Papers

Timely information dissemination is essential in modern communication systems, where low informational latency is critical for networked control, monitoring, and real-time decision-making. The Age of Information (AoI) has emerged as a fundamental metric for quantifying information freshness at the receiver. A key challenge in AoI optimization is selecting the update generation rate: infrequent updates lead to stale information, whereas overly frequent updates increase queueing delays. We develop a structure-aware, data-driven framework for adaptive AoI optimization modeled as single-server queues with unknown system statistics. Our key idea is to exploit analytical representations of the AoI gradient, which depend on a finite-dimensional collection of system quantities that can be estimated from observations collected during system operation. This reduces the online learning problem to estimating the quantities required by the gradient, rather than learning the AoI objective as a black box. We propose an online stochastic approximation algorithm that jointly estimates these quantities and optimizes the update generation rate. We prove almost sure convergence of the algorithm to the optimal update generation rate under mild regularity and moment conditions, despite the temporal dependence induced by the underlying queueing process. We demonstrate the framework in four representative incomplete-information settings, including FCFS queues with unknown service characteristics or hidden background traffic, a preemptive LCFS system with partial observations, and a generate-at-will system with threshold waiting.

In the literature, research on representing the non-dominated set has focused almost exclusively on multi-objective optimization problems with Pareto-dominance. However, general ordering cones are used to incorporate preferences of decision makers into the optimization process and to model other specific types of optimization problems, such as ordinal optimization problems. In this work, we consider several representation approaches based on representation relations in the context of multi-objective optimization problems with polyhedral ordering cones. Our main research question is whether existing theory and solution approaches developed for the Pareto case can be used to address the more general setting.

We recall several derivations of temperature-dependent Cahn-Hilliard equations. The resulting equations are systems of mixed order. For a suitable subclass of equations we discuss its local-in-time well-posedness based on $L^p$-maximal regularity theory. We also discuss stability of equilibria.

We establish the strict Dirichlet Pólya inequality on every bounded simply connected planar domain. The proof uses a unitary boundary transport formulation of the Helmholtz equation and relates Dirichlet resonances to unitary eigenphase crossings through the Dirichlet-to-Neumann map. An exact area identity yields the sharp Weyl coefficient, while a Ky Fan estimate leaves a strictly positive remainder in the counting estimate. The general case follows by conformal exhaustion.

Evaluating 5G positioning on a physical testbed ties every measurement to one propagation environment and one antenna layout, so any question about an alternative deployment requires a new field campaign. We present a ray-traced digital twin of the GEO-5G testbed at EURECOM that drives an unmodified OpenAirInterface (OAI) 5G stack with per-antenna channels computed by Sionna-RT over a 3D model of the site. The distributed gNB antennas are modelled as independent radio devices rather than as elements of a compact array, which preserves the per-antenna absolute delays on which time-difference-of-arrival (TDoA) positioning depends. With them preserved, the twin drives OAI's SRS-based UL-TDoA pipeline to metre-level accuracy consistent with that reported for the physical deployment. We then use the twin to answer a question the physical testbed cannot. Widening the carrier truncates the tail of the horizontal error distribution without improving a typical position, and leaves the vertical error untouched, because the latter is limited by the array geometry rather than by the measurement. That limit is predictable from the antenna coordinates alone, and a height-diverse layout reduces the RMS height error by a factor of four.

We establish a principle that converts qualitative regularity into uniform interior estimates for degenerate equations. For the $p$-Laplace equation, $p>2$, we prove that local $C^{1,\gamma}$ regularity of every solution, without any uniform bound on the corresponding seminorms, implies uniform $C^{1,\beta}$ estimates for every $0<\beta<\gamma$. Thus, the optimal qualitative and quantitative gradient Hölder exponents coincide as suprema in every dimension. In the plane, this yields estimates controlled by the $L^p$ norm of the solution for every exponent below the sharp Iwaniec–Manfredi threshold. Applied to Xu's recent qualitative $C^{1,1/3}$ regularity for planar infinity-harmonic functions, we obtain uniform interior $C^{1,\beta}$ estimates for every $0<\beta<1/3$. For the nonlocal superconductivity equation $|Du|(-\Delta)^s u=0$, with $1/2<s<1$, we prove that pointwise differentiability of all solutions with integrable tails implies uniform interior $C^{1,\alpha}$ estimates for bounded solutions, for some $\alpha>0$. The proofs combine compactness for shifted equations with an improvement-of-flatness iteration at variable scales.

We investigate generalized Stieltjes constants attached to Hurwitz-type Dirichlet series of the form $$ L(s, f, x)=\sum_{n \geq 0} \frac{f(n)}{(n+x)^s} $$ where $f$ is a periodic arithmetic function and $x\in\mathbb C\backslash\{0,-1,-2,\dots\}.$ We prove that these series admit meromorphic continuation to the whole complex plane, with at most a simple pole at $s=1$. We further derive functional equations satisfied by $L(s,f,x).$ Our main result gives an explicit formulas of the Stieltjes constants $\gamma_n(f, x)$ appearing in the Laurent expansion at $s=1$. In particular, we obtain explicit evaluations in the case of rational $x$. As an application, we derive explicit closed-form expressions for the first three generalized Stieltjes constants associated with nonprincipal real primitive odd Dirichlet characters arising from quadratic fields. These expressions are given in terms of the class number of the corresponding quadratic number field. These results extend and unify recent work of Blagouchine and others on generalized Stieltjes constants and Hurwitz-type $L$-functions.

The entanglement-breaking index of a channel is the number of self-compositions after which all entanglement with a reference system is destroyed. In a thermal qubit collision loop whose round is affine in the bath polarisation p, some circuits have thermal valleys: on an interior window of p the index falls below its infinite-temperature value. Rounds at different p compose exactly, so every schedule of polarisations defines a composite channel whose entanglement-breaking status can be certified in ball arithmetic. A gated search of 33 million random circuits finds no valley with contrast above about 3e-3, with an interior maximum near index 90, and yields six certified targets. At the reference target, held inside its window for 88 rounds, a centred excursion to p = 0.05 restores entanglement at exactly eight rounds: every shorter centred pulse certifiably fails, and so do the same eight rounds at the start. A 53-round pulse certifiably beats the static protocol, a depth-three valley is walked down its staircase by pulse duration, and a 1.2 per cent coherent tilt of the ancilla opens the switch at fixed population. Closing the switch costs 21 to 35 rounds. The asymmetry is not one of local rates, which agree to within about two per cent at matched circuit positions (certified). It is one of distance and response: closing must remove an entanglement margin 2.2 times larger, and while hot rounds added to a window hold act almost independently, window rounds added to a hot hold act collectively. Near the crossing, the lever of the log-margin is that of a linear approach to the entanglement-breaking boundary, with no free parameter. At the reference target, depolarising noise of about $4\nu/n$ per round kills a signature of size $\nu$ read after $n$ rounds: 1.3e-4 for the static signal, 6.5e-6 for the eight-round switch.

We present Neurograf v1, an experimental end-to-end Romanian-first language-model pipeline developed to investigate the feasibility of training language models from scratch for a lower-resource European language. The pipeline comprises RoSfera, a 635 GB Romanian corpus containing 175.7 million documents; an 80k byte-level BPE tokenizer; five decoder-only models ranging from 160M to 6B parameters, trained on 131 billion tokens; three instruction-tuned assistants; and two contrastively trained embedding models. Training and evaluation were conducted using computing infrastructure at ELI-NP/IFIN-HH. The model family exhibits consistent scaling of validation loss and increasing Romanian factual and cultural knowledge with model size. On the 1,000-question WWTBM benchmark, Neurograf-6B achieves 69.89% accuracy on Romanian-context questions, while Neurograf-3.2B reaches 67.38%. Compute-optimal analysis yields a parameter-scaling exponent of approximately 0.423. Neurograf v1 serves as an experimental demonstrator rather than a production system. The study identifies practical challenges involving diacritic preservation, corpus quality, data composition, and tokenization efficiency, while demonstrating the feasibility of developing Romanian-first base models, instruction-following assistants, and semantic embedding models within a unified training pipeline.

Demand management policies aim to shift individual behavior toward sustainable choices (e.g., goods, transport modes). Personalized incentive policies tailor subsidies to agents' needs and constraints, but personalization also exposes the regulator to agents' strategic manipulative behavior. The extent to which agents can profitably manipulate such policies remains poorly understood, which limits their applicability. This paper establishes analytically the conditions under which manipulation benefits the agent. We formalize a Stackelberg game in which an agent may simulate a different profile to extract higher incentives from a regulator. Previous work typically assumes that the regulator would instantaneously classify agents into the simulated user profile. In contrast, we explicitly model the regulator's process of learning the agent profile, based on observed past behavior. This makes manipulation harder, since it forces the agent to consistently behave according to a simulated profile, thus renouncing following their true preferences. Manipulative agents thus incur a Cost of Manipulation, which has typically been neglected. By accounting for it, we get a counterintuitive finding: although classification uncertainty reduces personalization accuracy, it may benefit the regulator by deterring manipulation. We establish an upper bound on the agent's gain from manipulation, which becomes stricter as profiles become harder to distinguish for the regulator. Simulations on the London Passenger Mode Choice dataset illustrate this deterrence effect. The framework can therefore be used to stress-test the robustness of personalized incentive policies before deployment.

LLM-as-a-Service has become a dominant paradigm for accessing large language models without requiring local deployment. However, the process raises severe privacy concerns when user queries may include sensitive or confidential contexts. Differential privacy (DP) provides a principled approach to protecting such inputs, but existing DP inference methods are primarily designed for short- or moderate-context settings and struggle with million-token-scale inputs due to limited context capacity and severe utility degradation under DP perturbation. In this paper, we propose DP-MIND, a memory-guided framework that scales differentially private inference to million-token contexts. DP-MIND enables private long-context inference by locally perturbing the input and processing it chunk by chunk with a compact memory that preserves task-relevant information across the sequence. To make this memory robust to DP noise, we optimize the memory construction policy directly on perturbed inputs using reinforcement learning, guided by both final answer correctness and memory consistency. We further recover perturbation-corrupted answer spans through a local denoising module without exposing the original context to the external model. We provide theoretical privacy analysis of DP-MIND and further conduct extensive empirical analysis across multiple long-context benchmarks with context lengths up to the million-token scale, demonstrating that DP-MIND consistently outperforms existing DP and long-context inference baselines in both scalability and utility.

Trajectory generation underpins urban applications such as traffic management and urban planning. Within road networks, vehicle movement along road segments is largely deterministic, whereas transitions at intersections are critical–multiple outgoing paths may appear equally plausible, yet they can lead to substantially different downstream routes. Existing generative methods such as diffusion models simulate location sequences and fail short of next transition selection, even producing trajectories that drift off the road network. To fill this gap, we advocate a fundamental shift: moving beyond location simulation toward behavior modeling on the road network. In this paper, we revisit trajectory generation through the lens of generative route planning, and introduce a new graph-constrained generative paradigm, Anticipatory Generative Modeling, which resolves next transition selection by incorporating the Anticipation of how the future route may unfold toward the destination. Concretely, we propose a novel generative model named Foresight that formulates a probabilistic model for an "anticipate–select–update–re-anticipate" loop that enables longer-term foresight while continuing sequential generation along the historical trajectory. Extensive experiments demonstrate that Foresight wins state-of-the-art baselines.

Datacenter congestion control (CC) and load balancing (LB) evolve rapidly as workload demands change, but deployment depends on RDMA hardware and transport interfaces. Emerging programmable RDMA engines tie policy code to vendor APIs. Operators must therefore port each policy for every platform they deploy. We address this with PCM, a programming model that separates portable congestion management from platform integration on datacenter NICs. Developers write event-driven handlers that they can debug in simulation and reuse on real hardware. PCM exposes flow telemetry as signals and uses triggers to support per-event reactions and coarser responses to accumulated feedback. Vendors provide optimized backends that map signals and output controls to native NIC mechanisms with defined feedback and control semantics. We demonstrate PCM's expressivity by implementing eleven well-known datacenter CC and LB algorithms. We deploy PCM across four distinct datapaths: an Ultra Ethernet simulator, CPU-based UCCL, and on-NIC offloads on NVIDIA ConnectX-8 and Microsoft 1P. On CX-8, PCM-Swift achieves processing capacity equivalent to one reaction per three 4 KiB packets at 800 Gbit/s, using one-fifth of the NIC's hardware threads. In UCCL, it matches native Swift's 1.56 Tbit/s goodput with 64 KiB chunk-level control. PCM's design informed ongoing Ultra Ethernet work toward a portable congestion management interface.

Dynamic Time Warping (DTW) is a common choice for measuring the discrepancy between time series. It proceeds by finding an optimal warping between timestamps through dynamic programming, which can be slow and is not differentiable. Consequently, several variants have been proposed. Among them, Soft-DTW proposes a smooth formulation of DTW, amenable to gradient computation, while Continuous-DTW (CDTW) introduces a continuous formulation through parametrizable warping functions. Yet, computational costs remain high for both methods and impractical for large datasets in modern machine learning pipelines. This paper revisits CDTW in several ways. After establishing its differentiability under mild conditions, we introduce FastCDTW, which is an efficient approximate solver with two different computational strategies: an exact closed-form scheme for piecewise-linear warping functions, and an unbiased Monte Carlo estimator for smooth ones, such as implicit neural networks. Both strategies provide gradients with respect to the warping function and the input time series, and the resulting algorithm is GPU-friendly. We demonstrate its benefits through its integration into more complex learning problems, including time series averaging, classification and deep domain adaptation.

Question answering over large heterogeneous document collections requires first finding the right passages and then reading them, yet current retrieval-augmented agents give one model both kinds of tools and let it schedule itself. We present KnowForge, a pipeline separated by job: a hybrid-retrieval knowledge-point index; a Locator agent that can only search; a code-scheduled bridge judge that forces the missing second hop; a typed, size-bounded evidence pool that merges every search's ranking; a Deep Reader agent that can only read text, images and tables; and a synthesizer with a format checker. With a self-hosted Qwen3.6-35B-A3B (3B active), KnowForge reaches F1 47.5 on the full open-domain M3DocVQA (2,441 questions, 3,368 PDFs), +11.0 over the published M3DocRAG system and +7.1 over one-shot retrieve-and-answer on the same index and model, and 53.5 with DeepSeek-V4-Flash. On MultiHop-RAG it reaches 73.5 accuracy, above the strongest published RAG and GraphRAG results (71.2); on the single-document MMLongBench-Doc it lifts the same model from 34.8 to 47.0, level with GPT-4o reading page images (46.3). Per-job ablations on the full benchmarks attribute the gain: the Deep Reader is worth 3.8, 9.8 and 0 points on the three corpora, tracking how visual each is; merging both agents into one with all tools and twice the rounds loses 1.9 F1 because it keeps searching instead of reading; a round-budget sweep shows a step at four Locator rounds and a plateau beyond; query latency scales with model calls rather than corpus size. We serve the three indexed benchmarks as read-only knowledge bases (free registration) and release all prompts and the runner code: https://github.com/metask-ai/knowforge-bench

Aims: To analyse the impact of dynamically formed binaries emerging naturally from various star cluster environments on already established planetary systems. Methods: Numerical simulations replicating the star-forming region NGC 1333 are conducted. We consider three models, each using identical cluster realisations and simulation parameters. The cluster itself is initially sub-virial and fractally distributed. Model~A assumes planetary systems are initially in mean motion resonance. In Model~B, the same planetary systems are moved out of resonance, but remain intrinsically stable. Finally, Model~C removes all planets from the simulation. Ten independent realisations are integrated per model until 30 Myr. Results: Dynamically formed binaries emerge at similar rates across models, indicating planets play little role in their formation. We also find a higher production rate of wide binaries compared to previous numerical studies which we attribute to the sub-virial and sub-clustered initial conditions. S-type and binary S-type planetary systems naturally emerge in our simulations, while P-type systems are rare. For Model~A, the fraction of S-type or binary S-type systems is f_S=0.02 and $f_Bin,\, S=0.05, respectively. Rogue planets originating from a binary system are ejected with a substantially larger velocity than those around singletons. Planet-star collisions and planet capture events occur disproportionately within binary systems. Finally, several \alphaCentauri-like systems emerge in our simulations, suggesting that the nearest stellar system to the Sun may have a dynamical origin.

This paper studies exit timing for SPX short-straddle portfolios using a two-stage supervised and reinforcement-learning framework. Historical episodes first generate ex-post optimal hold/exit labels for training a multilayer perceptron (MLP) using only contemporaneously available state variables at deployment. The trained MLP then initializes reinforcement-learning actors that are fine-tuned with a risk-aware trading reward. Out-of-sample evaluation over 2024-2026 shows that the supervised MLP substantially outperforms passive holding and representative fixed-time and profit-and-loss-based exit rules. Across seven independent seeds, MLP-initialized Actor-Critic averages 37.54% cumulative return and a Sharpe ratio of 1.705, compared with 35.01% and 1.651 for the supervised MLP. Actor-Critic trained from scratch averages only 0.18% return. These findings show that oracle-guided supervised learning provides a stable foundation for reward-based policy fine-tuning in a sparse and path-dependent option-exit problem.

Unified Memory Architecture (UMA) enables the CPU and GPU to share a single physical memory. However, its complexity and opacity often drive applications back to inefficient copying or to incorrect direct access that sacrifices performance or even correctness. This paper presents a comprehensive study of CPU-GPU memory sharing on UMA, covering correctness and performance across data-plane transmission and control-plane coordination on four platforms. It yields four findings that guide libumsh, a portable library that delivers correct and efficient memory sharing across platforms and extends to emerging ones (e.g., NVIDIA DGX Spark). Applied to nine real-world systems, libumsh reduces data-plane time by 16-75% and control-plane overhead by 7-90%.

Two-dimensional (2D) superconducting systems have recently emerged as a frontier in condensed matter physics, driven by their unique physical properties and potential applications. Using first-principles calculations, we show that monolayer ZrX$_2$ (X = Cl, Br, I), which is a family of 2D semiconducting electrides, hosts phonon-mediated superconductivity under both electron and hole dopings. Electron doping in ZrX$_{2}$ can achieve high superconducting transition temperatures at very low concentrations, with ZrBr$_2$ reaching $T_c = 30.0$ K at 0.01 e/f.u., followed by ZrCl$_2$ (22.5 K at 0.05 e/f.u.) and ZrI$_2$ (18.1 K at 0.01 e/f.u.), for which the high $T_c$ originates from the flat conduction band minimum, with Van Hove singularities nearby enhancing the density of states. In contrast, hole doping requires higher concentrations (0.2–0.25 h/f.u.) to activate superconductivity, yielding $T_c \sim 2$ K, and approaches a magnetic instability at elevated doping levels. By comparing with InSe and LaBr$_2$, we identify the spatial location and orbital character of excess electrons as key factors determining the strength of hole-doped superconductivity and the nature of competing orders. These findings establish monolayer ZrX$_2$ as a platform for bipolar superconductivity in a 2D electride and provide insights into flat-band engineering for high-$T_c$ superconductors.

A solid understanding of the predictive success of deep neural networks (DNNs) remains elusive. For example, while many data-dependent metrics exist that seek to predict model quality of DNNs, these metrics may be expensive to compute for large datasets and/or may be impossible to compute when training datasets are not released. A practical and effective predictor of model quality involves analyzing DNN weight matrices, as prior work has established that heavy-tailed spectral distributions in trained weight matrices can be strongly predictive of model quality. However, such metrics lack a rigorous statistical justification. In this work, we aim to understand why these metrics perform well, from a statistical perspective. To do so, we employ a probabilistic framework to derive the marginal likelihood for a DNN, evaluated using only the final layer weights. Across large-scale convolutional DNNs and LLMs, we find that this marginal likelihood predicts model quality. To understand the role heavy-tailed spectra play in predictive performance, we derive the limiting log-marginal likelihood for spectra that follow the (heavy-tailed) High-Temperature Marchenko-Pastur (HTMP) distribution. We show that this limiting quantity is convex in a heaviness parameter, and we derive an optimal heaviness that empirically predicts over-fitting for large-scale DNNs.

Vision-and-language navigation (VLN) requires an agent to track progress through an extended instruction, identify landmarks relevant to its current objective, and understand the surrounding spatial structure to act. End-to-end navigation policies typically rely on action supervision to implicitly learn these capabilities. Without explicit guidance on the evidence underlying each decision, however, policies can overfit to correlations predictive of training actions rather than learn the task-relevant grounding and geometry needed for generalization. Our insight is to guide the representations that connect instructions to actions: identifying what remains to be done determines which visual evidence matters, while understanding the geometry around that evidence informs how to move. We introduce GuidedNav, a unified framework that explicitly shapes these representations through training-time guidance. Subtask and landmark guidance steer native pre-action attention toward the active instruction segment and relevant visual regions, grounding action prediction in the current objective. Spatial guidance aligns the policy's internal representations with geometric features that capture long-horizon context and local spatial structure. Together, these objectives strengthen navigation capabilities while avoiding the inference overhead of intermediate textual reasoning and online geometric encoding. We also introduce Guided-R2R and Guided-RxR, two annotation-enriched navigation datasets that support more interpretable VLN research. GuidedNav establishes state-of-the-art performance among the compared monocular RGB policies, improving success rates by 8.1% and 13.1% relative to the strongest prior results on R2R-CE and RxR-CE val-unseen, respectively, while achieving faster inference. Extensive experiments further demonstrate the effectiveness and real-world transferability of our approach.

Cyberbullying is a growing concern in online communities, yet existing Bengali corpora remain limited in contextual information, annotation transparency, scale, accessibility, and support for multidimensional analysis, thereby hindering robust detection and comprehensive study of harasser behaviors and networks. To address these gaps, we introduce the Bengali Contextual and Diverse Cyberbullying (BenCoDiC) corpus, a publicly available contextual cyberbullying resource for Bengali political news discourse comprising 50,250 textual instances collected from 100 YouTube videos and their associated discussions. The corpus preserves caption–comment–reply relationships and is organized through a nine-dimensional star schema covering users, videos, textual content, context, time, engagement, political parties, target identities, and language categories. Five AI chatbots were initially evaluated for annotation but did not meet the predefined reliability criterion. Consequently, a two-tier human annotation process with expert verification was adopted, achieving Fleiss' $\kappa=0.86$ and Krippendorff's $\alpha=0.86$. Statistical analyses characterize the corpus, while transformer-based experiments show that incorporating conversational context yields modest but consistent improvements over isolated-text classification. BenCoDiC supports context-aware cyberbullying detection and perpetrator-oriented behavioral and network analysis. The corpus is available at https://github.com/bencodic/BenCoDiC-Corpus.

Decoding dynamic visual content from EEG has advanced rapidly for visual perception, yet it remains unclear which aspects of neural representations are preserved when people internally replay dynamic scenes. We introduce PAIR, an EEG benchmark collected from 20 participants using an interleaved, clip-wise paradigm that pairs naturalistic video viewing with immediate visual recall. This design enables a comparison between externally driven perception and internally generated recall for the same stimulus content within a fixed trial sequence. We benchmark multiple decoding models on hierarchical semantic classification at both fine-grained and coarse-grained levels, meta-feature decoding tasks covering motion, color, object number, face, and human presence, and a state-decoding task that discriminates perception from recall. Within-subject validation results show numerically similar performance for several model–task combinations, alongside higher perception accuracy in two coarse-grained comparisons after multiple-comparison correction. In contrast, perception and recall are highly separable at the state level under the present validation protocol. State accuracy remains high after alpha-band ablation, although this does not exclude ocular or other task-related contributions.

Collecting dexterous manipulation demonstrations for each robot hand requires repeated task-specific data acquisition. We ask whether one shared corpus of natural human demonstrations can instead serve as a reusable data interface for different dexterous hands, without collecting task demonstrations on each target robot. We present Human2Dex, a wrist-centric human demonstration interface that separates shared human observation and motion semantics from embodiment-specific execution. It aligns observations through local hand-appearance adaptation and a shared 21-point hand representation with consistent structural semantics across human and robot observations. It constructs executable supervision by retargeting fused human hand motion to each target hand, and introduces the Grasp-Pocket Interaction Representation (GPIR) to express tracked objects relative to a hand-centered Functional Grasp Pocket. The corpus is converted into separate training sets for Linker O6 and Wuji, while action labels, normalization, training, and diffusion-policy checkpoints remain hand-specific. Across six real-world tasks, Human2Dex achieves 109/120 (90.8%) successes on Linker O6 and 97/120 (80.8%) on Wuji. Across both hands, Human-Raw achieves 63/240 (26.3%) successes, whereas the complete Human2Dex configuration achieves 206/240 (85.8%). These results support the use of wrist-centric human demonstrations as a reusable interface across calibrated dexterous hands while retaining embodiment-specific policy learning. The project website is publicly available at https://human2dex.pages.dev/.

We study when communication emerges between autonomous agents and what role an innate action selection architecture plays. Two e-puck robots, reconstructed in PyBullet from their Webots model and validated against it, must find a feeding zone that can only be detected when a robot passes over it. Each robot can emit a one-bit signal whose direction is detected by the receiver. Control is provided by a GRU recurrent network with a residual path, either operating alone or as a channel within an Embedded Action Selection Architecture (EASA). Fitness measures only the time the robots spend over the food. When the robots receive a compass pointing to the food, use of the signal disappears completely. Without a compass, in a small arena, three independent runs achieved similar performance using different strategies: a referential signal, a permanent beacon, or no use of the signal. As the arena expands, the strategy based on a referential signal generalizes best. In contrast, evolving directly in the large arena does not increase the frequency of communication, but it does separate the strategies, and the referential signal clearly outperforms solitary searching. Within EASA, innate reflexes constrain the network and reduce performance. As its priorities evolve, evolution consistently weakens those reflexes. The results suggest that communication emerges in proportion to the value of the information it transmits, and that an innate scaffolding designed for isolated agents can be an impediment in a social task.

Vision-language-action (VLA) models and world action models (WAMs) have advanced robotic policy learning by building on pretrained vision-language models (VLMs) and video generative models (VGMs), respectively. Yet VLAs learn without explicit supervision from future world states, while WAMs that condition actions on predicted futures incur extra inference cost and let visual prediction errors propagate into actions. We present Forward Dynamics Models (FDM), a family of robotic action models that couples action and future world-state prediction through an action-to-future paradigm. Future prediction is conditioned on action features, never the reverse: end-to-end training lets ground-truth future states supervise the policy through these features, while prediction errors cannot propagate into actions. At inference, FDM skips future generation entirely, avoiding its computational cost. The paradigm is backbone-agnostic, and we instantiate it with both VLM and VGM backbones. Experiments in simulation and on real robots demonstrate the effectiveness of FDM across both backbone families, achieving a state-of-the-art 99.4% average success rate on LIBERO and absolute gains of up to 13% over VLA and WAM methods on real-world tasks.

As large language models (LLMs) acquire advanced biological design capabilities, they offer substantial research benefits but simultaneously introduce severe dual-use risks, such as engineering enhanced pathogens or drug-resistant bacteria. While biosecurity frameworks recognize these threats, the actual risk of LLMs executing dangerous biological modifications remains unquantified, particularly when augmented with specialized biological design tools (BDTs). Existing evaluations primarily assess hazardous knowledge possession rather than practical application, overlook risk amplification from BDTs, and rely on unreliable, expensive, or hazardous validation methods. To address these gaps, we introduce DRYLAB-Bench (Design Risk Yielded by LLM Agent in Biodesign), the first comprehensive benchmark evaluating the high-risk biological mutation design capabilities of LLM agents. Moving beyond basic question-answering, DRYLAB-Bench tasks agents with 30 genuinely hazardous mutation-design scenarios distilled from biosecurity guidelines. It evaluates models in a multi-round, tool-augmented environment where LLMs orchestrate BDTs to iteratively refine candidates. Crucially, it scores designs directly against 187,695 empirical wet-lab measurements from ProteinGym and ViroGym using our proposed Biosecurity Risk Score (BRS), quantifying risk-associated effects while enabling scalable and safe ground-truth verification. Evaluations of 15 frontier models reveal substantial baseline threats: without tools, 67.6% of one-shot designs exceed the average measured variant, and 21.4% reach the highest-risk decile. Equipping models with BDTs significantly amplifies this risk, raising the pooled risk by \(\Delta BRS=+0.116\) and the top-decile share to 23.5%. Code: https://github.com/drylab-bio/DRYLAB-Bench

The phase of an electron wave is sensitive to the local electrostatic environment and can therefore provide information about atomic-scale structural variations. In electron ptychography, however, specimen thickness, multiple scattering and projected atomic overlap can complicate the quantitative interpretation of the reconstructed phase. Here, we use a twisted trilayer WS2 thickness staircase containing monolayer, bilayer and trilayer regions as an internal reference, with all regions measured under identical experimental conditions. The mean W-column phase increases by approximately 13.2% from the monolayer to the bilayer region, compared with only 4.7% from the bilayer to the trilayer region, revealing a nonlinear thickness dependence. W-S overlap has little effect on the W-column phase, whereas short-range projected W-W overlap increases it by approximately 8-9%. These results provide a benchmark for evaluating quantitative phase measurement in multilayer WS2 and for assessing phase-based mapping of local electrostatic potential, charge-density and three-dimensional atomic-scale polarization.

Graph matrices that combine structural and metric information have attracted considerable attention in spectral graph theory. In this paper, we introduce the Degree–Adjacency–Diametrical (DAD) matrix of a connected graph $G$. The DAD matrix is defined by \[ (DAD(G))_{ij}= \begin{cases} d(v_i), & \text{if } i=j, 1, & \text{if } v_i\sim v_j, \operatorname{diam}(G), & \text{if } d(v_i,v_j)=\operatorname{diam}(G), 0, & \text{otherwise}, \end{cases} \] where $d(v_i)$ denotes the degree of the vertex $v_i$. We establish fundamental spectral properties of the DAD matrix by deriving trace identities, characteristic polynomials, and spectra for several classes of graphs. Furthermore, we obtain a spectral expression for the Wiener index of graphs with diameter two, derive upper bounds for the largest DAD eigenvalue, and establish lower bounds for the DAD energy. Explicit spectral characterizations are also obtained for complete multipartite graphs. The proposed matrix provides a unified spectral framework that simultaneously incorporates degree, adjacency, and diametrical information, thereby enriching the study of graph spectra and introducing a new graph matrix for spectral graph theory.

Continual post-training faces a dual challenge of acquisition and retention: models must learn new capabilities while preserving prior knowledge. While Self-Distillation Fine-Tuning (SDFT) mitigates this using self-supervision with in-context demonstrations, it relies on strong in-context learning and instruction-following abilities. Consequently, SDFT fails on smaller models critical for low-latency, on-device use. To enable continual learning at smaller scales, we first identify the root cause of failure. We show that poor acquisition and forgetting correlate linearly with the post-trained model's distribution shift, revealing the necessity of in-distribution learning. We thus introduce Synthetic Self-Distillation (SynthSelf), imitating SDFT's in-distribution rewrite by outsourcing the task to a minimally larger, same-family aligner. Coupled with off-policy supervision and rejection sampling, this provides the high-quality supervision small models require without breaking their original distribution. Evaluations spanning 4 acquisition datasets (including a mixture) and 14 retention benchmarks show SynthSelf consistently outperforms 6 methods (including off- and on-policy distillation) across 3 small model scales. Extensive further analyses confirm SynthSelf paves the way toward effective small-model continual learning.