Science & Mathematics

Many learning systems are performative: once deployed, their predictions change the data distribution on which future models are trained. In multi-agent settings, this feedback is routed through a network of interacting learners, so stability depends not only on the strength of performativity but also on who affects whom. We study repeated retraining in multi-agent performative prediction and identify a spectral threshold that separates stable learning from worst-case learning obstruction. Prior work often collapses cross-agent feedback into a scalar contraction constant, missing this network structure. We instead introduce the reflexivity matrix $Γ$, which records how strongly one agent's deployment can move another agent's best response, defined from per-agent gradient sensitivities and curvature parameters. The threshold occurs when the spectral radius of this matrix crosses one. Below the threshold, repeated retraining converges geometrically to a unique performatively stable equilibrium, with cumulative excess loss bounded uniformly in the time horizon. Above the threshold, we construct noisy supercritical instances in which every local-information learner incurs expected cumulative excess loss growing linearly for horizons up to order $1/(ρ(Γ)-1)$. Together, these results characterize the threshold behavior through upper and lower bounds and recover the single-agent scalar threshold from prior work as a special case.

We explain an attempt to define semi-simplicial types (SST) in homotopy type theory by first defining a small type theory whose contexts are exactly the data of semi-simplicial types, constructing $\mathrm{SST}\,n$ uniformly in that syntax, and then interpreting the syntax in the universe. The syntax can be defined without any truncation and without K, and the construction of $\mathrm{SST}\,n$ goes through. The naive interpretation does not: it transports along the semantic images of equations that hold only propositionally, and making these transports agree needs uniqueness of identity proofs (UIP) for the semantic types. Moving the problem into the syntax, by making composition a constructor of a higher inductive-inductive type (HIIT), does not remove it either: it reappears as a coherence problem in the syntax. Fording the HIIT (storing equations instead of computed indices), with conditional rules, makes the syntax provably a set. This is UIP for the syntax, obtained by encode-decode, and it does not by itself restrict the semantics; but the naive interpretation of the conditional rules again needs UIP for the semantic types. A semantics into groupoids that avoids UIP will be discussed in a separate note. The Agda development accompanying this note is in the repository https://github.com/txa/sst-new.

Legged robots can extend planetary exploration to rugged and discontinuous terrain that is inaccessible to conventional rovers. In lunar gravity, jumping further expands this mobility by enabling robots to cross obstacles and gaps that cannot be traversed by walking alone. We present a multistage reinforcement-learning approach that produces a single end-to-end control policy for commanded point-to-point jumping under lunar gravity. The resulting policy has a simple architecture and relies exclusively on mission-compatible onboard sensing, making its deployment on a real lunar mission plausible. In simulation, the learned policy performs accurate point-to-point jumps with landing errors smaller than the robot's body size. We analyze the torque and power demands of lunar jumping and characterize performance degradation under extreme outlier events, including control-update freezes of up to \SI{1}{s} during flight. Hardware transfer demonstrates that the policy can successfully execute a vertical jump under lunar-gravity offloading. Furthermore, the measured torque and power profiles are consistent with the simulation results and remain within the robot's hardware limits.

Reliable video perception over bandwidth-constrained wireless links requires transmitting information that the receiver cannot already recover from its evolving scene representation. This paper introduces WorldGT, a world-model-guided semantic packet transmission framework that prioritizes information according to its contribution to the receiver's perception. A shared feature-space world model predicts the receiver's scene state through affine motion compensation and learned residual displacement, while sender-side verification identifies reusable predictions. An acknowledgment-synchronized receiver twin estimates task-impact scores through two counterfactual evaluations, guiding packet selection and unequal error protection over LDPC-coded wireless channels. A compact 3-bit feature representation increases spatial coverage under stringent bandwidth constraints, while predictive fallback sustains perception during packet outages. Experiments on KITTI and VisDrone under Rayleigh fading and AWGN demonstrate substantial gains in low-SNR conditions. On KITTI at a 4.13% reference bandwidth budget, WorldGT achieves 47.5% J&F at 0 dB Rayleigh SNR, outperforming H.264 by 11.3 percentage points. Controlled ablations show that world-model prediction contributes up to 8.7 points under Rayleigh fading. Object-detection experiments further demonstrate the transferability of receiver-relative predictive transmission across visual perception tasks. These results establish world-model-guided, receiver-state-aware transmission as an effective approach to robust video perception under severe communication constraints.

Mobile edge services can combine local computation with satellite and cellular access to deliver time-sensitive results in areas with intermittent terrestrial coverage. Meeting their deadlines and quality requirements calls for joint computation and transmission scheduling that accounts for task dependencies, changing link capacity, and shared resource budgets. However, optimizing resource allocations or tuning parameters within a fixed policy structure offers limited scope to correct failures rooted in the scheduling logic. We propose SatCraft, a framework in which a large language model (LLM) agent uses execution feedback to revise the code of joint scheduling policies. Each revision undergoes policy checks and complete-run evaluation to assess its validity, runtime, and effects on service completion. Because policy performance varies with access conditions, SatCraft retains valid alternatives in a policy library and selects one for each episode using initial access observations and the known resource configuration. Experiments using Starlink and cellular measurement data show that SatCraft improves cumulative completed-service value by 3.61%-16.89% over baselines, including recent state-of-the-art scheduling methods ADPRL, EdgeOPT, and TF-DDRL. It achieves the highest completed-service value in all four resource profiles and averages 25.02 ms per scheduling call. These results demonstrate how LLM-guided policy optimization can improve complete service delivery under the computing, communication, and quality constraints of satellite-cellular networks.

Graph neural networks (GNNs) are among the standard tools for molecular property prediction, but the atom-level explanations of their predictions are hard for medicinal chemists to act on. We propose a change of input representation rather than a new model: molecules are encoded as fragment-level graphs from BRICS decomposition, where each node is a synthetically meaningful fragment whose features are the element-wise sum of its atoms' features. Because only the representation changes, any message-passing GNN and any post-hoc attribution method produces explanations at the resolution of functional groups instead of single atoms. Across four architectures, four prediction tasks, and two explainers under scaffold splits, fragment-level encoding matches atom-level test accuracy on roughly five-fold smaller graphs. Integrated-Gradients attributions recover known pharmacophores above chance on two enzyme targets, namely BACE at 1.5 times the chance rate, and PTR1 at 3.2 times. A like-for-like comparison against an aggregated atom-level baseline shows that post-hoc aggregation is not a stable substitute: its BACE hit rate range both exceeds and falls below the fragment model's hit rate depending on two arbitrary aggregation choices, whereas fragment-level training yields a single convention-free attribution at no predictive cost. The code is available at: https://github.com/4D-Lab/frame

Cinematic multi-shot video generation requires both narrative planning and iterative refinement. Existing methods either lack narrative reasoning or rely on feed-forward workflows without policy updates. In this paper, we propose CineEvolve, an agentic self-evolution framework for cinematic multi-shot video generation that unifies a Director and a Critic within a single vision-language policy. Emulating real-world filmmaking that couples iterative direction with critical feedback, we train CineEvolve via alternating reinforcement learning, where the Director surfaces flaws to sharpen the Critic, and the updated Critic guides the Director's targeted revisions. To stabilize co-evolution, we introduce Evolutionary GRPO (E-GRPO), leveraging historical anchoring and revision safeguarding to mitigate advantage inflation. Additionally, we contribute CineTraj, a 40K cinematic trajectory dataset spanning creation, critique, and revision. Extensive experiments demonstrate that CineEvolve outperforms existing approaches. Our code will be available at https://github.com/JingZhe-8/CineEvolve.

In this paper, we study quantile distributional policy evaluation, where the return distribution under a given policy is approximated by the quantile Bellman fixed point $\bmη_m$. We investigate the statistical accuracy of estimating $\bmη_m$ across quantile resolutions under Wasserstein distances. Assuming access to a generative model, we construct a quantile fixed point estimator $\bmη_m^{(n)}$ from an empirical Markov decision process and establish non-asymptotic error bounds under Wasserstein distances. When reward densities are uniformly bounded away from zero, we obtain a root-$n$ high-probability bound simultaneously over all resolutions $m$ and all Wasserstein orders $1\le p\le\infty$. When reward densities vanish at the support endpoints, we show that continuity and endpoint monotonicity suffice to preserve the same resolution-independent rate for $1\le p\le2$. For $p>2$, we establish a root-$n$ bound with coefficient $m^{1/2-1/p}$, together with a slower, resolution-independent $n^{-1/p}$ bound for every finite $p>2$. We further obtain resolution-independent $W_\infty$ consistency governed by the continuity modulus of reward quantile functions. To investigate the sharpness of these bounds, we characterize the exact local asymptotic minimax risk at each fixed resolution as the expected Wasserstein-coordinate norm of the efficient Gaussian limit and prove that our estimator attains it. We then examine the resolution dependence of this optimal local risk. A regular one-state instance has resolution-independent local risk constants, whereas a single endpoint-vanishing two-state instance exhibits growth of order $m^{1/2-1/p}$, up to logarithmic factors along a resolution subsequence, for every $p>2$.

This study investigates the microstructural evolution of the powder metallurgical high-speed steel PM HS 3-3-4 during isothermal holding within the transformation gap between pearlite and bainite formation. The focus is on the mechanism of carbide spheroidization and the proposed formation of a chemically graded carbide rim. The heat treatments were carried out in a quenching dilatometer, enabling precise control of the thermal cycles and isothermal holding conditions. Spatially resolved analytic microscopy techniques, including scanning transmission electron microscopy (STEM), Auger electron spectroscopy (AES) and atom probe tomography (APT), image analysis and complementary thermodynamic simulations (MatCalc) were used to characterize the carbide evolution. Isothermal holding increased carbide circularity, carbide volume fraction, and hardness, while MatCalc calculations predict a strong thermodynamic driving force for additional carbide precipitation. The analytic techniques revealed a compositional gradient of approximately 15 nm at the carbide-matrix interface in the isothermally treated condition, compared with a sharp transition of only 2-3 nm in the directly quenched state. This nanometer-scale gradient provides strong evidence for epitaxial, interface-controlled carbide growth and the formation of a chemically graded carbide rim. The associated solute depletion of the surrounding austenite provides a plausible explanation for its reduced stability, reflected by an increase in the martensite start temperature. The combined results demonstrate that isothermal holding promotes interfacial chemical reorganization, offering new insights into carbide spheroidization and providing a basis for tailoring carbide morphology and the resulting mechanical properties through heat treatment.

Diffusion language models generate by iteratively denoising a masked sequence under bidirectional attention, without the strict left-to-right commitment of autoregressive decoding. Whether this difference affects how reliably they remain grounded in structured inputs is largely unexplored. We study this question in table-to-text generation using ToTTo, where highlighted source cells make factual grounding directly testable. We compare the masked diffusion models MDLM and LLaDA with GPT-2 and GPT-Neo under an identical retrieval-based few-shot protocol, together with a QLoRA-adapted LLaDA, and we score reference similarity and source grounding as separate axes. The two pull apart. Within LLaDA, additional demonstrations raise BLEU from 11.3 to 22.0 while NLI entailment against the source table falls from 0.78 to 0.52; QLoRA reaches the best BLEU of the study, 24.15, and grounds no better for it. Manual annotation shows the failure modes differ in kind: the autoregressive baselines tend to wander off the table, while LLaDA-based models more often stay on topic and swap out individual entities. Reference-based metrics alone are insufficient to assess faithfulness in table-to-text generation.

A frozen multi-horizon forecaster can be corrected online from its own residuals. However, under label delay the target of a block arrives only after the whole horizon has passed, and the corrector's retained state grows with the number of channels. Aiming to combine a small retained state with a large error reduction, we propose DRiC (DCT Ridge Corrector), which fits a ridge regression of the leading discrete cosine transform (DCT) coefficients of the next residual block on those of the frozen forecast and of the latest input block, with sufficient statistics updated recursively. Its local variant, DRiC-L, keeps one such regression per channel; its global variant, DRiC-G, pools all channels into one regression, whose statistics do not depend on the channel count. A guard tracks a running scale of the frozen forecast's residual for each channel; it skips updates on blocks far outside that scale and clips the issued correction. We evaluate on seven datasets (four ETT series, Weather with its missing-value codes filled, Electricity, and Traffic) with DLinear, PatchTST, and iTransformer backbones, against 10 naive, gradient-trained, and closed-form methods. DRiC-L with eight DCT coefficients (DRiC-L8) reduces the mean squared error (MSE) by 14.55% on average over 126 conditions, against 12.98% for ELF, the most accurate comparator, while retaining 15-17 times less state, although ELF is more accurate on Traffic. DRiC-G retains 14.6 kB on the 321-channel Electricity data and 18.9 kB on the 862-channel Traffic data and is the most accurate method at or below its state on both. However, 5 to 7 of the 10 comparators, all with at least 10 times its state, are more accurate there. On the unfilled Weather file, DRiC-L8 without the guard has on average 8.0 times the frozen forecast's MSE, whereas with the guard it reduces the MSE by 7.94%. Code: https://github.com/keiotakmin/dric.

Stochastic flow maps promise few-step sampling of stochastic differential equations (SDEs) by learning to directly jump between times along the solution path, yet training these models to learn this solution map remains difficult. The primary obstacle is the Brownian path itself: it is an infinite-dimensional object which needs to be passed to the neural network and is nowhere differentiable, which makes the implementation of derivative-based losses that power the state-of-the-art deterministic flow maps just out of reach. We identify two core design principles which address these issues: a) represent the Brownian path with a basis which carries higher-order local information, and b) train with an objective that forces the network to use the full representation of the path, not just the first order. To address this, we leverage the shifted Legendre polynomial expansion with a Wong-Zakai styled objective, unlocking derivative-based losses. Finally, we propose a recipe for fine-tuning large pre-trained flow models into strong stochastic flow maps. We ablate through this design space on several toy problems and use our findings to transform a large flow model into a strong stochastic flow map on ImageNet, achieving roughly a three-fold improvement in FD-DINOv2 over SSFM at four NFE.

Automatic phenotype information extraction plays an important role in rare disease diagnosis, genotype-phenotype analysis, and precision medicine. However, accurately identifying phenotype concepts and determining their associations with specific patients remain challenging, particularly in full-text biomedical articles involving multiple patients and complex clinical descriptions. This paper presents an overview of CHIP 2026 Shared Task 1, which aims to promote the development of methods for patient-specific phenotype concept recognition from full-text biomedical literature. The task consists of two subtasks: full-text phenotype concept recognition and patient-specific phenotype association. For this evaluation, we used 200 annotated full-text biomedical articles from the PatientPheX dataset, containing 506 patients, 15,193 phenotype mentions, and 5,656 patient-phenotype associations. A total of 455 teams registered for the competition, with 189 and 103 valid submissions received on the Phase A and Phase B leaderboards, respectively. The best-performing system achieved an overall score of 0.7389 on the final test set, outperforming the baseline by 8.37 percentage points. The top-performing teams explored diverse approaches combining biomedical language models, large language models (LLMs), and ensemble strategies. The evaluation results highlight both the potential of these approaches and the remaining challenges in patient attribution and phenotype normalization from full-text biomedical literature. The post-evaluation phase remains open at https://tianchi.aliyun.com/competition/entrance/532512.

Deep neural networks deployed on smart glasses must adapt to changes in the distribution of incoming data while adhering to strict privacy and device constraints. We present No Frames, Just Frequencies, a bandwidth-efficient source-assisted adaptation framework that characterizes distribution shifts using compact frequency descriptors computed on-device. Rather than transmitting camera frames, the glasses compute and send frequency descriptors to a host device that contains a replica of the model. The host device maps the received frequency descriptor to a distribution shift operator that synthesizes out-of-distribution data and adapts a model replica using a pre-defined adaptation protocol. We evaluate the framework on mixtures of ImageNet-C corruptions using convolution-based (ConvNeXtV2-Femto), self-attention-based (DeiT-Tiny), and hybrid (TinyViT-5M) vision models, on person classification (Visual Wake Words), and on object detection (MS-COCO). Our results reveal that adaptation driven solely by the transmitted descriptors recovers $93-95\%$ of the supervised oracle adaptation gain on all three architectures, $82\%$ on Visual Wake Words and $94$--$98\%$ on MS-COCO. Notably, adaptation quality degrades gracefully as the descriptor resolution decreases from $224 $ to $28$ pixels, consistently across architectures, enabling bandwidth reductions from $213\times$ to $784\times$ across the frequency descriptors.

In automation pipelines, a GUI agent's confident success report can let downstream stages proceed even when the recorded execution trace does not show that the agent produced the requested outcome. We call this case a dangerous misreport: an unhedged success claim that the trace does not support as causal accomplishment during the run. We study report--trace auditing, which checks an agent's completion claim against an independently recorded trace. A controlled aware/blind intervention reveals self-report leakage: the agent's own self-report changes how completion is judged. The same judge reads the same task and trace; only that self-report is included or withheld. Withholding it improves trace-only dangerous-misreport recall by 43.4 percentage points with Qwen3.5-35B-A3B and 13.3 percentage points with DeepSeek-284B-A13B, with the same trend on held-out Batch 2. This motivates report isolation. The report records what the agent claims, but it must not influence the trace-state judgment. We implement this separation with a structured auditor that separates claim interpretation from claim-blind trace analysis and applies deterministic rules to extracted evidence. Because evidence extraction is separated from the policy that consumes it, the same audit record can be replayed under alternative policies without new model calls. The auditor achieves F1 0.75--0.78 on the 465-record audit-target set and 0.77--0.81 on held-out Batch 2, outperforming claim-aware and prompt-blind controls in recall and F1. Under audit-stage gating, where flag and review both block automatic pass, it blocks 74--90% of dangerous misreports at 45--56% inspection load, or 16--19% of the full in-scope stream after coarse status triage.