reasoning from only the question and requested compression level at inference. Across two Qwen3 Student scales, Standard-CoT and Long-CoT groups, and three compression levels, SynLat matches or exceeds the strongest evaluated baseline in all 12 task-group aggregates and strictly leads in 11 under the reported achieved-CR selection protocol. Overall gains reach 3.6/2.6 points at MEDIUM and 7.0/5.5 points at HIGH for Qwen3-8B/14B, with larger advantages under stronger compression, particularly on Long-CoT groups.

Context: The environmental impact of ICT is constantly growing. Enhancing its environmental sustainability requires dealing with energy-hungry software systems, by possibly reducing their footprint. Objective: To better frame the problem and its persistent nature, software can be viewed as carrying structural inefficiencies regardless of how advanced its functionality may be. Identifying these inefficiencies as "green smells" provides a basis for refactoring code to mitigate its ecological impact. We aim at identifying well-known green smells impacting environmental sustainable software and map them to corresponding refactoring aimed at reducing the impact therein. Method: In this article, we conduct a systematic literature review on the articles that discuss green smells and their possible refactorings. We analyse and classify 66 primary studies published up to March 2026. Results: We identify 20 smells impacting environmental sustainability of software. Each smell is associated with one or more refactoring techniques known to mitigate their negative effects on the environmental sustainability of software. Conclusions: The literature witnesses the interest in resolving the green smells affecting software systems. The catalogued green smells and the corresponding refactorings can provide pragmatic value for practitioners in their daily development workflows, while also offering researchers a starting point to establish new research directions in sustainable software engineering.

Answering a question of Madill and Rampersad, we give an example a of a $2$-automatic sequence whose abelian complexity is not $2$-regular.

Pinching-antenna systems attach dielectric pinches to a waveguide, creating line-of-sight radiation points within a few meters of any user. The pinches on one waveguide, however, radiate the same signal, unlike a conventional array, where each element transmits a dedicated signal. Separating the signals of multiple users therefore becomes challenging and requires a new perspective on multiple access. Most existing studies rest on an idealized model, and no survey examines the access schemes under realistic impairments. This article surveys multiple access in pinching-antenna systems and organizes fifty-six works into a unified taxonomy by spatial degrees of freedom. The users of one waveguide can be separated only in time, frequency, power, or guided mode, while multiple waveguides add the space domain. The article then identifies the impairments critical to this separation, which perturb the effective channel, leak interference across orthogonal resources, or couple the pinches on the uplink, and summarizes them in an impairment-versus-method map. A unified evaluation of time-, frequency-, power-, and space-domain access under identical channel and impairment models reveals that the degradation is method specific. The analytical and numerical results give insight into multiple access in these systems and provide a basis for its impairment-aware design, a necessary step toward their adoption in practical networks.

For a bounded operator $T$ between Banach spaces, let $d_+(T)$ be its distance to the operators that are not upper semi-Fredholm, and let $\inn(T)$ be Schechter's quantity $\inf\|T|_M\|$ over infinite-dimensional subspaces $M$. One always has $\inn(T)\le d_+(T)$, and González and Martinón showed that the two quantities are not equivalent in general. We prove that $d_+(T)$ is the infimum of the norms of operators on the whole space that extend a restriction of $T$ to some infinite-dimensional subspace. The gap between $\inn$ and $d_+$ thus becomes an extension problem, and projection constants enter naturally. If the domain is uniformly subprojective with constant $\lambda$, then $d_+\le\lambda\,\inn$. In the other direction we prove a quantitative obstruction: a complemented Hilbertian subspace all of whose infinite-dimensional subspaces have projection constant at least $\Lambda$ produces operators with $d_+/\inn\ge\Lambda/2$. Consequently, on the subprojective but not uniformly subprojective space $(\bigoplus_nL_{p_n})_{\ell_2}$ with $p_n\to\infty$, there is no uniform comparison between $\inn$ and $d_+$. We conjecture that uniform comparison is equivalent to uniform subprojectivity. Finally, we compute $\inn$ for inverses and for shears in terms of the quantity $\tau(T)=\sup_M j(T|_M)$, and we use the arbitrary distortability of Hilbert space to show that a rectangular diagonal operator $\operatorname{diag}(A,B)$ built from two isomorphisms with $\inn(A)=\inn(B)=1$ can have arbitrarily small $\inn$. In contrast, every operator $T$ on a complex Banach space satisfies $\inn(T)\le\tau(T)$, so every automorphism satisfies $\inn(T)\,\inn(T^{-1})\le1$ and the same mechanism cannot work for endomorphisms.

Short-term bike-sharing demand forecasting is complicated by spatial-temporal non-stationarity and the practical difficulty of incorporating unstructured external text into numerical pipelines. Conventional approaches rely on historical flow sequences and fixed graph structures, thereby constraining their accuracy when anomalous social events perturb normal travel patterns. We propose a forecasting framework in which a Large Language Model (LLM) drives a semantic shockwave mechanism that converts free-form urban text, such as municipal event schedules, local news, and transit bulletins, into quantified spatial-temporal perturbation fields. The LLM extracts three physically interpretable parameters per event (intensity, spatial reach, and temporal lag), from which Gaussian decay fields are constructed and injected into a Zero-Inflated Adaptive Spatio-Temporal Graph Convolutional Network (ZI-ASTGCN). To handle the pronounced sparsity of grid-level measurements, the model couples a dual-branch output head with a multi-task zero-inflated loss that jointly trains a gating probability and a conditional flow intensity. Experiments on the operational Barcelona Bicing dataset show that ZI-ASTGCN outperforms established neural baselines, with particularly strong gains during high-demand periods, validating the utility of physics-grounded semantic signals in spatial-temporal mobility forecasting.

A novel discovery is one which is both useful and surprising: a generative model's output is a useful discovery if it has a low probability of being generated (it's surprising) and a high reward (it's useful). Global optimization can directly increase the probability of sampling high rewards but typically requires updating model weights. Such gradient based optimization is expensive and bars using capable closed-source models. Instead, modern search methods for discovery sacrifice the global target, and use evolutionary algorithms with local reward maximizing objectives, permitting the search to focus only on high probability samples. In this paper, we interpret various evolutionary algorithms as approximate Markov Chain Monte Carlo, an optimization-free method to sample from complex distributions. This interpretation allows developing Distribution Matching Evolutionary Algorithms (DME), a class of search methods which sample from a global target distribution without updating weights. Empirically, DME has a higher sample efficiency than existing methods on problems requiring many samples to find a solution.

A leading explanation for why humans cooperate with strangers is indirect reciprocity: when conduct is observed, individuals earn good reputations by cooperating with the deserving. Previous research has identified many "moral systems" that sustain cooperation through indirect reciprocity - each pairing a strategy for how to act with a norm for how to judge others - yet what makes them succeed, and whether that success generalizes beyond the donation game, has remained unclear. Here we evaluate all 1,048,576 moral systems whose norms condition on both players' reputations and actions, across every two-player social dilemma. We find that every highly stable, cooperative system pairs a "discriminating" strategy - one that cooperates with the good and defects against the bad when in good standing - with a "consistent" norm that deems good exactly those who act as the strategy prescribes. Such systems "practice what they preach" and preach what they practice. In the limit of small errors, only one strategy paired with a consistent norm remains stable across all social dilemmas: the "contrite discriminator", which cooperates to make amends when its standing is bad. Our results unify decades of research on indirect reciprocity, revealing moral consistency as the common ingredient of every stable, cooperative moral system.

Capable systems often show jagged competence: low average error alongside failures on particular inputs. We ask where it comes from in a task built from two known layers. A lower layer A computes five per-slot sums from records; an upper layer B uses the slot-1 sum and a mode carried over from earlier boards to predict the next board's category, so B cannot be computed from the current A alone: B is a strict extension of A. We train small recurrent networks on B and measure, against exact ground truth, what they acquire of A. The central finding is that learning B gives the network a jagged version of A, measured through probe readability and supervised outputs. It is readable where B needs it (slot 1, 97-98% by a linear probe) and becomes less readable where B does not (the other slots, 5-10%); category misreads concentrate near the slot-1 cutoffs; and when A is trained explicitly it comes out only approximately right. The networks show jagged competence in B: natural KL below $3\times10^{-4}$ bits coexists with a maximum law TV of about 0.25 on constructed histories. Matched experiments show that even a good A is not enough: connecting a learned A to B cuts misreads 1.3 to 17-fold, the same exact A gives fewer misreads as one-hot inputs (0-2) than as numerical inputs (10-317), one B update makes an exact A inexact unless the B gradient is blocked, and exact A still leaves some B failures. In this task, uneven acquisition, access and preservation of the lower theory explain part of the jagged competence of a network that learns the theory built on it; some failures remain unexplained. Whether the same holds in larger systems is a test to run.

In partially observable Markov decision processes, the optimal policy generally depends on the history of observations and past actions. Asymmetric actor-critic methods have become popular to learn such policies when additional information, such as the true state of the environment, is available during training. The critic, which is not needed at execution, is given access to the state. A critic conditioned on the state alone is generally ill-defined and yields biased policy gradients. Conditioning on the state and the history, the history-state critic restores both. In this paper, we show that conditioning the critic on the state and the policy's own memory, i.e., the internal representation of the history through which the policy selects its actions, is already well-defined and gives unbiased policy gradients, removing the need for a second recurrent approximator of the history. We call it the memory state critic. It follows that a critic based on the policy's memory need not backpropagate its loss into that memory, even though the memory is a lossy encoding of the history. We evaluate the memory-state critic in a vision-based pursuit-evasion environment between two quadrotors across two arena types. The pursuer is the learning agent, and the evader is sampled per episode from a fixed pool of heuristic behaviours. The results show that the memory-state critic outperforms the history-state critic and converges faster. In addition to being unbiased compared to the state-only critic, it maintains a slight edge in the wall arena, where the actor's history carries information that the privileged state alone does not.

Real-world outpatient oncology notes contain specialised terminology, tumour staging expressions, treatment names, toxicity descriptions, and institution-specific de-identification markers that may not be represented efficiently by general biomedical or adjacent clinical language models. We developed and evaluated oncology-specific BERT-style encoders using a governed UK outpatient oncology corpus comprising 290,026 notes from 21,564 patients treated for lung and head-and-neck cancer. We compared RadBERT and PathologyBERT with two local strategies: OncoNote-RadBERT, produced by continued masked language model pretraining, and OncoNoteBERT, trained from scratch with an oncology WordPiece tokenizer. Models were evaluated using masked language modelling loss and perplexity on the validation set, tokenizer fragmentation metrics, clinical term tokenisation, masked-token probes, and exploratory representation analysis. Both external encoders fit the oncology corpus poorly in zero-shot evaluation (perplexity 113.04 for RadBERT; 2035.03 for PathologyBERT), while continued pretraining produced the strongest fit (2.10 for OncoNote-RadBERT). OncoNoteBERT achieved perplexity 2.83 but produced the most efficient tokenisation, with lower subword fertility and shorter normalised sequence length. It also returned a clinically acceptable prediction for 12 of 13 masked-token probes, compared with 7 of 13 for OncoNote-RadBERT. This divergence between corpus-level fit and masked-token performance was partly attributable to tokenizer fragmentation rather than learned semantics alone. Both locally developed models represented the institutional placeholder as a single learnable token. These findings show that continued adaptation and bespoke tokenisation provide complementary benefits, and that representation-layer design matters before adjacent-domain encoders are applied to oncology NLP.

Quadruped robots entering hospitals, care homes, and quiet offices must be context-appropriate not only in where they move but in how loudly they move: a legged robot's locomotion noise, dominated by foot-ground impacts, is itself a social variable. Prior social navigation respects human space but treats the robot as acoustically uniform, while quiet-locomotion methods reduce noise to an operator-specified, context-blind level. We present TACET, a context-appropriate acoustic-social navigation method that infers social context from the robot's egocentric view and decides both where it walks and how loudly, coupling a slow fine-tuned vision-language reasoner to a fast reactive controller through a single compact behavior token, <gait, speed, social_cost>. The same token conditions both a social costmap (where to go) and a quiet locomotion policy (how loudly to move), while a structured out-of-view memory keeps recently seen people in the reasoner's context after they leave the camera view. On a real quadruped, context-conditioned locomotion lowers locomotion noise by up to 9.3 dBA at matched speed, and across our scenarios the full method keeps personal-space compliance at 100% with low acoustic intrusion (<=2.9 dBA), jointly improving spatial and acoustic performance in the evaluated scenarios. The project page is available at https://rcilab.khu.ac.kr/tacet/.

Similarity scores are often offered as evidence that a model shares structure with the brain or across languages. We ask what such a score reads when that structure is removed, or when the instrument measuring it does not work, in two settings. Across seventeen languages, a grammaticality probe transfers worse between more distant languages (r = -0.66). But the probe's own accuracy falls along the same axis and is at chance in four languages, four of the five most distant. Dropping them halves the explained variance, and because it also narrows the range of distances, the design cannot say how much of the gradient is the instrument. Counting the 272 language pairs as independent gives p = 0.0006 for a steering effect that is null when the seventeen languages are the unit (p = 0.155). In brain alignment, a training objective raises a language model's similarity to fMRI responses from 0.10 to 0.34 (reliability ceiling 0.54), yet targets with the brain correspondence destroyed still reach 0.31. A shuffled target scores about k/n against the real one, for rank k and n sentences, a baseline that can be computed before any model is trained. Both interventions work, yet we detect no distance-graded steering effect and no syntactic benefit from alignment. Before reading a correspondence score, measure how much of it survives without the correspondence.

Benchmark scores license claims about the capabilities being evaluated. We audit the inference licensed by an evaluation protocol, rather than the predictive model alone. Using procedural action anticipation as a controlled case study, we study a broader evaluation-validity failure mode: a protocol can remain temporally causal and free of classical target leakage while still supplying a privileged intermediate representation. The failure is not merely optimistic accuracy, but a mismatch between the capability claimed and the construct actually measured. On Breakfast, matched recognizer-history and GT-history regimes score 30.3% and 62.3% (Delta_PH(R_BA) = +32.0, 95% Student-t interval [+24.1, +39.8]). A fixed-checkpoint 2 x 2 intervention isolates a +14.6-point test-time oracle contrast; a matched no-video provenance probe yields a +26.8-point gap; and a boundary-independent query/history control retains a +19.4-point gap. We refer to this discrepancy as a privileged-information gap, defined relative to a specified non-oracle recovery pipeline, and propose a reusable four-condition audit. Across five evaluation settings the gap is heterogeneous; a standard 50 Salads long-term-anticipation re-implementation provides cautious support beyond the audit-specific next-action construction.