Papers

Ecological fitting describes a situation where species interact despite no shared evolutionary history. It is both a null model for adaptation and an important process in its own right, for example, the establishment of a non-native species is ecological fitting, and it is a major mechanism in the formation of Novel Ecosystems. In this paper we introduce and analyse a model of ecological fitting. We demonstrate both analytically and through simulation that the main factors determining invasion success are the invaded ecosystem's population and diversity, which set the size of the barrier for invaders. We also study the situation where many evolutionarily unrelated species simultaneously colonise an empty area. We show that a successful equilibrium is likely to be established, but the diversity tends to be very much less than the number of species in the starting pool and discuss the implications of this.

Interstellar extinction at high Galactic latitude is often traced by thermal dust emission, but the conversion from infrared surface brightness to reddening is sensitive to grain properties and temperature. We test dust-scattered far-ultraviolet (FUV) and near-ultraviolet (NUV) diffuse emission as a high-resolution extinction tracer towards the North Galactic Pole (NGP), where E(B-V) < 0.1 mag. Using the GALEX diffuse maps of Murthy (2014), we build FUV and NUV intensity maps at 6' resolution over the region b > 70 deg and compare them with the Planck E(B-V) map at matched resolution. A direct pixel-to-pixel regression gives only moderate correlations (rho = 0.536 for FUV, 0.243 for NUV) owing to high per-pixel scatter from H2 fluorescence, line emission, and photon-counting noise; we therefore construct per-pixel GALEX uncertainty maps and apply a centroid-fitting method that fits the mean intensity in bins of E(B-V) after rejection of outliers via iterative sigma-clipping, recovering tight relations (rho = 0.993 for FUV, 0.976 for NUV). We find the best-fit relations for the FUV and NUV intensity per 6'x6' pixel: FUV = 2285.3 E(B-V) + 278.6 and NUV = 1517.6 E(B-V) + 547.2 (photons/cm^2/s/sr/A), inverted to produce a 601x601 GALEX reddening map. GALEX E(B-V) correlates strongly with HI column density from the all-sky HI4PI survey (rho = 0.991). We validate the maps using SDSS quasars as extragalactic extinction calibrators: quasar extinctions from the GALEX and Planck maps agree with a correlation of 0.99 and a slope of 0.986. Comparison with the Pan-STARRS1 stellar reddening map of Schlafly et al. (2014) shows the GALEX-based values have a smaller mean difference (0.039 mag) than the Planck-based values (0.064 mag). These results establish dust-scattered FUV diffuse emission as a viable alternative to thermal-dust extinction maps at high Galactic latitude.

The Surface Water and Ocean Topography (SWOT) mission offers unprecedented freshwater monitoring capabilities through its innovative wide-swath measurement system, which generates several data products, including the high-resolution pixel cloud (PIXC) product. However, the native PIXC water classification remains prone to systematic misclassification in urban environments, where strong radar returns from non-water surfaces are the predominant error sources. We present a deep learning approach that enhances land-water classification directly on SWOT PIXC data based on a dynamic graph convolutional neural network that simultaneously exploits spatial proximity and feature similarity to derive binary land-water class labels. The model is trained on a full year of PIXC data for the Dallas-Fort Worth metropolitan area using pixel-level ground-truth class labels derived from the DSWx-HLS product, which provides water-class labels based on Landsat and Sentinel-2 at a 30\,m spatial resolution. Against the DSWx-derived reference, the proposed method increases the mean scene-level F1 score against the DSWx-derived reference from 0.52 to 0.86 on the temporally independent test set and from 0.43 to 0.74 on the spatiotemporal test set, relative to the native PIXC classification. These results demonstrate that dynamic graph neural networks are well-suited to the irregular, point-cloud-like structure of PIXC data and offer a scalable path toward more reliable urban flood monitoring and freshwater mapping at the high spatial resolution provided by the SWOT PIXC product.

Large-scale learning programmes generate records that make learner participation observable across different activities. Participation points are commonly used to record and encourage such participation, but their value may extend beyond the activities for which points are awarded. Existing evaluations often examine gamification outcomes within the activities or learning environments in which the game elements are implemented, providing limited evidence about whether early participation points contain information about later participation outside the points system. This study examines whether early participation points can provide information about learners' later participation in a self-paced learning track that does not award participation points. Using anonymised records from 876 learners in a large-scale remote software-upskilling internship, we examined participation points generated from live-session attendance and poll responses against later self-paced course completion. The primary analysis used the 438 learners who earned at least one point during the first week, while the full cohort was retained for the no-point analysis. Week-one participation points distinguished learners who later completed a self-paced course with an AUC of 0.89, increasing to 0.95 by the fourth week. Similar AUCs were observed at both stages of the self-paced course sequence, while the absence of week-one points identified learners who did not start or did not complete a self-paced course with 95% precision. These findings indicate that early participation points can provide information about later participation outside the activities that generate the points. Such information can help large-scale learning programmes identify learners who may require timely attention while learning is still in progress, without treating participation points as a measure of overall learner engagement.

We study the consequences of membership comparators that exclude one possible membership vector, deterministically or with a relative advantage over uniform guessing. For every polynomially bounded arity, a randomized polynomial-time comparator of error at most $(1-1/poly(n))2^{-t}$ gives $ NP/ poly\cap coNP/ poly$ recognition with common advice. The proof uses limited independence, polynomial occurrence certificates, and a self-contained positive-relation advice transfer. For SAT at arity $O((\log n)^d)$, both this relative-gap hypothesis and deterministic comparability imply $PH=S^{NP}$, the uniform bound $PH\subseteq BPTIME(2^{O((\log n)^{d^2})})$, and symmetric verification with polynomial-length certificates and an oracle-free deterministic $2^{O((\log n)^d)}$ predicate. Polynomial-advice deterministic decoding has the same exponent $d$. Applying the randomized simulation to an unconditional diagonal language yields, for every fixed $\varepsilon>0$, $\mathrm{BPP}\subsetneq BPTIME(2^{O((\log n)^{d^2+\varepsilon})})$, without advice. The larger clock remains subexponential under every fixed number of self-compositions. A layered oracle satisfies deterministic comparability and $NP^O=coNP^O$ but excludes randomized NP algorithms with smaller logarithmic power, establishing a relativized limit on the SAT exponent $d$. This expanded version also develops the full weak-advantage regime, where saving $2^{-O((log n)^d)}$ gives randomized SAT exponent $d$ and PH exponent $d^k$ at fixed level $k$; the quasipolynomial and exponential hierarchy consequences; binary-comparator advice bounds; and the certificate-length boundary between the randomized regimes. Under deterministic comparability, uniform deterministic promise-unique search additionally gives $UEXP=EXP$. The ordinary second-level collapse $PH=\Sigma_2^p$ for $d>1$ remains unproved.

In 2023, AI models answered from training data and hallucinated when it ran out, and businesses were told to seed that knowledge. Models' training knowledge has since given way to live web search, and the advice followed it there: answer-engine optimization, or AEO, now tells businesses to scatter breadcrumbs across forum threads, listicles, and off-site citations, so AI engines are likelier to surface and recommend them. But being surfaced is no longer enough: an agent opens the results and reads them before deciding, and one buyer question sends it through several rounds of search and fetch. What decides the outcome at this drill-down step is whether the agent can fetch and read the business's own site: agent experience (AX). We argue that AX is the new AEO. We run 37,927 agent journeys, each a buyer question about a business, across four independent harnesses over 1,056 real businesses, matched on fame, prior model knowledge, and two AEO proxies, then split based on their AX level. Only 7-10% of the finished answer comes from the model's training knowledge, whether or not the site is readable. Agent-ready businesses have answers built from their own pages 78% of the time against 56% and are clearly recommended 1.9x more often, while a grounded answer about a not-agent-ready business costs the agent 64% more on average. Holding business, harness, and question fixed, answers built from the site are 41% more accurate on average. The dominant failure is not fabrication but omission: web-built answers are 3.7x more likely to contain none of the facts the buyer asked for. Baselines differ sharply across the four harnesses, with clear-recommendation rates varying sevenfold from stack to stack, yet the recommendation gap holds in every one. In the agentic web era, being readable beats being talked about, and improving a site's AX is the strongest lever a business has.

AI-powered lifelogging wearables are emerging as a new class of consumer devices that transform everyday experience into searchable, AI-curated memory archives. We study early public sensemaking around these systems at the moment of their market entry, using the Looki L1 as an empirical lens. Analysing large-scale Chinese-language and English-language social media discourse (N = 5,053 comments), we combine topic clustering with inductive thematic analysis to examine how users interpret the social, moral, and political implications of AI-mediated memory. Across contexts, users reference dystopian surveillance imaginaries, express privacy resignation and bystander concerns, and debate assistive value alongside consumer logics. English-language comments more often framed these devices through interpersonal power, evidentiary use, and hacking anxieties, while Chinese-language comments more often foregrounded labour exploitation, governance surveillance, and technological inevitability.

Few-shot visual anomaly detection is fundamentally a visual comparison task, requiring fine-grained inspection of a query against normal references. Many recent methods based on large vision-language models (LVLMs) emphasize comparative reasoning through language chain-of-thought. Yet discrete, abstract descriptions may underrepresent dense, fine-grained visual differences, leaving a gap between visual comparison and its expression in language. To address this gap, we propose Visual Difference DeepStack (VD-DeepStack), which explicitly conditions language reasoning on query-reference visual differences. Specifically, we fuse DINO features with the LVLM visual hierarchy to strengthen fine-grained representations, then construct dense difference evidence from residuals between query features and softly matched reference features. The difference-evidence path injects spatially weighted difference vectors into query-image states at multiple decoder depths, while an auxiliary visual-context path provides fine-grained appearance information to support their interpretation. Experiments on 4 industrial and 2 medical anomaly benchmarks demonstrate substantial improvements in few-shot anomaly detection over baselines relying on textual comparative reasoning. These results support mitigating the visual comparison-reasoning gap through the joint design of comparison representations and their integration into the decoder. Code will be released upon acceptance.

Proactive dialogue requires agents to continually adapt their policies to user feedback while progressing toward task objectives over multiple turns. To move beyond imitation learning on static datasets, recent approaches use user simulators to collect interactive data for policy optimization. However, many simulators do not explicitly model the evolution of user cognition, limiting the consistency and state dependence of feedback across turns. Moreover, representing each action only by a high-level strategy label overlooks the large utterance space and cannot distinguish alternative realizations of the same strategy. To this end, we jointly design a $\textbf{Cog}$nitive User $\textbf{Sim}$ulator $\textbf{(Cog-Sim)}$ and $\textbf{C}$ognitive-$\textbf{S}$tate $\textbf{T}$ransition–Driven $\textbf{P}$olicy $\textbf{O}$ptimization $\textbf{(CSTPO)}$. Cog-Sim maintains the user's cognitive and affective states and generates responses through constrained state transitions across turns, so feedback depends on both the realized utterance and the user's current state. CSTPO organizes each action as a hierarchical strategy–utterance representation: a high-level strategy label constrains utterance sampling, and utterances are optimized within each label. Sparse complete-branch sampling reuses shared dialogue prefixes and estimates separate strategy-level and utterance-level advantages, enabling fine-grained optimization at both levels. Across three tasks, Cog-Sim exhibits monotonic dose–response relationships and is preferred over prompt-based simulators for naturalness. CSTPO improves Qwen3-14B's performance to a level comparable to that of GPT-5.5-based planning methods.

The radio – soft $\gamma$-ray data of the Galactic black hole binary XTE J1859+226 showed that discrete jets are launched when the inner radius of the accretion disk ($R_{\rm in}$) rapidly approaches the innermost stable circular orbit (ISCO). Given the correlations among the X-ray fractional variability amplitude (rms), $R_{\rm in}$, and the X-ray hardness, previous phenomenological findings related to jet ejections (“jet line” and sharp drops of rms) also indicate that the rapid shrinkage of $R_{\rm in}$ down to the ISCO turns on the jet formation. In order to make clear how these properties appear in other objects, we investigated the X-ray and radio data of XTE J1550-564, GX 339-4, and MAXI J1820+070. We found that all the four sources show drops of $R_{\rm in}$ and rms 1-10 days ahead of a jet activity, and exhibit no huge jumps of the accretion rates prior to jet ejections. XTE J1859 and MAXI J1820 show the jet activities under sub-Eddington accretion. Models for the jet production that require an enhancement of $\dot{M}$ or a super-Eddington accretion rate are unlikely. Another behavior common to multiple sources is the curved shape of the rms–$R_{\rm in}$ correlation. Therefore, monitoring of $R_{\rm in}$ seems to be suitable to predict the timing of jet ejections. When both $R_{\rm in}$ and rms continuously decrease and when $R_{\rm in}$ is about to reach the ISCO, alerts can be circulated as a “jet forecast” for triggering Target-of-Opportunity observations.

In this work, we provide a complete characterisation of Pitt's inequality for metaplectic operators. The symplectic geometry underlying a metaplectic operator distinguishes effective directions, along which its action is Fourier-type, from singular directions, along which concentration is preserved. For this reason, we adopt two complementary perspectives, thereby obtaining both an isotropic Pitt's inequality, emulating the classical theorem of Pitt, and an anisotropic inequality that adapts to the geometric features of the metaplectic group. As a by-product, we obtain Pitt's inequality for Fourier transforms along subspaces of $\rd$ and new quantitative time-dependent boundedness results for metaplectic operators on homogeneous Sobolev spaces, with applications to Schrödinger evolutions generated by quadratic Hamiltonians. Additionally, we derive logarithmic and entropic uncertainty principles for the metaplectic group. We show that the classical entropic uncertainty principle fails when both the effective and singular directions are present. In this case, a generally unbounded corrective entropy term accounts for the concentration-preserving directions and restores a meaningful lower bound. For quadratic Schrödinger evolutions, this correction follows the time-dependent geometry of the nondispersive directions and adjusts the uncertainty estimate accordingly.

An effective field theory (EFT) of gravity with a cubic Riemann invariant coupled to an axion-like field is investigated and shown to propagate no additional independent low-energy degrees of freedom. This interaction generates a curvature-induced scalar profile around black holes, characterised by a secondary axion charge fixed by regularity of the axion field at the horizon, while the requirement of a regular event horizon constrains the asymptotic axion background. We construct the corresponding black hole solutions and show that the backreaction of the axion field could affect the metric only at subleading order in the EFT expansion. The modified geometry impacts black hole thermodynamics and Hawking evaporation, thus providing a controlled realisation of black holes with higher-curvature contributions and secondary axion hair motivated by ultraviolet physics. The modified Hawking temperature allows for the existence of black hole remnants at the end of the evaporation within the regime of validity of the EFT.

Flow-based Vision-Language-Action (VLA) policies are typically trained by behavior cloning and thus do not explicitly optimize long-term task return. Critic guidance steers generation toward higher-value actions, but existing methods differentiate the critic through a one-step surrogate of the sampler and back-propagate a critic ensemble at every flow step. In contrast, here we propose Adjoint Guidance Flow (AGF), which amortizes trajectory-aware critic guidance into a lightweight guidance network while preserving the pretrained VLA policy. Specifically, we formulate critic-guided flow generation as a deterministic optimal control problem, whose optimal guidance is a costate that carries the terminal critic gradient back through the remaining flow, and regress the guidance network onto this costate while keeping both the VLA and critic frozen. This design provides favorable memory and throughput scaling during training, and inference needs one guidance-network forward pass per step, without the critic ensemble, back-propagation, or adjoint computation. Across LIBERO, RoboCasa, and LIBERO-Pro, AGF consistently improves pretrained VLAs, remains competitive with critic-guidance and policy-fine-tuning baselines, and is the most robust method when a single guidance strength is deployed across tasks. Compared with QGF, AGF runs $3.6\times$ faster per guidance step with $7.0\times$ fewer parameters, with comparable and even better performance, showing that critic guidance can be trajectory-aware and lightweight.

Several external sources (the rings, the satellites, cometary impacts, interplanetary dust particles) supply material to the upper atmosphere of Saturn. The signature of this external influx has been documented on the ionospheric electronic and ionic densities, and in the stratosphere on oxygenated molecules (CO, CO2, H2O). We search for the chemical signature of the external sources in nitrogen-bearing species in Saturn's stratosphere. We have targeted the HCN(4-3) rotational line and mapped its emission line at the planet limb with ALMA in 2018 with configuration C43-3, providing a synthetic beam of 0.61"x0.50". HCN was detected at all observable latitudes from 20{\deg}S to 60{\deg}N. The HCN abundance was measured at two independent pressure levels, 0.1 and 1 hPa. The measured volume mixing ratios are latitudinally uniform within uncertainties at both pressure levels. The mean vmr is $(7\pm 2)\times 10^{-11}$ and $(4\pm 1)\times 10^{-12}$ at 0.1 and 1 hPa respectively. This corresponds to a mean column-density of $(4\pm 1)\times 10^{12}$ cm$^{-2}$, and a total nitrogen mass of $(3.7 \pm 0.9)\times 10^{10}$ g. The steep vertical profile suggests that another loss mechanism, possibly sticking or heterogeneous chemistry on aerosols, is active besides photodissociation. The huge influx of material detected in the ring plane by Cassini left no detectable chemical signature on HCN. The detected HCN mass seems compatible with several sources: the ring rain at mid-latitudes, Enceladus, or an old cometary impact. The detected HCN could also be a chemical product of internal N2 transported from the deep atmosphere up to the stratosphere.

The inference efficiency of Multimodal Large Language Models (MLLMs) is severely constrained by massive visual token sequences induced by high-resolution inputs, with computational cost scaling quadratically. Existing approaches primarily focus on downstream token compression, while overlooking a fundamental upstream inefficiency: input resolution is treated as a static, task-agnostic hyperparameter. We propose Task-Conditioned Resolution Routing (TCRR), which formulates visual compression as a task-conditioned decision and employs a lightweight cross-modal router that conditions backbone visual representations on textual semantics via feature-wise modulation and cross-attention to predict the minimal sufficient compression level per query. To support this, we curate a dataset of 500k samples across 12 task categories, labeled via a teacher-oracle pipeline to approximate Pareto-optimal compression scales. Extensive experiments across diverse architectures show that TCRR achieves a superior efficiency frontier, specifically reducing visual FLOPs by 40.9% and latency by 53.7% on Qwen3-VL-8B while preserving competitive performance. Further analysis of scaling behavior confirms that dynamically routing visual compression enables optimal resource allocation without modifying the MLLM backbone.

Even-layer topological antiferromagnetic thin films such as MnBi$_2$Te$_4$ exhibit an unconventional layer Hall effect (LHE) arising from a layer-locked Berry curvature. This effect manifests as a layer-polarized anomalous Hall effect that can be controlled by a vertical displacement field. However, achieving a quantized version of the LHE remains experimentally challenging. In this work, we propose that moiré engineering, using an electrostatic superlattice potential from control layers, can realize the quantum layer Hall effect (QLHE). Specifically, by placing moiré control layers in close proximity to the surfaces of a MnBi$_2$Te$_4$ thin film without breaking $PT$ symmetry, the resulting system exhibits a vanishing net Hall conductance but a quantized layer Hall conductance of $e^2/h$. A weak gate electric field then breaks $PT$ symmetry and drives the system into a layer-polarized quantum anomalous Hall (QAH) phase. Furthermore, when a single control layer is introduced on only one surface, the layer-polarized QAH emerges even in the absence of an external electric field. We identify the microscopic origin of the QLHE as an emergent layer-$U(1)$ gauge field generated by the scalar moiré potential, which in turn produces opposite periodic pseudomagnetic fields on the top and bottom layers. Our work provides a feasible pathway toward realizing the QLHE in antiferromagnetic thin films.

Long gamma-ray bursts (LGRBs) are believed to originate from the core collapse of massive Wolf-Rayet (WR) stars, leading to the formation of a spinning black hole that powers a relativistic jet. The prompt emission is produced once the jet emerges from the stellar envelope. We perform 2D axisymmetric RHD simulations using the PLUTO code to model semi-self-consistent, accretion-powered jets launched by Kerr black holes. For the two progenitor models, we adopt black hole masses of $4 M_{\odot}$ and $5 M_{\odot}$ corresponding to WR stars of $10 M_{\odot}$ and $25 M_{\odot}$, respectively. Unlike previous studies, our simulations employ continuous jet injection, with the injected power linked to black hole spin and accretion through an empirical relation motivated by GRMHD simulations, enabling a controlled exploration of jet energetics over $10^{48}$-$10^{52}\,\mathrm{erg\,s^{-1}}$. We investigate jet breakout for WR progenitors of $10\,M_{\odot}$ and $25\,M_{\odot}$ and identify a critical spin threshold: jets are choked for $a \leq 0.001$ and successfully break out for $a > 0.001$. The breakout time ($t_B$) shows a correlation with jet luminosity ($L_{\rm jet}$) and spin ($a$), revealing three regimes: Newtonian ($a \lesssim 0.03$), relativistic ($a \gtrsim 0.3$), and intermediate. A dichotomy is observed in jet head velocity at breakout, with $\beta_h \gtrsim 0.8$ for high-spin and $\beta_h \sim 0.35$-$0.8$ for lower spins. In the Newtonian regime, simulations yield $t_B \propto L_{\rm jet}^{-0.48}$, steeper than analytical predictions, with systematically longer breakout times indicating modified jet-head scaling. In the relativistic regime, calibrated analytical models show improved agreement, with simulations indicating efficient jet propagation.

Future exogenous variables provide valuable information for forecasting endogenous time series. Existing covariate-aware methods primarily learn the direct influence of exogenous variables on endogenous variables. However, these effects can be complex and change with the pattern of the exogenous variables, making them difficult to capture. Beyond this perspective, we observe that a given exogenous pattern often co-occurs with only a small set of endogenous response patterns. These associations motivate a strategy that matches future and historical exogenous patterns and uses the corresponding endogenous patterns to enhance forecasting. However, in real-world forecasting scenarios with multiple exogenous variables, each exogenous variable provides a distinct dimension for matching, creating a dilemma for this strategy between precise matching and sufficient historical support. To bridge this gap, we propose XMatch (EXogenous MATCHing), a covariate-aware forecasting model that realizes the aforementioned strategy through a tree-structured matching process that adaptively adjusts the number of exogenous variables used as matching conditions. Specifically, we first introduce the ProtoTree Creator, which organizes historical correspondences between exogenous and endogenous patterns into a ProtoTree, whose deeper levels incorporate additional exogenous variables for matching. For forecasting, we then design the ProtoTree Matcher, which uses future exogenous variables to query the ProtoTree and adaptively determines how many exogenous variables to use for matching based on exogenous pattern similarity and historical support. Finally, the matched endogenous patterns are used as explicit historical evidence to enhance forecasting. Extensive experiments on 12 real-world datasets demonstrate that XMatch outperforms state-of-the-art baselines.

Exponential stabilization is investigated for a hetero-dimensional parabolic system in which a one-dimensional shoreline subsystem, subject to a constant input delay, is bidirectionally coupled with unactuated two-dimensional bathymetric dynamics. A spectral decomposition separates the finite-dimensional critical shoreline modes from the exponentially stable residual dynamics. State feedback is designed on the critical subspace and lifted to the full shoreline space. An auxiliary transport equation represents the delay, while a predictor removes the explicit input delay from the nominal predictor dynamics. Exponential stability of the full coupled system is established by combining shoreline stabilization and bathymetric coercivity using weighted input–output estimates and a small-gain condition on the hetero-dimensional interaction. Numerical simulations show that uncompensated delayed feedback can strongly amplify both shoreline and bathymetric responses, whereas predictor compensation restores decay of the critical shoreline dynamics and prevents delay-induced amplification from propagating to the unactuated bathymetric subsystem.

We investigate the twist-4 generalized transverse momentum dependent parton distributions (GTMDs) of $\bar{u}$ and $\bar{d}$ quarks inside the proton at zero skewness by adopting the overlap representation within the light-cone formalism. Using the light-cone wave functions of the proton derived from the baryon-meson fluctuation model in terms of the $|q\bar{q}B\rangle$ Fock states, we calculate the twist-4 GTMDs of $\bar{u}$ and $\bar{d}$ quarks. Additionally, the intricate relations between the GTMDs and the generalized parton distributions are also explored in our research. Numerical results for these twist-4 GTMDs and twist-4 GPDs at zero skewness are analyzed and presented.

It is unclear how Neural Language Models (NLMs) acquire the structural meaning encoded by grammatical structures that is independent of lexical semantics. We propose a statistical learning process in which learned dependency structures themselves become new distributional units for subsequent statistical learning. Under this account, once a dependency structure is acquired, the model tracks its contextual distributions. These contextual features reflect the semantic properties of a composite structure. To test this hypothesis, we design a synthetic grammar in which each grammatical structure has distinct contextual distributions that cannot be recovered from the distributional statistics of their component tokens alone. We train a series of BERT-style masked language models on this grammar and examine their developmental trajectory. The results show that models can successfully learn the contextual distributions of composite dependency structures even though they cannot be inferred from token statistics alone. Developmental analysis further reveals a clear developmental trajectory. The learning of the dependency relations that define a grammatical structure consistently precedes the learning of its contextual features. These findings suggest that statistical learning in NLMs is not merely the accumulation of token co-occurrence statistics, but a process in which learned dependency structures become new units of distributional learning. We argue that this process provides a statistical-learning account of how NLMs solve the compositionality problem in language. Finally, we discuss the possibility that this statistical learning process provides an explanatory theory on how language cognition could emerge from pure distributional statistics.

Classical simulation of quantum random access memory (QRAM) under noise can be accelerated dramatically by branch pruning: noise histories are sampled as trajectory ensembles, and good branches, whose routing paths avoid every sampled fault, need not be evolved because their final states are analytically predictable. Under amplitude damping this predictability is not automatic, because the no-jump operator $K_0$ acts on every branch at every time slice. Here we give a complete account of damping-channel predictability in bucket-brigade QRAM simulation, for both qutrit and qubit encodings. For the qutrit encoding, $K_0$ is diagonal in the node basis, and the wait state $|W\rangle$ is its fixed point, so a good branch follows the noiseless orbit up to a scalar attenuation. In the qubit encoding, phase-kickback memory fetching between two Hadamard walls makes $K_0$ no longer diagonal, yet the resulting structure is exactly solvable in closed form. Conditional on no jump, the in-branch infidelity of a good branch is second order in the damping rate $\gamma$, with coherent leakage $\sim n^2\gamma^2/4$ per data qubit, and the closed form captures it without approximation. We also revise the pruning criterion for the qubit encoding. The resulting algorithm evolves only the bad branches plus one reference branch and shares every other cost with the full evolution it replaces. It is validated by end-to-end benchmarks with pruned-over-full speedups up to $285\times$ and by trajectory-level tests confirming every closed form to machine precision.

Recent 3D generation models can produce accurate geometries while still struggling to reconstruct detailed textures. We propose a diffusion-based native 3D material generation model TaoTex, which faithfully recovers intricate textures through tailored strategies and improvements. First, we develop a data construction agent to create high-frequency textured 3D assets to bridge the data gap in public datasets. Training with these data significantly enhances the ability of TaoTex to recover challenging details such as text and patterns. Second, we design a multi-level feature fusion (MLFF) module to adaptively integrate local and global features of the conditional input, providing more complete texture cues for the diffusion model and thereby enhancing reconstruction fidelity. To alleviate VAE reconstruction errors, we adopt a latent-to-pixel space loss transition, further improving the pixel-level details and generation quality. Finally, we scale TaoTex to multi-view inputs by incorporating learnable viewpoint embeddings, achieving accurate and consistent material reconstruction across views. Extensive experiments demonstrate that our method significantly outperforms existing approaches in preserving texture details in both single- and multi-view settings.

Memorization has been proposed as a mechanism to explain how language models fit the tail of their training distributions, but its training dynamics are not understood well. In this work, we take a fine-grained look at memorization by decomposing the loss trajectory of memorized sequences over training and model parameters. Across the Pythia family, we study memorization of duplicated training sequences (recitation) and rare ones (recollection). We find that memorization in both cases is characterized by sequence-level gradient alignment, though recitation suffers from misalignment with other training influences which causes forgetting, explaining the necessity for higher duplication of these examples. We further show that the lower model layers are the most involved in memorization and forgetting. Predicting memorization, our decomposition improves over a cross-entropy baseline, especially in larger models and early in training. Intervening on a small set of highly influential parameters we are able to ablate memorization in the final model. Together, these findings advance our understanding of how memorization develops during training and offer insights for predicting and intervening on it.

Planetesimal formation via the streaming instability of small pebbles in protoplanetary discs requires enhanced solid concentrations relative to the solar metallicity, $Z\simeq 0.01$. We investigate here whether pile-ups in weak axisymmetric pressure bumps are sufficient to trigger strong particle concentrations via the streaming instability. Using high-resolution 2D shearing box simulations with millimetre-to-centimetre pebbles, we explore the behaviour of the streaming instability in the presence of a non-reinforced pressure bump. We find that even very weak bumps, with gas density amplitudes as low as $A = 0.04$ relative to the background, produce dense particle filaments via the streaming instability for all tested Stokes numbers at solar metallicity in the inner disc. Outer disc regions require slightly stronger bumps ($A \geq 0.14$) to form filaments at a solar metallicity, though the necessary bump amplitude is substantially lowered when the metallicity is increased to $Z = 0.02$. Our results suggest that weak, non-reinforced pressure bumps can act as focal points for planetesimal formation via the streaming instability in small pebbles, in contrast to previous studies which used reinforcement to maintain the pressure bump. Additionally, we introduce a pressure-bump-dependent clumping criterion, $(Z/\chi)_\mathrm{crit} \approx 0.3$, where $\chi = \Pi_\mathrm{min}^2/\Pi_0$ reduces to the background pressure gradient $\Pi_0$ in the absence of a pressure bump. This criterion encapsulates the scale of solid pile-ups at lower pressure gradients and accurately predicts the onset of strong clumping based on the results of our simulations. With pressure bump signatures being common features of observed young discs and magnetohydrodynamical simulations, the results of our 2D simulations imply that weak pressure bumps may be major cradles for planetesimal formation in protoplanetary discs.

Machine learning has made strong progress on music tasks, both as assistive tools and as creative partners. However, most systems train on multitrack corpora that emphasize pop and rock. Jazz, with improvisation at the core of its practice, still lacks a well-annotated corpus of clean per-stem combo recordings on standards. We introduce JazzSAMBA (Jazz Synchronous and Asynchronous Multi-take Band Audio) to fill this gap: the first originally recorded jazz-combo multitrack dataset of standards with asynchronous (overdubbed) and synchronous (live ensemble) protocols, preferred and alternate takes chosen by the musicians, and timed annotations for bars, chords, sections, and soloists. JazzSAMBA covers 76 standards by eight musicians on drums, bass, piano, trumpet, and saxophone, with per-stem audio, mixtures, and MIDI. It can support chart-conditioned accompaniment, combo source separation, and form-aware music information retrieval. We demonstrate the dataset on two tasks: a jazz combo source-separation baseline and a chart-conditioned accompaniment ablation. The dataset, code, and samples are linked from the project demo page.

Large language model (LLM) agents are evolving from tool-calling systems that execute isolated instructions into task-oriented agents that pursue user goals through sustained, multi-step interactions. However, existing benchmarks for personalized tool use largely assess isolated calls or reactive execution, leaving unclear whether agents can formulate, execute, and revise an explicit plan while preserving user preferences throughout long-term interaction. To address this gap, we introduce PDEU-Bench (Personalized plan Definition, plan Execution, and plan Update Benchmark), a benchmark for evaluating the complete planning lifecycle of personalized tool-using agents. PDEU-Bench comprises 214 long-horizon interaction tasks spanning 12 everyday domains and 94 tools, with stage-specific assessments of preference adherence and plan quality. Extensive evaluations of 15 representative open-source and closed-source LLMs reveal a pronounced gap between local tool execution and dynamic planning: LLMs can often instantiate preferences in individual calls, yet struggle to construct coherent plan definition and plan update. We further evaluate mainstream personalization and memory-augmentation methods. Although these methods improve particular stages, none of the evaluated methods reliably propagates user preferences throughout the complete lifecycle, and their gains frequently fail to transfer to subsequent execution. Fine-grained error analysis further reveals that preference omissions and conflicts persist throughout the planning lifecycle, highlighting the need for future research to parameterize LLMs with preference-aware information retrieval and memory capabilities. We provide the relevant code and data in the appendix to support future research.

Language models are increasingly used to sample from a specified distribution, for instance, to simulate survey respondents or generate synthetic data. Instruction-tuned models can state such a distribution correctly and still fail to sample from it. Prompting and changes to decoding reduce this mismatch only partly, which motivates training with policy optimization. Group relative policy optimization (GRPO) is a natural fit for this problem because it already samples a group of rollouts per prompt, and the group's empirical distribution can be compared with the target. However, scoring the group as a whole gives every rollout the same reward. Group-relative centering then sets all advantages to zero, and the model receives no learning signal. To give each rollout its own signal, we introduce the witness advantage, a per-rollout advantage derived from maximum mean discrepancy (MMD). It trains a model to match a target distribution over a finite set of outcomes. The MMD between the model's distribution and the target has a witness function that measures how over- or under-produced each outcome is. Each rollout's advantage estimates the negative witness at its outcome, so a rollout is rewarded for an outcome the group under-produces and penalized for one it over-produces. The witness advantage is computed in closed form from the group's outcome counts, and we use it as the reward in GRPO. On unseen target distributions, training with the witness advantage substantially reduces the total variation distance to the target while largely preserving the model's general capabilities.

Structural analysis is a core method in Petri Net (PN) research, offering a perspective complementary to state-space techniques while avoiding many of their limitations. It is well studied for classical PNs but far less for High-Level Petri Nets (HLPN). Symmetric Nets (SN), a common HLPN formalism, use compact annotations to encode behavioral symmetries, enabling the construction of a symbolic reachability graph (and a lumped Markov chain in stochastic SN) and the execution of symbolic discrete-event simulations. During the past two decades, structural techniques tailored to SN have been developed, notably supported by the SNexpression tool. This tool implements a formal calculus designed for the computation of symbolic structural relations, including, but not limited to, conflict relations and causal dependencies. Here, we focus on using this calculus to verify semi-automatically symbolic structural invariants, a task currently feasible only for certain restricted SN subclasses. We focus specifically on (semi)flows and briefly discuss an approach through which a flow generative family can be generated, at least theoretically. We further briefly outline a framework for the formal verification of a broader class of invariant properties. An extended formulation of the SN formalism is employed, which demonstrably satisfies the closure property with respect to fundamental functional operators. The core concepts are elucidated by means of representative examples throughout the exposition.

We prove that the state space of every well-formed session type, quotiented by strongly connected components, forms a bounded lattice; n-ary parallel composition yields product lattices. Two consequences follow: duality preserves the lattice up to isomorphism, and Gay-Hole width subtyping corresponds to lattice embedding for non-recursive types. We validate this on 108 benchmark protocols across networking, databases, distributed systems, AI, and fault tolerance: all form lattices, 93 distributive, 15 non-distributive. Mechanised in Lean 4 with two independently developed tool implementations.