Papers

We study the circle-relation subspace $O_n^\circ(V)$ in the defining relations of the level-$n$ Zhu algebra of a vertex operator algebra $V$. We assemble the residue relations generating $O_n^\circ(V)$ into generating functions and obtain an explicit recursion that replaces every sufficiently deep mode by retained modes with indices at least $-2n-1$, modulo $O_n^\circ(V)$. This recursion is encoded in a left truncated field $Y_L$. Since such mode replacements are valid only in the leftmost position of an operator product, we prove the replacement, derivative, commutator, and one-step iterate identities for $Y_L$ needed for finite-length calculations. For a finite homogeneous strong generating set $S$, we define the finite-length circle subspaces $O_n^\circ(N,S,V)$ and their plus subspaces. We formulate a finite-length conjecture under an explicit auxiliary-degree hypothesis and a separate termwise conjecture for the correction residues of one-sided iterates. We prove reductions to the $k=0$ strip and to last vacuum mode $-1$, recover the length-one relation spaces, and give explicit Virasoro computations at word length two for $n=1$ and $n=2$. The $Y_L$ identities are the input for the companion paper (Part II).

Hydrodynamic fluctuations in the primordial plasma generate a gravitational-wave (GW) background that grows like $f^3_\mathrm{\scriptscriptstyle 0}$ in terms of the current frequency. However, within the ($10^{-4}_{ } ... 10^4_{ })\,$Hz band of upcoming interferometers, the modes likely re-entered the Hubble horizon when the hydrodynamic noise was no longer at its maximal value. To address this situation, we define a model-independent transfer function for GWs generated by hydrodynamic (or other white-noise) fluctuations at any time. Employing appropriate variables, we show how this transfer function, together with the simpler vacuum one, can be determined numerically for a general expansion history. The results are illustrated for Standard Model embedded warm inflation scenarios, including both the vacuum and hydrodynamic parts. Following previous estimates, we find that a pre-exit-generated hydrodynamic tail becomes almost flat below $(10^5_{ } ... 10^7_{ })\;$Hz. Its small-$f^{ }_\mathrm{\scriptscriptstyle 0}$ intercept remains finite, and can then moderately influence observational constraints through the tensor-to-scalar ratio ($r$).

Text-to-speech (TTS) based on masked diffusion models (MDMs) achieves competitive zero-shot speech synthesis performance. Unlike autoregressive (AR) models, MDMs define a corruption process that maps clean speech-token sequences to masked maps and learn a reverse denoising process to recover the reference tokens. However, MDMs commit to intermediate predictions that cannot be revised in subsequent inference steps. Consequently, prediction errors accumulate and propagate, degrading synthesized speech quality. We present Revise-TTS, a single-stage framework that applies hierarchical binary quantization (HBQ) and generative refinement networks (GRN) to zero-shot speech generation. First, we replace finite scalar quantization (FSQ) with HBQ to introduce a scalable binary representation for speech tokens with an explicit coarse-to-fine structure. Second, GRN is trained to reconstruct the token sequence from a mixture of ground-truth and random tokens. During inference, iterative refinement enables self-correction by randomly discarding a small proportion of previously predicted tokens with new predictions. Experiments and ablation studies demonstrate that Revise-TTS achieves competitive zero-shot speech synthesis performance using HBQ. It can further be extended to any MDM-based TTS task as an alternative training and inference method.

Model signing authenticates the full-precision weights W an LLM is released with, yet users run a quantized artifact Q whose producing transformation is not uniquely determined under current disclosure. We separate the non-environment inputs from the execution environment E and measure a quantity that, to our knowledge, has not been measured before: the Independent Verifiability Rate (IVR), the fraction of a popularity-sampled GGUF ecosystem that a third party can re-derive bit-for-bit. Across the 200 most-downloaded Hugging Face GGUF repositories (198 in scope), no artifact or model card records the build environment, so under a recorded-inputs protocol the IVR is 0/198. Yet the artifacts are not irreproducible. On four in-sample repositories that disclose their inputs, the card's information alone reproduces no Q4_K_M file, but two coordinates the card omits - the build's floating-point-contraction setting and, for hybrid models, the per-tensor quantization recipe - re-derive 7/8 target-quant files bit-for-bit on x86 (5/8 on arm64); the fp16 re-derives from the provider's checkpoint, but the calibration imatrix must be shipped and hash-bound, since recomputing it did not match. An older release additionally pins the instruction set and did not re-derive. Reproducing the divergence reported with llama.cpp PR #25353, we show that E is a necessary coordinate and that one compiler setting yields a unique output across the tested platforms at negligible quality cost (+0.06% perplexity). As a security characterization, benign floating-point non-determinism separates from uniformly placed nibble-flip tampering above a detection floor that grows with the environments a verifier must allow, but a targeted embedding edit evades the statistic, so the proposed protocol rests on exact re-derivation; adversarial-fp16 attacks lie outside it. We release the audit data and a re-runnable harness.

Data augmentation (DA) plays a vital role in improving model performance in low-resource or limited-supervision scenarios within Natural Language Processing (NLP). Existing DA methods, such as synonym replacement and back-translation, have demonstrated effectiveness on the word or sentence level; however, they frequently neglect discourse-level coherence and logical flow, which are essential for complex tasks dependent on inter-sentential relationships. In this paper, we propose a structure-preserving document-level data augmentation framework based on large language models (LLMs) and document-level discourse structure parsing. Our approach identifies rhetorical relations between sentence pairs and extracts key phrases, which are then replaced with semantically coherent but topically shifted content while preserving the original discourse structure. Experimental results on text summarization and question answering show that the proposed augmented data provides complementary supervision and improves combined-data training across the evaluated tasks. Comparisons with a standard RST parser and controlled ablations further support the proposed discourse representation.

We study how retained ownership supports the value of an asset that pays no dividends and is held only for resale. A sponsor chooses hidden maintenance that determines market survival, bears its full cost, and captures only the benefit accruing to its retained tokens. A positive stationary equilibrium exists if and only if retention meets an endogenous minimum. Under additional curvature restrictions, retention above this threshold supports two positive stationary equilibria. Conditional on survival, prices below the upper equilibrium converge to the lower one while the upper price is the largest feasible initial valuation. Greater retention expands the feasible price interval. Credible news that future retention will be below the minimum causes an immediate collapse through backward unraveling. The model identifies a governance channel through which insider unlocks can affect valuation without changing total token supply or fundamental cash flows.

Retrieving useful context for a robot requires matching its task, execution stage, and the continuation that should follow. We introduce Robota-Embed, to our knowledge the first vision-language-action (VLA) embedding model to jointly encode all three modalities in both current-state queries and recorded future continuations. Predictive retrieval training aligns these endpoints, enabling historical continuations to serve as context at inference time. We construct a multi-source corpus of 1,695,684 cleaned training anchors and introduce Robota-Embed-Bench, with 4,000 queries and a shared 8,000-entry gallery. Robota-Embed achieves 47.33% clip-level and 81.77% episode-level R@1, exceeding action-free Gemini Embedding 2 by 22.09 and 33.47 percentage points. Controlled retrieval studies show that removing native action degrades retrieval more than removing vision. We study two complementary applications. Embed2ICL uses retrieved experience for training-free and training-based policy conditioning, increasing success from 47.0% to 56.3% in DROID-Sim and from 86.4% to 92.8% on the LIBERO difficult subset, while improving task progress on real robots. Embed2Evolve uses execution feedback to grow task-local memory and refine context use around a frozen GPT-6 Astra policy. Across five ManiSkill tasks, it achieves 68% success, compared with 24% for GPT-6 Astra Direct and 33% for a harness independently evolved with Omni-Embed-Nemotron-3B under the same protocol. Relative to the latter, it uses approximately 79.5% fewer policy tokens per test rollout, counting both successes and failures. These results show that a unified, action-aware retrieval space can support robot in-context learning and task-local context evolution.

Real-time monocular human pose estimation is an important sensing component for human-centered intelligent systems. While recent camera-based methods have achieved impressive accuracy, fundamental challenges persist, particularly when visual observations are incomplete, such as during lower-body occlusion or short out-of-frame intervals. In these challenging but common scenarios, vision-only systems often struggle to reconstruct physically plausible lower-limb motion and reliable foot-ground contact. Due to the highly unobtrusive and wearable nature of smart insoles, this paper investigates whether simple plantar pressure and acceleration sensors can effectively resolve these visually induced limitations. To systematically investigate this, we first collect a synchronized multimodal dataset and then propose a part-decoupled temporal fusion framework. The proposed model fuses foot-sensing features into the lower-body branch to compensate for incomplete visual evidence, while a lightweight coordinator exchanges information between the upper- and lower-body temporal representations to maintain whole-body consistency. The framework is evaluated with three representative real-time visual backbones based on 2D keypoints, 3D landmarks, and SMPL-based human mesh recovery. We evaluate our framework on three representative real-time visual backbones. Quantitative experiments demonstrate that incorporating simple foot sensors consistently improves pose accuracy, motion smoothness, and gait-related measurements over vision-only baselines, with the most significant gains observed under occlusion and short visual-dropout conditions. Ultimately, our results demonstrate that unobtrusive foot sensing provides vital and complementary physical observability when camera observations are temporarily unavailable.

Rapid prediction of pressure-driven flow in microfluidic porous media is important for experimental design, tracer-test interpretation, and digital-twin development for subsurface energy and critical mineral applications. High-fidelity simulations resolve geometry-controlled flow but are computationally expensive across large geometry ensembles. Here, we introduce GeONet, a Geometry-Encoded Operator Network that predicts steady-state pressure fields by jointly representing material properties, spatial coordinates, and solid–void geometry. GeONet extends the Branch–Trunk architecture with a dedicated Leaf Network encoding the binary domain mask through multiscale features. The Branch Network represents porosity and permeability, the Trunk Network combines coordinates with local geometric descriptors and random Fourier features, and a gated DeepMixer integrates these representations to predict pressure. GeONet was trained and evaluated using 5,000 PFLOTRAN simulations of heterogeneous Reduced Complexity Platform domains, comprising nearly 30 TB of data. Comparisons with pore-scale Stokes solutions across three geometries yielded Pearson correlations of 0.987–0.991 and $R^2$ values of 0.974–0.982, supporting PFLOTRAN as reference data. Under matched partitions and two-stage hyperparameter optimization, GeONet substantially outperformed a boundary-condition-aware DeepONet baseline. Across three held-out domains, GeONet achieved NRMSE of 0.0381–0.0500, MAPE of 1.76–2.59%, and $R^2$ of 0.9673–0.9822, versus 0.1483–0.2040, 7.81–10.45%, and 0.5658–0.7126 for DeepONet, respectively. On three additional unseen geometries, GeONet achieved NRMSE of 0.0448–0.0676 and $R^2$ of 0.9511–0.9770. These results demonstrate improved accuracy, pressure-range preservation, and cross-geometry generalization through explicit geometry encoding.

Streaming Inverse Text Normalization (ITN) is vital for converting spoken-form outputs from streaming Automatic Speech Recognition into formatted written text. Existing streaming ITN methods rely on hybrid systems combining neural tagging with expert-crafted finite-state transducer rules, limiting scalability across domains and languages. While end-to-end models offer superior scalability by learning directly from data, standard encoder-decoder architectures are inherently non-streaming due to global attention. In this paper, we propose an efficient streaming end-to-end ITN system, adapting a pretrained text-to-text model to leverage its robust linguistic knowledge. To enable streaming, we introduce architectural adaptations, a specialized training strategy, a Read-Tag-Write decoding policy, and inference optimizations. Experiments on Vietnamese datasets show accuracy comparable to non-streaming baselines, outperforming hybrid methods while satisfying real-time latency requirements.

The rise of quantum computing presents substantial challenges to the security of Peer-to-Peer (P2P) energy trading systems. This paper introduces an innovative mutual authenticated session key agreement protocol aimed at protecting edge-IoT P2P energy trading communications in the quantum era. Each peer is an IoT-enabled smart metering device that is registered by an edge registration authority and acts as both producer and consumer. The proposed protocol incorporates CRYSTALS-Kyber's quantum-resistant cryptographic suite to guarantee the authenticity and confidentiality of negotiated session keys in the presence of quantum threats. Furthermore, it leverages the Hedera Hashgraph public blockchain, which is recognized for its quantum-resilient design, to ensure transparency and support efficient revocation mechanisms. Extensive comparative analyses have been conducted to demonstrate the protocol's superiority over existing alternatives, including both traditional schemes based on classical cryptography and emerging early-stage post-quantum approaches, in terms of performance and security features. By combining robust security and high efficiency, the proposed protocol fosters the development of resilient decentralized energy markets, paving the way for secure and sustainable energy distribution in the future.

Digital-domain self-interference cancellation is essential in in-band full-duplex radios after passive suppression and analog cancellation. It is particularly challenging when the desired signal is strong relative to the residual self-interference. In this regime, conventional iterative digital cancellers update their coefficients from desired-signal-contaminated received samples and may suffer substantial performance degradation. This paper reformulates digital-domain self-interference cancellation as an operator-learning problem. The proposed framework learns the temporal evolution of the residual self-interference channel from historical observations and uses the learned operator together with the known local transmit signal to reconstruct the current residual self-interference. A Transformer neural operator is used to realize this framework. The resulting method is trained in a desired-signal-free setting and then applied in the presence of a desired signal. A continuity-based analysis characterizes the sensitivity of the learned operator to the resulting input perturbation. Numerical experiments in static and time-varying channels, including time-varying SIR conditions, demonstrate stable cancellation performance over a wide range of interference levels. At 30 dB SIR, the proposed method improves the cancellation performance by 5–15 dB over the considered conventional methods.

Sparse matrices, in which most entries are zero, are central to scientific simulation, graph analytics and machine learning, and their multiplication with dense inputs is a core parallel kernel. In many applications the matrix stays fixed while only a few input rows change per step, so the result can be updated from those rows alone, which we call the dirty set. Sparse kernels often add a second optimization based on the inspector-executor principle, in which a tiled schedule is prepared once and reused across calls. However, when this reuse remains profitable as the dirty set moves is not well characterized. In this paper, we show that retained tiled execution rarely beats direct CSC in incremental sparse matrix multiplication, although incremental execution itself remains highly profitable. Across 21 SuiteSparse matrices and controlled patterns of motion, a direct column-wise incremental update is 2.1-8.2x faster than the best of eight full-recomputation codes, among them Intel oneMKL and PyTorch. Yet the same direct update outperforms a tiled executor that retains tile preparation and caches assembled schedules in 779 of 837 workloads. Separating reuse into per-tile preparation and cached schedules shows that the benefit comes from preparation; extra cache capacity seldom helps because a tile set rarely returns after a different one has intervened. Tiling is faster only when tiles merge many entries into each output write, a structural property that we predicted before measurement and confirmed on four held-out matrices of the same family. These results show that the structure of the matrix and the motion of the dirty set, rather than the sophistication of the schedule, determine whether reuse is profitable. We anticipate that our methodology and guidelines will help developers decide when to build schedule reuse into incremental sparse runtimes.

We examine the value of information in extensive-form decision problems: the largest multiplicative gain that a decision maker can obtain from an information refinement, taken over all payoff environments. This measures the structural importance of information disentangled from any particular payoff specification, and generalizes the notion of value of recall recently introduced by Berker et al. (AAAI'25). We establish a unifying characterization of the value of information under any information refinement, uncovering an intriguing connection to the maximum fractional chromatic number over a certain family of conflict graphs. We then employ our characterization to elucidate the complexity of two central information refinements. First, for the value of perfect information in perfect-recall decision problems, we find that the relevant conflict graphs are cographs, so the fractional chromatic number collapses to the clique number. Even though the clique number is tractable for a given cograph, we show that the additional layer of optimization (over the family of graphs) makes computing the value of perfect information NP-complete. On the positive side, we prove fixed-parameter tractability in terms of the degree of imperfect information. Second, for the value of recall in imperfect-recall decision problems (with no absentmindedness), we first show that the value of recall can realize the fractional chromatic number of any graph. Interestingly, this implies that, unlike the value of perfect information, the value of recall can take fractional values. Building on this reduction, we then establish S2P-completeness. This addresses questions left open by Berker et al. (AAAI'25)—both on the characterization and the complexity of the value of recall—and shows that computing the value of recall is strictly harder than the value of perfect information.

Mandlebrot's coastline paradox reframes measurement as a joint property of an object and its ruler. Here we ask how observational resolution shapes the apparent structure of neural computation. We introduce a scale-indexed decomposition that separates the predictive information captured by direct dependencies from the residual outside that model family. Exact constructions show that projection can convert higher-order dependence into pairwise structure. Projection can increase the direct fraction of surviving information even as total predictive information decreases. A schematic branch-to-soma cascade grounds these effects in local coincidence detection, saturating integration and somatic mixing. Resolving organized nonlinear subunits improves approximation and raises effective representational rank. In a nested hierarchy, these gains occur without increasing information about the target. This distinction separates information availability from access by a restricted readout. A point-neuron or direct population model therefore establishes an effective description at a specified scale. It does not uniquely identify the finer computation that generated the observed relation. Which resolution is adequate depends on the observable and perturbation of interest.

Velocity space discretization is a primary source of computational cost in deterministic methods for multiscale gas flows. A fixed velocity grid often retains redundant kinetic degrees of freedom (DOFs), particularly under near equilibrium conditions. To address this issue, this work develops a nonequilibrium-capturing formulation based on the unified gas kinetic scheme (UGKS), in which the number of kinetic DOFs adapts locally to the degree of nonequilibrium. A discrete orthogonal basis is constructed for numerical implementation, and the velocity distribution function is decomposed into an analytical local equilibrium and a basis expanded nonequilibrium component. The retained, locally adapted DOFs are evolved directly, substantially lowering computational overhead while preserving the discrete conservation of mass, momentum, and energy, as well as accurately capturing nonequilibrium states. The formulation retains the core multiscale coupling of molecular transport and collision inherent to UGKS, providing explicit and implicit schemes for unsteady and steady flows, respectively. Numerical tests across a wide range of flow speeds and rarefaction regimes (from continuum to highly rarefied) demonstrate close agreement with benchmark solutions. For unsteady flows, the explicit scheme achieves speedups of 12.64 to 58.50 over the standard UGKS. In steady state tests, only about $2\%$ of the original kinetic DOFs are required in continuum regions, accelerating each implicit iteration by a factor of 33.87. Even in the most rarefied cases, the scheme captures nonequilibrium behavior using only about $27\%$ of the original DOFs.

The gradient flow exact renormalization group (GFERG) is a formulation of the exact renormalization group for gauge theories that provides a renormalization group flow preserving the gauge Ward–Takahashi identity exactly. In this paper, we nonperturbatively solve the GFERG flow equation for the simplest nontrivial gauge-invariant Wilson action, introduced in a previous paper, for $D$-dimensional quantum electrodynamics with~$N_f$ flavors (for all~$D$, flavors are counted in units of four-component spinors). For $D=3$ and~$D=2$, a nontrivial infrared fixed point exists even for~$N_f=1$, and the structure of the critical exponents at the fixed point is qualitatively the same as that in the large-$N_f$ limit. We also obtain the gauge-field propagator in the continuum limit; for~$D=2$, it reproduces the dynamical mass of the Schwinger model within a few percent. To the best of our knowledge, apart from the large-$N_f$ approximation, this is the first example in which the Wilson renormalization group in a gauge theory is solved nonperturbatively while manifestly preserving gauge symmetry.

Sequential conformal prediction needs to decide which past errors should calibrate uncertainty for the current target. Recency and online error feedback are useful when recent observations remain relevant, but the recent past can be a poor reference when comparable states recur at nonadjacent times, with the recurrence pattern generally unknown. We propose State-Matched Sequence Conformal Prediction (SMSCP), which constructs an explicit one-sided local-linear state from pre-target information, uses state similarity to select a matched reference set of past calibration scores, and then applies the ordinary unweighted conformal order statistic. In contrast to existing approaches, SMSCP uses state similarity to identify calibration observations relevant to the target score quantile. This motivates a theoretical analysis of whether nearby states have comparable target score quantiles under temporal dependence. We establish control of the resulting conformal cutoff and characterize predictive information not captured by the chosen state. Simulations show that state matching is most effective when score-relevant states recur across nonadjacent times, while revealing settings in which its advantage weakens or reverses. Across solar irradiance and bike-demand benchmarks, SMSCP improves interval score over recency, weighted, adaptive, and HopCPT comparators, and achieves coverage closer to the nominal 0.90 level than ResCP on five of six benchmarks.

In this work, we study spectral functions of heavy and exotic mesons in a self-consistent Einstein-Maxwell-dilaton model at finite temperature and chemical potential, where the dilaton is allowed to back-react on the geometry. A non-quadratic dilaton profile is introduced to reproduce the nonlinear Regge trajectories of heavy-flavor mesons, and the corresponding spectral functions are computed for charmonium, bottomonium, tetraquark, and hybrid channels. In all cases, increasing either temperature or chemical potential suppresses and broadens the quasiparticle peaks, signaling sequential in-medium dissociation; for each radial family, the 2S-like state melts before the 1S-like state. The predicted melting temperatures obey the mass ordering $T_{\rm melt}^{\pi_1(1400)}<T_{\rm melt}^{J/\psi}<T_{\rm melt}^{Z_c(3900)}<T_{\rm melt}^{\Upsilon(1S)}$, with $T_{\rm melt}\simeq 1.08,\ 2.96,\ 4.44$, and $5.93\,T_c$, respectively, so heavier states remain bound up to higher temperatures. These results provide quantitative, testable predictions for heavy-ion experiments probing tetraquark and hybrid mesons in hot and dense QCD matter.

Response-spectrum matching is an underdetermined waveform inverse problem in which spectral compatibility must be achieved without uncontrolled changes in other engineering-relevant ground-motion characteristics. This study formulates the problem as differentiable waveform inverse design using a fixed SDOF dynamics engine and configurable optimization modules. An envelope-guided parameterization defines the correction waveform, while a hard projection enforces zero correction-induced terminal velocity and displacement. Differentiable SDOF responses then provide gradients for waveform optimization. Spectral fidelity and Arias-intensity accumulation are treated as adaptive objectives. Spectral feasibility, PGA, total Arias intensity, and displacement-path requirements are incorporated as constraints. The differentiable spectrum operator agrees with an exact solution of the equation of motion to within 0.002%. The complete five-damping configuration achieves a maximum spectral error of 5.00% over 375 ordinates. On a common benchmark, maximum errors of 2.61% and 8.47% are obtained for single- and multi-damping matching, respectively, while retaining the additional waveform constraints. Three-component applications further show that different engineering acceptance requirements can be accommodated through module reconfiguration without changing the forward dynamics.

Wrist-worn inertial measurement units (IMUs) provide a common sensing modality for tasks ranging from activity and gesture recognition to pose estimation, inertial odometry and clinical motor assessment. Existing wearable benchmarks evaluate performance across multiple tasks, but leave open how supervision for one task affects transfer to others and whether these tasks benefit from a shared representation. We present IMU-Bench, a benchmark of ten tasks from eight datasets for studying IMU multi-task performance and cross-task transfer. We compare six self-supervised methods pretrained on unlabeled recorded and simulated wrist motion, evaluating classification, pose and velocity estimation, and IMU-language retrieval. We find that our masked autoencoder, InrtlWrist, achieves the best or joint-best frozen-probe performance on nine of ten tasks, and fine-tuning improves on training from scratch on nine of ten tasks, including full-body pose, odometry and retrieval. Across transfer and joint-training experiments, supervision that benefits one target can reduce performance on others, with supervised sources transferring within kinematic estimation while, under frozen probing, none approaches the pretrained encoder on language retrieval, and joint-training gains that come at a cost to the paired task. These findings distinguish broad downstream usefulness from complementary task supervision. Code, task cards and a leaderboard are publicly available.

Tangles of soft filaments are ubiquitous, from living worm blobs to the cables in a desk drawer. How do such tangles, held together by friction and entanglement, respond when one embedded filament is shaken? We suspend disordered tangles of tunable cohesion and ordered stitched meshes from a single thread (the tether) and oscillate it vertically at fixed amplitude. Extraction peaks at an intermediate drive frequency ($\sim 15$ Hz) and declines at higher frequencies, although the peak drive acceleration grows quadratically. Here we combine experiments, high-speed imaging, and a tension-only spring-mass model to show that this non-monotonic response is set by timing rather than resonance. The extraction window is bounded below by a slackening threshold, where the peak drive acceleration exceeds gravity and the tether, slack for part of each cycle, creates a tensile impulse each time it snaps taut. The window is bounded above by a transmission crossover, where the transmission time across the tangle exceeds a quarter of the drive cycle, so that the tangle no longer responds as a whole and the drive's work dissipates internally. The slackening threshold depends only on the drive and gravity, while the transmission crossover is set only by the tangle's size and transmission speed. By tracing extraction to the overlap of impulse creation and impulse transmission, we identify a generic principle of driven filamentous matter and suggest frequency-selective routes to its disassembly.

Sim-and-real co-training is an effective way to scale robot learning by augmenting limited real-world demonstrations with simulation data. However, existing approaches typically treat simulators as fixed data sources, relying on uniform generation, offline retrieval, or manual parameter alignment by domain experts. We formulate Active Generative Simulation: optimizing a simulator's data-generation distribution to maximize downstream real-world performance. To solve this problem, we propose MetaGenSim, a meta-learning framework that trains an auxiliary rater network via meta-gradients to automatically discover high-utility data-generation distributions within a user-specified simulation parameter space, without manual parameter alignment. Across simulation and real-world manipulation tasks, MetaGenSim outperforms uniform generation and retrieval baselines, achieves a 150% relative improvement in real-world full-task success over uniform generation, and matches expert manual alignment. We further show that distributions curated with a lightweight proxy policy transfer to VLA fine-tuning, and that MetaGenSim can exploit partial alignment through atomic skill stitching and object-receptacle composition when exact sim-and-real task alignment is unavailable. Project website: https://meta-gensim.github.io

Agricultural systems make decisions over heterogeneous entity types, textual judgments, and knowledge-question options. Generating a textual answer adds decoding and extraction steps to decisions that ultimately require a candidate choice. We introduce AgriJev, a family of adapted Qwen3.5 models at 0.8B, 2B, 4B, and 9B. A shared interface conditions on agricultural text, a judgment question, and candidate definitions, returning scores and relative probabilities in one forward call. We prepare source-aware training records and frozen evaluations from existing agricultural resources, retaining original groups and annotation provenance. Comparisons include original Qwen3.5 models and matched agricultural SFT models, both evaluated through native answer generation. Across 17 external test views, AgriJev-9B reaches 75.44% mean accuracy, exceeding the original and SFT generators by 5.89 and 3.79 percentage points. It outperforms the original generator on 13 views and narrows the mean gap to hosted Jev 1.13.0. On a fixed A100 workload, its median scoring latency is approximately half that of same-scale answer generation. Task-dependent gains and failures characterize a practical choice among agricultural model weights and inference contracts.

We study factorized gradient descent (FGD) for overparameterized asymmetric matrix sensing with the balancing regularizer $\frac18\|L^\top L-R^\top R\|_F^2$. An exact indefinite lifting of the population objective decomposes the dynamics of FGD into positive, negative, and null spectral subspaces. The first two components have linear restoring terms, whereas the null component contracts only quadratically. For independent isotropic sub-Gaussian sensing matrices and independent centered sub-Gaussian noise with scale proxy $\sigma$, we prove the local bound \[ \|L_tR_t^\top-X^\star\|_2 \lesssim\max\left\{\frac{1}{\eta t+1/\sigma_r}, \kappa\sigma\sqrt{\frac{d_1+d_2}{n}}\right\}, \] under a sufficiently small initialization error, the number of linear measurements $n\gtrsim\kappa^2(d_1+d_2)\max\{k^2,\sigma^2/\sigma_r^2\}$, and step size $\eta\lesssim\sigma_1^{-1}$. Here $\sigma_1$ and $\sigma_r$ are the largest and smallest nonzero singular values of the target $X^\star\in \mathbb{R}^{d_1\times d_2}$ respectively, and $\kappa=\sigma_1/\sigma_r$. Furthermore, a population trajectory was constructed to have tight product error $\Theta(1/t)$. These results provide a local rate characterization for ordinary gradient descent for overparameterized asymmetric matrix sensing with balancing regularizer.

Post-training often improves the performance of large audio-language models on audio understanding and reasoning benchmarks. However, these improvements may reflect adaptation to specific question formats, such as multiple-choice questions, rather than genuine advances in audio understanding and reasoning capabilities. To investigate what LALMs actually learn from post-training, we conduct controlled experiments that disentangle improvements in audio understanding from adaptation to specific question formats. Across different models and post-training methods, we find that much of the reported improvement persists even without informative audio, but largely disappears when the question format changes. These findings suggest that post-training often teaches models to exploit task-specific textual cues and answer patterns rather than developing generalizable audio understanding and reasoning capabilities.

Multimodal in-context learning (ICL) is an important capability of multimodal large language models (MLLMs). Its performance is closely tied to the demonstrations it uses, making demonstration retrieval a core problem for ICL. Common methods use fixed scoring methods to retrieve all the demonstrations at once, which may introduce redundancy and noise. Context-aware retrieval methods have emerged to address this problem by allowing incremental retrieval from a growing context. However, none of them lets the MLLM learn how to judge what new demonstrations it needs and then carry out progressive retrieval on its own. To this end, we introduce ActICL, which trains an MLLM agent to acquire and use an ICL context via tool-augmented reasoning. An ActICL agent analyzes its own information gaps during iterative reasoning and tool-use actions. Based on these gaps, it performs targeted retrieval to acquire a set of initial candidates. It can then use the existing context to further filter the candidates and refine the context, or it can iteratively perform a new retrieval. When the agent judges that the context is sufficient for the task, or when the resource limit is reached with the user request met, it sends the acquired context to an answering model, either the agent itself or another MLLM. We use a two-stage post-training procedure that jointly optimizes the agent's acquisition and answering. With an adaptive number of demonstrations, ActICL improves average performance by 11.6 percentage points over standard similarity-based retrieval across seven benchmarks and three MLLMs, and by 4.4 points compared to the strongest prior baseline. The contexts acquired by ActICL agents can also be used effectively by frontier models. When GPT-5.6-Sol performs 4-shot ICL, ActICL's contexts improve its performance by 3.5 points over standard retrieval, averaged over three benchmarks.

Questions about an agent's history often require aggregating events, which retrieval alone does not resolve. We combine source-preserving evidence access with executable aggregation. Trajectory Memory (TM) retains the original trajectory and constructs a source-addressable reading view at question time. TallyMem augments it with a reusable event ledger, model-generated event specifications and deterministic execution. With a commit-specific adaptation on 776 questions from 99 held-out MINTEval repositories, TallyMem raises Qwen3-32B exact match from 35.2% to 49.0%, a 13.8-point gain over TM; gains span 12.8-13.9 points across three readers. A controlled probe with event sets parsed independently from raw logs shows an 11.0-point gain, from 82.8% to 93.8%, without routing. On full AMA-Bench, the routed system yields a smaller 1.3-point gain, from 69.8% to 71.1%. These results support executable aggregation on tasks that require combining events, with benefits dependent on the task and event specification, the largest residual system-error category in our diagnostic analysis.

Quantifying the biological effects of radiation is essential for radiotherapy and radiation protection. While research has centered on statistical analyses of experimental data, this approach often fails to elucidate underlying mechanisms, limiting the development of generalizable models. Describing time dependent phenomena, such as sub-lethal damage (SLD) recovery during fractionated irradiation, presents a particularly significant challenge for existing static models. To address this, we developed a mechanistic mathematical model that extends the classical multi-target, single-hit model by introducing explicit recovery transitions among damage states, allowing the simultaneous description of radiation-induced damage accumulation and biological recovery. We applied this model to Elkind and Sutton's classical split-dose experimental data to determine its parameters. The model reproduced the major features of split-dose survival curves for treatments with 0-, 2.5-, and 23-hour recovery intervals. We visualized the temporal changes in damage state distributions within the cell populations. These simulations provided a process-level interpretation of the characteristic shoulder-like response appearing after an inter-fraction recovery interval. Furthermore, the estimated recovery half-life was about 120 minutes, which is broadly consistent with reported timescales of DNA double-strand break repair in mammalian cells, supporting the biological interpretability of the model parameters. The proposed framework provides a simple, interpretable description of radiation-induced cell death dynamics, including SLD recovery, and may serve as a basis for the improved mechanistic modeling of radiobiological effects under different irradiation protocols.

Long-term memory remains a fundamental bottleneck for large language model (LLM) agents across sessions. Existing systems typically adopt a write-then-retrieve paradigm, storing context-dependent dialogue fragments and retrieving them via nearest-neighbor search in a continuous embedding space. This design yields unstable access units and systematically misses logically related but lexically distant evidence. We propose MemCode, a long-term agent memory framework that couples retrieval-oriented writing with discrete associative retrieval. MemCode rewrites dialogue into self-contained atomic memory units (AMUs), maps AMUs into a learnable codebook with isomorphic semantic quantization (ISQ), and uses codebook co-occurrence topology as a retrieval path complementary to vector search. For long-term memory stores, MemCode separates online atomic writes from asynchronous offline consolidation. On LongMemEval-S and LoCoMo with GPT-4o-mini and Qwen3-30B-A3B-Instruct-2507 backbones, MemCode achieves the best overall accuracy while keeping per-dialogue memory construction within roughly 40~130 seconds, making it 3~21 times faster than LightMem and over an order of magnitude faster than A-MEM and MemoryOS.