Papers

The Efimov effect manifests discrete scale invariance through a geometric series of three-body bound states. Analogous discrete scale symmetry has been observed in the Efimovian expansion of scale-invariant strongly interacting Fermi gases. However, it is unclear whether strong correlation is necessary. Here we investigate the spontaneous breaking of continuous scale invariance in a simple driven-dissipative harmonic oscillator whose driving and dissipation parameters scale as $1/t$. By transforming to logarithmic time, we show that the dynamics is governed by a non-Hermitian superoperator whose spectral decomposition reveals the emergence of a pair of eigenmodes with opposite real parts and equal imaginary parts. This signals the spontaneous breaking of continuous scale invariance and gives rise to Efimovian-like oscillations: log-periodic decay oscillations obeying discrete scaling laws. The onset of oscillations coincides with a PT-symmetry breaking transition in logarithmic time. The Efimovian-like oscillations are robust against weak nonlinear interactions and perturbations. Our work implies that the spontaneous breaking of continuous temporal scale symmetry and the emergence of Efimovian-like dynamics can even be observed in a simple driven-dissipative system without many-body interactions.

Asynchronous communication is a fundamental feature of modern distributed systems, where messages are emitted without requiring immediate synchronization with receivers. In process calculi, this behaviour is typically modelled by separating message emission from message consumption. At the same time, reversible computation has emerged as an important paradigm for analysing concurrent systems, enabling computations to be undone while preserving causal dependencies between actions. While reversible semantics have been extensively studied for synchronous process calculi such as CCS, their integration with asynchronous communication remains largely unexplored. In this paper we investigate the interaction between asynchrony and reversibility in the setting of CCS. We first introduce CCSa, an asynchronous variant of CCS in which output actions generate explicit message entities that can later be consumed by matching input actions. We then define rCCSa, a reversible extension of CCSa obtained by adapting the framework of Phillips and Ulidowski. In rCCSa, prefixes and messages are annotated with unique keys that record message emission and consumption events, allowing computations to be reversed while preserving causal dependencies. We show that the resulting reversible semantics satisfies causal consistency, ensuring that computations can be reversed exactly up to causal equivalence. The proof relies on the axiomatic framework for reversible computation proposed by Lanese et al.

An auditor who checks whether a system's weights are frozen is checking the wrong thing. Our stationarity dichotomy says that iterative self-modification hits strict diminishing returns whenever the agent's reachable set of edits stays fixed, and can escape only if that set expands. Rewriting scaffolding (tools, verifiers, decomposition) expands what an agent reaches without touching a weight, so frozen weights buy an eventual ceiling but no stationarity along the way. The criterion also separates three regimes usually merged: search within a fixed class, test-time training that raises the ceiling itself, and scaffold rewriting between them. Audit the scaffold, not the checkpoint. The same ceiling binds sideways. Best-of-$k$ orchestration realizes the best worker's ceiling exactly: width buys rate, not budget. Re-consulting a fixed pool has a horizon computable in advance, decided by the pool alone, and the one arrangement that would beat it, a weighted vote, needs diversity real workers lack: on 30 same-family workers the failure overlap sits at its maximum, and a majority fails 23/55 (42%) of tasks. We obtain the criterion by reading refinement as gradient boosting on the residual error between draft and target, a patch or git diff, and then measuring where that reading breaks: patches compose instead of standing beside each other to be voted on, and failures overlap. What we measure is saturation. Per-round improvement decays toward zero on SWE-bench, and churn decays geometrically across 401 production sessions, a shape shared with a pre-AI human baseline that establishes the regime without identifying its cause. Both breaks are engineering choices rather than laws about code, so together they specify a harness worth building.

A major problem in geometric analysis is to understand the behaviour of the Lagrangian mean curvature flow. This has proved to be a challenging problem, in particular, not many explicit examples of flows that exist for all time and converge to a minimal Lagrangian are known. In this paper we construct new examples of Lagrangian mean curvature flows in the Kummer K3 surface that converge to special Lagrangian spheres. We reduce the construction to a scalar perturbation problem that can be solved by a fixed point argument.

The Participants-as-Interfaces (PaI) methodology for system composition proposes that system participants can be regarded as interfaces. For each system in a given collection, one participant is designated to serve as its interface. When the systems are composed, these interface participants are replaced with gateways that communicate with one another by forwarding messages. We generalise the approach to partial gateways, where gateways can forward only a chosen set of messages. As for the standard PaI approach, we exploit such extended version for systems of communicating finite state machines (CFSMs). This extension is fully detailed for the binary case, and can be scaled up to the multicomposition case. We prove that many relevant communication properties (deadlock-freeness, reception-error-freeness, etc.) are preserved by PaI composition via partial gateways in case also the connection policy (i.e. the system representing the way we wish gateways interact with each other) enjoys the same properties. Such a proof turns out to be just a corollary of a property preservation result for a restricted and partial composition method, dubbed fusion-composition. Fusion-composition hence turns out to be at the heart of the PaI composition approach.

Predicting Drug-Target Interactions~(DTIs) is a central task in computational drug discovery, with direct applications in virtual screening, drug repurposing, and therapeutic candidate prioritization. Although recent deep learning methods have improved DTI prediction, many sequence-based models still process drugs and proteins independently and only combine their representations at a late prediction stage. This limits their ability to explicitly model cross-molecular dependencies between chemical substructures and protein sequence regions. In this paper, we propose a sequence-only DTI prediction architecture that combines two pre-trained language models, ChemBERTa for drug SMILES strings and ESM-2 for protein amino acid sequences, with a hierarchical interaction module. The proposed model first extracts contextual representations using pre-trained encoders, then applies 1D convolutional layers to condense local sequence patterns, followed by a sequential bidirectional cross-attention mechanism inspired by the induced-fit view of molecular recognition. Finally, attention-based pooling constructs fixed-size interaction-aware vectors for binary prediction. Experiments on BIOSNAP, Davis, and BindingDB show that the proposed model achieves the best performance on BIOSNAP, matches the best AUROC on Davis, and remains competitive on BindingDB while using only 25.2 million trainable parameters. Ablation results confirm the contribution of both the CNN and cross-attention modules, and cold-start experiments indicate promising generalization to unseen proteins and drugs.

On modern production text-to-image systems, successful policy violations are rare, and previously effective human-written seeds are often patched out. Current automated red-teamers are poorly matched to this regime in two ways: unreliable success measurement and poor exploration. First, we find that judges widely used in prior T2I red-teaming work are unreliable under vague unsafe-content targets: they either miss true violations or reward benign borderline images on hardened APIs. We therefore define strict category-specific success criteria and calibrate strong VLM judges against human labels. Second, we show that broadly used prompt-modification pipelines do not solve the exploration problem: on harder guardrail settings they remain tied to seed prompts, fail to transfer, or cannot bootstrap positive examples. We introduce RISE, which evolves reusable strategies used to generate prompts rather than rewriting them one by one. The best discovered strategies are then reused to generate attacks across new scenarios. On DALL-E 3, Nano Banana 2 (Google) and GPT-Image-2, RISE reaches up to 13% human-verified ASR; under the same calibrated evaluation, prior methods with reported ASR as high as roughly 30% fall to near zero.

The short wavelength of terahertz (THz) waves makes the acceleration of electrons, from typical electron gun energies to fully relativistic energies, challenging due to the need to match the velocities over comparably long distances. However, the inherent laser synchronisation of a photogun to the THz drive laser offers significant advantage in applications where timing is critical. We present a novel design process for high-gradient THz-driven dielectric-lined waveguide injectors and demonstrate designs for rectangular and cylindrical geometries. These structures utilize novel tapering and stepping schemes to manipulate the phase velocity, controlling the position of the electron bunch relative to the phase of the accelerating field. We demonstrate a hybrid optimisation method utilizing firstly a multi-objective genetic algorithm based on analytic models of the waveguide accelerating modes, and then high-detail particle-in-cell simulations. Our example designs achieve electron acceleration from 100 keV to 1 MeV over a distance of 20 mm by interacting with a multicycle 0.5 mJ, 0.2 THz pulse, paving the way for the realisation of THz-based injectors. The exit beam shows excellent beam quality with 50 fs bunch lengths, sub 0.5% energy spread and emittances of under 0.15 {\mu}rad. We also present an analysis of the robustness of the designs to errors in machining and operational parameters demonstrating the feasibility of the concept. We show that using multi-objective genetic algorithm optimization and robust design process, we can achieve high-quality 1 MeV beams.

For functions that are convex of higher order, classical extremalities identify Gaussian quadrature as a lower bound and Gauss–Lobatto quadrature as an upper bound for the integral. We sharpen this bracket in every odd order: if $f$ is $(2n-1)$-convex on $[-1,1]$, then the exact integral $I[f]:=\int_{-1}^{1}f(x)\,dx$ lies not merely between the $n$-point Gauss–Legendre rule $G_n[f]$ and the $(n+1)$-point Gauss–Lobatto rule $L_{n+1}[f]$, but already between $G_n[f]$ and their midpoint. This yields the certified estimator $Q_n[f]:=\frac{3}{4}G_n[f]+\frac{1}{4}L_{n+1}[f]$, with $|I[f]-Q_n[f]|\leqslant\frac{1}{4}|L_{n+1}[f]-G_n[f]|$, and the constant $\frac{1}{4}$ is sharp. We also record a complementary phenomenon in even order: for $2n$-convex functions the Gauss–Radau endpoint rules still bracket the integral, but the midpoint of the two Radau rules cannot yield a one-sided refinement. The obstruction is explained by an odd, sign-changing Radau quadrature kernel. Finally, we distinguish numerical certification from ordinary high-order approximation: we include concrete examples from spline, moment and statistical models and report experiments for truncated powers and near-pole Stieltjes kernels, where the shape certificate remains meaningful when classical smoothness-based error constants are unavailable or severely pessimistic.

We investigate dynamical dark energy emerging from quantum gravity within the framework of group field theory (GFT) condensate cosmology. We generalise previous models of quantum gravity induced cosmic acceleration by considering a broader parameter space, including an additional scalar matter field besides the relational clock, and accounting simultaneously for the phase dynamics and the contribution of multiple condensate modes. Working in a mean-field approximation with coherent peaked states and restricting to Hermitian GFT interactions, we derive the corresponding effective cosmological dynamics and analyse its late-time behaviour. We show that the system approaches an asymptotic de Sitter regime, dynamically dominated by a single condensate mode, while the approach to this regime can exhibit a non-trivial dynamical dark-energy evolution. In particular, both the condensate phase and subdominant modes can generate deviations from the asymptotic equation of state $w=-1$, including phantom behaviour and, under suitable conditions, a crossing of the phantom divide. The additional scalar matter field enters directly into these conditions, providing a link between the matter content and the quantum gravity induced dark-energy dynamics. These results extend the robustness and generality of the GFT mechanism for dynamical dark energy and provide a basis for confronting the underlying quantum gravity dynamics with cosmological observations.

Accurate prediction of avalanche motion is essential for hazard assessment in mountainous terrain. This study develops and evaluates a physics-informed neural network (PINN) framework for the Savage-Hutter model of depth-averaged granular flow, progressing from 1D analytical verification to 2D experimental validation. First, three 1D problems of increasing complexity were verified against the analytical solution: height prediction with prescribed velocity, velocity prediction with prescribed height, and coupled prediction of both fields using the conservative formulation. The decoupled tests accurately reconstructed the spatio-temporal evolution of each field when the other was prescribed. The coupled formulation learned both fields without prescribed data, achieving mean height and velocity RMSEs of 0.043 and 0.079 in non-dimensional units. A hyperparameter sensitivity study evaluated the effects of network depth, width, collocation density, learning rate, and epochs. The framework was then extended to 2D and validated against laboratory experiments of a cylindrical granular pile collapsing on an inclined plane, with TITAN2D providing numerical comparisons. Purely physics-based training converged to the trivial zero solution; augmenting the loss with 10 sparse training points from final deposit profiles produced a physics-informed, data-assisted hybrid framework. Peak flow depth, depth-averaged velocity, RMSE, and wetted-area IoU evaluated global and local agreement. Global height RMSE ranged from 2.7 to 6.7 mm across four experimental cases, while mean wetted-area IoU ranged from 69 to 81 %, demonstrating consistent performance across variations in pile mass and slope angle.

Muon is increasingly used for language-model pretraining, yet its large-step dynamics are not captured by the classical edge-of-stability (EoS) picture of gradient descent (GD). In GD, loss neutrality, equal-magnitude update reversal, and marginal stability meet at a single learning-rate-dependent edge. We show that Muon breaks this coupling. For stochastic no-momentum Muon, we derive a coherence-corrected conditional loss-neutral boundary $2\rho_b/\eta$, while temporal alignment follows a separate geometry. Controlled experiments show that loss balance and temporal alignment respond differently to learning rate and batch size. Across our language model experiments, the 130M Llama-like LLM runs exhibit loss-boundary tracking with weak negative alignment, whereas the studied 1B LLM configuration shows stronger partial cancellation; in both settings, directions remain far from coherent reversal while training continues to improve. These results support a split EoS picture for Muon: a stochastic loss-neutral edge survives, but it is not accompanied by a universal temporal-direction signature. The source code for reproducing the experiments can be found in https://github.com/cyzebra/Muon-Sublates-the-Edge-of-Stability-in-LLM-Pretraining

We develop a rational-function approximation for the structural properties of multicomponent fluids with narrow square-well and square-shoulder interactions, together with weakly nonadditive hard-sphere mixtures. The theory is obtained by reformulating the analytical Percus–Yevick solution for additive sticky hard-sphere mixtures so as to account for finite interaction widths, leading to a unified semianalytical framework for these classes of fluids. The approximation satisfies the exact low-density limit and preserves the continuity of the cavity functions; a simple local correction restores the continuity of their first derivatives. Comparison with Monte Carlo simulations for binary mixtures shows that the theory accurately describes both the radial distribution functions and their Fourier transforms for narrow wells and shoulders and for weak nonadditivity. The proposed formulation extends previous sticky-hard-sphere-based approaches while retaining their analytical simplicity.

Tool-using language-model agents can review enterprise projects as governance boards do: they read the evidence, apply written rules and decide whether the project may proceed. Part of that evidence comes from suppliers and project members with a stake in the decision. DGF-Bench is a benchmark in which a board of agents (specialist gates and a General gate that consolidates their decisions) reviews synthetic dossiers while an attacker plants deceptive content in evidence the organization does not vouch for. Dossiers are generated from canonical facts under 61 executable rules, with 42 authoritative records and 32 narrative documents; every gate is certified decidable from those records. Attacks never change an authoritative value, so an attacked dossier keeps the reference decisions of its clean copy. A success is attributable only when the agent receives the injection and takes the exact injected action, which it does not take on the paired clean dossier; the DGF score is the share of applicable fixed attacks a model blocks. Reading documents and records themselves, five of six models were outcome-strict (disposition, findings, actions and authorization all correct) on 82 to 85 of 85 gates. Over 2,622 attacked gate runs, seven direct-order, false-data and false-authority attacks obtained one attributable success against these five, whereas task-aligned attacks imitating the organization's own process passed against four of them: a record note citing a fake review procedure lowered GPT-6 Luna Pro from 34 to 6 outcome-strict gates and DeepSeek V4 Pro from 33 to 7. DGF scores ranged from 96.2 to 26.9, and a policy-aware adaptive attacker writing in records succeeded against five of six models. The approval tool executed no forged approval, yet deceived agents submitted approvals that the rules forbid. The open-source package dgf-bench computes the DGF score with one command.

The idea of engaging critically and creatively with physics theories is a stretch for many physicists and physics educators. Most physicists become enculturated into the belief that physics definitions, laws, and equations are read from the book of nature. However, this belief obscures how theories are collectively negotiated by people in specific socio-political and personal conditions, and how students' particular ideas and intuitions will inevitably shape their own practice of science. To help students build their skills for collective theory-building, we suggest a pedagogy for physics in which students actively negotiate the laws and definitions of theories.

Modern robot policies predict a chunk of future actions from a single observation, execute only a prefix, and discard the rest before replanning. Choosing the length of this prefix, the execution horizon, poses a trade-off between reactivity and efficiency. A short horizon keeps the policy reactive to the environment, but requires frequent policy calls. Recent test-time methods adaptively select the horizon for each chunk, but they either read model internals, where the signal must be chosen for each architecture, or draw extra samples, which adds cost. We propose Action Upcycling, a training-free algorithm that reuses actions the policy would otherwise discard, without accessing model internals or drawing extra samples. We find that discarded actions stay close to their replanned versions as long as the action velocity remains smooth. Action Upcycling therefore extends the execution horizon up to the point where the velocity begins to fluctuate. Extensive experiments on simulated and real-world manipulation tasks show that Action Upcycling reduces policy calls by 1.2-1.7x with no loss in success rate, across multiple Vision-Language-Action Models (VLAs) and even a World Action Model (WAM). It applies to any chunked policy at negligible cost and is orthogonal to other policy acceleration methods such as few-step sampling and streaming action decoding, opening a new axis for policy acceleration.

The generalized packing–covering conjecture of Elimelech, Firer and Schwartz asserts that, for every linear code $\mathcal{C}$ and every admissible order $t$, the $t$-th generalized Hamming weight $d_t(\mathcal{C})$ and the $t$-th generalized covering radius $R_t(\mathcal{C})$ satisfy $d_t(\mathcal{C})\le 2R_t(\mathcal{C})+2$. We give a computer-assisted proof of the conjecture for every linear code over every finite field and every admissible order. Combining a parity-check reformulation of the conjecture, bounds on the length of putative counterexamples, and successive puncturing arguments, we settle all orders $t\ge 32$ and reduce the remaining orders to finitely many parameter tuples, which we exclude by an exact computer verification.

We develop a holographic dual to the integrable field theories built from $4d$ Chern-Simons theory. The dual theory is a topological string on a generalized Calabi-Yau manifold, which is entirely encoded in a geometric object we call a planar spectral curve. We describe the RG flow for a planar integrable field theory as a geometric flow on the moduli of planar spectral curves. This description is valid at all orders in the 't Hooft coupling. We compute this flow explicitly up to two loops in many examples, and find an exact match with known field theory results. At one loop this includes a wide class of models. At two loops we reproduce the RG flow for the principal chiral model with WZ term, and for a class of integrable $\sigma$-models with target $U(N) \times U(N)$.

We employ a projection operator technique to study the equilibrium phases and the non-equilibrium quench and ramp dynamics of ultracold bosons described by a two-dimensional (2D) Bose-Hubbard model coupled to a high-finesse cavity with competing local short-range and cavity-mediated long-range interactions. Our analysis relies on a systematic elimination of high-energy virtual hopping processes; this enables us to capture the effects of short-range quantum fluctuations on the phase diagram of the model while treating the cavity-induced long-range interaction within mean-field theory. We find that the phase boundaries between the superfluid (SF), supersolid (SS), charge-density-wave (CDW) and Mott insulating (MI) phases are significantly modified by the presence of these fluctuations. We use this approach to study the dynamics of bosons following a quench/ramp from the CDW to the SS and SF phases; our analysis indicates the presence of oscillatory dynamics characterized by a single frequency due to a quench which takes the system near the critical point. In contrast, for a quench deep inside the SF phase, the dynamics involves multiple frequencies. An explicit comparison with Gutzwiller mean-field theory shows the importance of quantum fluctuations in shaping such post-quench dynamics; this feature may be tested experimentally as we discuss. We also study the ramp dynamics of these bosons and identify a bound on the ramp rate above which the present method is expected to produce accurate results.

Bathroom acoustic-event recognition can support ambient assisted living in settings where continuous video monitoring is undesirable. However, practical deployment requires models that are compact, interpretable, and robust to changes in the recording environment. This work introduces \dataset{}, a seven-class bathroom acoustic-event dataset containing 21{,}387 annotated clips recorded across five environments, and proposes SincDPNet, a compact raw-waveform classifier with a learnable sinc filter bank followed by a depthwise-separable convolutional body. Each sinc filter is controlled by two frequency parameters, allowing the learned passbands to be inspected directly in hertz while keeping the front end small. To reduce room-specific leakage, recording sessions and environments are separated before overlapping windows are assigned to the training, validation, and test partitions. We further use multi-objective Bayesian optimization as a design tool to examine the validation performance–model-size trade-off across 24 configurations. The selected designs span different operating points: the best-performing model achieves 80.2% accuracy and 0.760 macro-F1 with 14{,}040 parameters, while the compact $N_f=25$ configuration uses only 2{,}848 parameters and achieves 75.7% accuracy, 0.661 macro-F1, and 0.716 MCC on the held-out environment. Analysis of the learned filters and confusion patterns shows that spectral overlap contributes to confusion among water-related events, while the Door/Walker/Crutch errors also reflect similarities in their transient temporal structure.

We establish noncommutative maximal inequalities for averages over submanifolds in the $L^p$-spaces associated with semifinite von Neumann algebras. For every $1<p\leq\infty$, we prove local maximal bounds for smooth submanifolds of finite type at the parameter origin, and global bounds over all positive scales for polynomial parametrizations, without a finite-type assumption and with constants depending only on $p$, the degree and the dimensions. For variable hypersurfaces in $\mathbb{R}^n$, $n\geq3$, satisfying a uniform rotational curvature condition, we obtain local maximal bounds for $p>n/(n-1)$. To our knowledge, these are the first such noncommutative inequalities for general lower-dimensional submanifolds and polynomial images. The key step is a weak type $(1,1)$ estimate for a regularized auxiliary family adapted to non-isotropic dilations. The geometric kernels need not satisfy the pointwise smoothness or $L^2$-Dini conditions required by existing noncommutative Calderón–Zygmund theories. We combine a Calderón–Zygmund decomposition based on Cuculescu projections with integral regularity estimates, a non-isotropic Littlewood–Paley decomposition and Fourier decay to prove this endpoint estimate. Marcinkiewicz interpolation yields maximal $L^p$ bounds for the regularized averages, and interpolation with Fourier-based $L^2$ bounds recovers the original averages. The variable-hypersurface result uses oscillatory $TT^*$ estimates and localization. As applications, we obtain maximal ergodic inequalities for trace-preserving actions of $\mathbb{R}^n$ and, under additional assumptions, bilaterally almost uniform convergence results for the associated normalized averages.

Power sampling has emerged as a training-free approach to LLM reasoning, eliciting capabilities comparable to reinforcement learning by sharpening the model distribution over complete responses. Despite this success, power sampling remains underexplored in large vision-language models (LVLMs). We transfer Power-SMC to LVLM decoding by defining a sequence-power target conditioned on both the image and the prompt. This direct transfer provides a strong training-free baseline, but leaves two aspects of finite-particle multimodal inference unaddressed. At the particle level, global resampling can collapse genealogies, while particle-based power sampling does not diversify trajectories through distinct visual cues in multimodal decoding, limiting exploration under a finite particle budget. At the answer level, sequence-level sharpening makes distinct reasoning trajectories compete even when they support the same answer. We introduce ReSight-SMC, a verifier-free two-stage power sampler for LVLM inference. Its first stage uses ancestry-isolated SMC islands to preserve independent trajectory families and routes a bounded set of prefix-conditioned visual scouts to prefix-relevant image regions while discouraging redundant overlap. Each scout temporarily increases attention to the image tokens and emphasizes its routed region. Exact importance correction preserves the base LVLM sequence-power target. The second stage aggregates terminal importance mass by canonical answer, powers the answer marginal, and samples an answer together with a supporting trajectory. Across four LVLM backbones and five benchmarks, ReSight-SMC achieves stronger aggregate performance than Power-SMC over both the reasoning and perception benchmark groups. Without post-training, it remains competitive in aggregate with backbone-matched models trained using reinforcement learning.

Graph neural networks (GNNs) need to exploit improved message passing without surrendering control over predictions already trusted in deployment. We introduce Reference-Tail Trust (RTT), a framework that admits learned updates inside a frozen GNN and certifies the prediction actually served. RTT couples graph-based proposal states with a constrained internal optimizer: each displacement is charged for its worst-case terminal cross-entropy increase through the incumbent's remaining message-passing layers. A trajectory-validated tube and an independent checker enforce per-node probability floors, $p^{\mathrm{s}}_{ic} \ge e^{-H_{\mathrm{row}}} p^{\mathrm{r}}_{ic}$, and a call-level budget, $\sum_i w_i D_\infty(p^{\mathrm{r}}_i \| p^{\mathrm{s}}_i) \le H^+$, uniformly over labels. Calls whose adapted outputs pass certification require no separate full incumbent rollout; failed certificates trigger whole-call fallback. We derive the exact probability-floor frontier by water-filling, characterize architecture-constrained efficiency, and establish conditions under which internal propagation exploits evidence unavailable to restricted output correctors. In the reported ogbn-arxiv audit, RTT achieves $6.5\times 10^{-3}$ nats of mean gain per call, with a one-sided 95% regression-rate upper bound of 0.95% and a 95% negative-flip upper bound of 0.51% on the uninspected part of the reserved node population. Its mean gain is 61% of a cross-fitted posterior-based frontier estimate and exceeds the strongest matched one-pass corrector by $+0.9\times 10^{-3}$ nats. Reported experiments span eight proposals, six graph-incumbent families, structural and temporal graph shifts, and molecular prediction, with additional image and tabular evaluations. RTT makes GNN adaptation a budgeted, certifiable inference decision rather than an unconditional model replacement.

JWST has uncovered compact high-redshift (z>=4) sources, dubbed Little Red Dots (LRDs), characterised by a distinct V-shaped continuum. Their nature remains debated since broad-line AGN signatures coexist with non-AGN continua, marked by weak X-ray and mid-IR emission. We investigate which properties of LRDs naturally emerge within simulations of massive AGN host galaxies and which require additional physics. We post-process two cosmological zoom-in simulations of the same massive halo, one including black hole growth and AGN feedback (AGNfid) and the other with only star formation and stellar feedback (SFonly). We analyse snapshots spanning z=6-9 to probe different evolutionary stages. Using SKIRTv8, we develop a forward-modelling framework exploring various dust models, dust-to-metal ratios, and intrinsic AGN spectra. We generate mock spectra and images to apply JWST observational LRD selection criteria. Not designed to model LRDs, our AGNfid simulations naturally reproduce their properties: V-shaped continua, compact morphologies with rest-UV complexity, UV luminosities, attenuations, overmassive BHs, sub-Eddington accretion, and Balmer break strengths (~1-2) typical of most LRDs. Spectral decomposition reveals that the rest-UV continuum is dominated by stellar emission, whereas the optical continuum stems from a dust-attenuated AGN. Dust geometry and AGN-to-stellar contribution drive photometric differences between LRDs and Little Blue Dots. Our results suggest that LRDs do not necessarily represent a distinct class of objects. Instead, their defining properties can arise in massive AGN host galaxies from the interplay between stellar emission, an obscured central engine, and dust geometry. Super-Eddington accretion, dust-free gas columns, evolved stellar populations, and photoionisation modelling are required to test whether this framework reproduces the diversity of observed LRDs.

Simultaneous machine translation must generate target tokens before the source input is complete. Existing approaches address this through post-hoc read-write policies, leaving the attention mechanism unaware of bidirectional stream dependencies. We propose a dual-stream attention framework that represents source and target streams as a two-dimensional grid of hidden states and models their interaction through four structurally distinct attention types merged via joint QK Softmax normalization. Two approximations—broadcast and Hadamard—reduce the per-layer complexity from O(X^2Y+XY^2) to O(X^2+Y^2+XY) with provably decaying error. Training uses a self-guided loop: a per-cell loss heatmap drives dynamic-programming path recovery, which generates read/write decision supervision labels without external alignment. An incremental KV cache with anchored rotary position embeddings enables efficient streaming inference. On Chinese-to-English simultaneous translation, the proposed model outperforms the Wait-k baseline by +5.66 BLEURT and +10.36 COMET at comparable latency, and surpasses the non-streaming reference on COMET at a fraction of the response delay.

We propose a domain-incremental learning framework for generative speech enhancement (SE) that learns from a sequence of datasets or domains recorded under diverse acoustic conditions. Fine-tuning a pretrained model on continuously evolving domains leads to catastrophic forgetting of previously acquired knowledge, while zero-shot generalization often fails to adequately adapt to unseen domains. To address these challenges, we first develop a novel language model-based generative SE model that we then use as a pretrained backbone and incrementally adapt it to acoustically mismatched domains using lightweight domain-specific Low-Rank Adaptation. The proposed framework enables the model to acquire enhancement capabilities for new domains while preserving performance on previously learned domains. Evaluated on four heterogeneous speech datasets, our approach effectively adapts to new domains without forgetting previously learned domains.

Scaling image-to-3D generation to ultra-high resolutions requires controlling rapidly growing computational costs without sacrificing fine geometric detail. We present Filigree3D, a sparse latent flow-matching framework that generates 3D geometry from a single image at voxel resolutions up to $2048^3$, with straightforward extensibility to $4096^3$. To make training tractable, we introduce Structure-Aware Sparse Scaling, which combines spatial bounding with alternating local-global attention to constrain token growth while preserving both fine-scale details and long-range structural context. To enhance detail reconstruction, we curate training samples based on their high-resolution geometric gains and inject multi-scale image features into a sparse 3D DiT, effectively coupling structural semantics with fine-grained visual cues. Furthermore, a visibility-aware voxel regularization strategy improves robustness against sparse perturbations and facilitates the completion of unobserved geometry. Under our default configuration, Filigree3D maintains peak GPU memory consumption within practical limits for contemporary hardware, enabling the generation of highly intricate 3D geometry in approximately one minute. Extensive experiments demonstrate that our method yields substantial improvements in overall geometric fidelity and fine-detail preservation compared to existing baselines, validating practical, detail-preserving 3D generation at unprecedented resolutions.

Multi-vector retrievers built on vision-language models lead visual document retrieval (VDR), but they run a multi-billion-parameter query encoder on every search. Distilling this encoder into a small student that queries the teacher's existing index would remove the bottleneck. The standard recipe, however, matches the teacher's MaxSim scores and so requires encoding and caching every training page, which can reach terabytes of page tokens. NanoVDR avoids pages entirely by training on the teacher's query embeddings alone, but only for single-vector retrievers. We present ColNanoVDR, to our knowledge the first framework to bring this document-free distillation to multi-vector VDR. Its objective, OTW (Optimal Transport with Learned Weights), aligns the student's query tokens with the teacher's by entropic optimal transport, with a learned weight for each student token, and needs no correspondence between the two tokenizations. We prove that the resulting alignment cost bounds the MaxSim score difference on every page. Distilled from five state-of-the-art teachers, the 149M text-only students retain about 95% of their teachers' NDCG@5 on ViDoRe v1-v3 while encoding queries up to 26x faster. Under identical training, OTW matches score distillation while encoding no page and reading 12.6x less cached teacher data.

Let $F$ be a fixed graph without isolated vertices. An edge-colored graph is $F$-rainbow saturated if it contains no rainbow copy of $F$, but the addition of any missing edge in any color creates a rainbow copy of $F$. The rainbow saturation number $rsat(n,F)$ is the minimum number of edges in such a graph on $n$ vertices. We prove a dichotomy governed by isolated edges: if $F$ contains an isolated edge, then $rsat(n,F)=O(1)$ for all sufficiently large $n$, while if $F$ has no isolated edge, then $rsat(n,F)=\Theta(n)$. The linear lower bound is expressed in terms of a directed weight parameter $\eta(F)$ and establishes the linear half of the dichotomy; in several cases it also strengthens the Cameron–Puleo type coefficient. For the bounded half, we construct rainbow saturated graphs for targets of the form $H\cup K_2$. As an application of these constructions, we determine the asymptotically tight behavior for the rainbow saturation number of the generalized friendship graph $F_{t,p,q}=tK_p\vee K_q$, proving that $ rsat(n,F_{t,p,q})=(p+q-1)n+O(1)$ for fixed $t\geq 2$, $p\geq 2$ and $q\geq 1$ as $n\to\infty$.

Whole slide image classification is a fundamental task in computational pathology, where patch representation quality directly affects downstream aggregation and slide-level discriminability. Pathology foundation models are widely adopted as frozen feature extractors for WSI classification; however, their fixed encoders may produce representations insufficiently adapted to target-specific tissue patterns and discriminative cues. Fine-tuning can improve target adaptation, but introduces a trade-off between pathology-specific representation capacity and adaptation efficiency, particularly in data-scarce settings. To address this, we propose a pathology role-guided mixture-of-experts feed-forward network (MoE-FFN) framework for efficient encoder-level representation learning. We design a two-stage training paradigm to establish and adapt pathology-aware expert specialization. In source-domain expert initialization, pathology-specific priors are distilled from a frozen Virchow2 teacher into a lightweight DINOv2-small student, while role prototypes serve as weak pathological anchors to encourage distinct expert functions. MoE-FFN blocks are introduced into selected high-level transformer layers to provide transformation diversity for heterogeneous pathological patterns. In target-domain adaptation, the initialized experts are refined through asymmetric prototype-guided optimization, enhancing task-relevant positive evidence and separating confusable hard negatives. The resulting encoder extracts offline patch representations that can be directly integrated with standard MIL aggregators. Experiments on the public BRACS dataset and a private PAROTID WSI dataset across five representative backbones demonstrate consistent improvements over the strongest baseline.