Papers

Convolutional neural networks (CNNs), used widely and successfully in audio enhancement, capture long-range time-frequency dependencies only indirectly, through successive convolution and pooling. Here, we present U-PAST, a hybrid transformer-U-Net architecture that addresses this limitation through self-attention dependency-modeling in the complex spectrogram domain. U-PAST tokenizes a complex STFT representation, similarly to the magnitude spectrogram tokenization of the Audio Spectrogram Transformer (AST), applies a multi-layer transformer encoder, and reconstructs the enhanced complex spectrogram with a U-Net-style decoder. We evaluate four architectural variants with between 1.17M and 2.40M parameters on the DNS Challenge, VoiceBank-DEMAND, and LibriMix corpora under matched, acoustic mismatch, and two-dataset mismatch conditions. U-PAST attains the best SI-SDR of any evaluated model under acoustic mismatch and closely trails substantially larger convolutional and time-domain baselines by 0.26 dB to 0.63 dB SI-SDR under the remaining three conditions while achieving the strongest perceptual (DNSMOS) quality under dataset mismatch. The largest evaluated configuration, U-PAST-H (2.40M parameters), is consistently the strongest variant of the family, offering an attractive performance-to-cost trade-off at a small parameter footprint.

Recent work uses large language models to generate systems code from specifications, treating the specification as the durable artifact and the implementation as disposable. Regenerating the implementation specializes it to each workload and device. However, that work lives in a forgiving setting: a component's externally visible effects, its writes and device commands, are recoverable, and the generator is honest, so trust is discharged by re-execution. We target the unforgiving setting: systems code whose effects are irreversible, produced by a generator that may be adversarial. There, re-execution cannot check an effect after the fact, and a proof fails silently when its assumptions do. We take the position that the only safe way to operate here is to deny the generated code the authority to act. The generated code only plans, while a fixed trusted mediator owns every effect and performs one only when the specification would have produced it. Because the guarantee lives in the mediator, not the code, it survives regeneration. We instantiate this as a reference monitor for regenerated device drivers, and characterize the mediability envelope, six conditions on the effect vocabulary: legibility, spec-input observability, correlatability, completeness, outcome enumerability, and explicit durability. They decide when such mediation is possible.

Observations increasingly reveal the coupled radiative, chemical, and dynamical processes that shape exoplanet atmospheres. Interpreting these atmospheres requires models that can capture this complexity. However, multidimensional models remain fundamentally limited by computational cost, and answering key questions requires simulating the governing physical mechanisms at speeds classical methods cannot achieve. As a result, models often rely on simplifying approximations, such as equilibrium chemistry, even when those assumptions miss important effects. There is a pressing need for fast and accurate chemical kinetics solvers to model planetary atmospheres. Here we present a machine learning local-box chemical kinetics solver for exoplanet atmospheres using a residual flow-map architecture. We demonstrate that this surrogate model is several orders of magnitude faster than a classical solver, achieving microsecond-scale inference while retaining percent-level accuracy. The surrogate model covers a parameter space that spans $T=300$-$3000$ K, $P=10^{-6}$-$10^{4}$ bar, $\Delta t=10^{-3}$-$10^{8}$ s, and compositions ranging from $10^{-2}$ to $10^{3}$ times solar in both C/O ratio and metallicity. Our model outperforms several commonly used machine learning architectures and performs robustly under the extreme stiffness characteristic of atmospheric chemistry. The machine learning framework presented here is a flexible and efficient approach to emulating state-to-state flow-map problems that commonly arise in numerical simulations.

The exponential growth of wireless devices is driving unprecedented spectrum demand, pushing spectrum management toward more fine-grained decisions across space, time, and device constraints. As a result, spectrum policymakers and engineers must process large volumes of data that come from diverse sources and take many different forms, such as text and tables. These data sources are often disaggregated and require significant time and effort to integrate, search, and interpret. Furthermore, most of this information is formatted for human understanding and is not readily accessible to automated systems. To address this challenge, we propose SpecMind, a novel Multi-Agent Retrieval-Augmented Generation (RAG) system for spectrum intelligence that performs reasoning over heterogeneous data sources. This system enables autonomous agents to coordinate specialized sub-agents that retrieve and synthesize knowledge across policy proceedings, legal regulations, and license databases. We develop SpecBench, a question and answer (Q&A) dataset based on real-world license records and policy proceedings, addressing the lack of evaluation resources for RAG systems in the spectrum domain. Experimental results demonstrate that SpecMind outperforms traditional, general-purpose RAG systems across spectrum-related tasks, achieving over 80% win rate against strong baselines. The agent-based design enables more accurate retrieval, better contextual reasoning, and improved task completion across diverse query types.

Reliable computational pathology depends on preprocessing methods that identify informative tissue regions while excluding artifacts and low-utility regions from whole-slide images (WSI). However, existing preprocessing pipelines often retain such regions or discard diagnostically relevant tissue, thereby limiting downstream model performance, reliability, and robustness across heterogeneous cohorts. Here, we systematically evaluate how these regions affect downstream AI model performance across multiple clinically relevant applications and introduce MUFASA, a generalizable, information utility-aware preprocessing framework for H&E-stained WSI that excludes artifacts and low-utility regions while preserving biologically meaningful tissue. MUFASA integrates slide-level artifact masking, stain-aware tile filtering, reconstruction-based utility stratification of tiles, and targeted recovery of tissue tiles that are over-filtered by earlier phases. Across tumor diagnosis, tumor subtyping, biomarker status prediction, and survival prognostication tasks in diverse cancer cohorts, MUFASA consistently improves downstream model performance relative to widely used preprocessing baselines. These gains are accompanied by reduced artifact-associated attribution in model heatmaps, indicating improved alignment between retained tissue and model attention. Our findings establish WSI preprocessing as a critical determinant of downstream model performance and validity, revealing that even accurate predictions can conceal important failure modes stemming from anatomically implausible reasoning driven by retained artifact-containing and low information-utility tiles.

In this paper, we establish exponential separations between Extended Frege systems for a range of intuitionistic substructural and linear logics. More precisely, for any logic $L$ below the intuitionistic logic obtained by extending $\mathbf{ILL}$ with structural rules, and any logic $M$ not contained in $L$, we construct a family of $\mathsf{FL_e}$-provable formulas that have short proofs in $M$-Frege but require proofs of exponential size in $L$-Extended Frege. The same result holds in the $!$-free settings, using $\mathbf{IMALL}$ and $\mathbf{FL_e}$ in place of $\mathbf{ILL}$. The key ingredient in proving these separations is a variant of the feasible disjunction property for $L$-Frege, which may be of independent interest.

ReRAM-based in-memory computing (IMC) architectures are widely regarded as a promising approach to alleviating the computational bottleneck of conventional architectures. Since ReRAM crossbars perform matrix-vector multiplication (MVM) in the analog domain, their computational energy consumption is highly dependent on weight and activation distributions. However, most existing ReRAM accelerators focus primarily on weight optimization while paying limited attention to the impact of activations on computational energy consumption, leaving the energy-saving potential of activation sparsity largely underexploited. In this paper, we propose a voltage-aware ReRAM-based accelerator (VARA), along with its accompanying design methodology. Specifically, we first introduce a voltage-aware training (VAT) algorithm that incorporates a preset threshold into the activation function to steer the activation distribution toward zero values, thereby enhancing activation sparsity. Building upon this, we further propose a co-zero activation reordering (CAR) scheme for crossbar-level computation skipping. CAR clusters activation dimensions based on their co-zero correlations and consistently reorders both the activation matrix and its corresponding weights. This process consolidates scattered zero activations into contiguous zero-valued regions to maximize the benefits of crossbar-level computation skipping. Extensive experimental results demonstrate that, with only marginal accuracy loss, VARA reduces the average total system energy consumption by 60.12% and improves the average system energy efficiency by 2.68$\times$ compared to the baseline, outperforming existing state-of-the-art accelerators for sparse-activation optimization.

The linear recurrent neural network (LRNN) is a simple model for studying how much memory a network builds up as it trains. For uncorrelated inputs, earlier work found that training itself settles the network between keeping the past and reacting only to the present. Real sequences are correlated, and we solve the learning dynamics exactly for correlated inputs. In the solution, keeping the past carries a cost. The whole effect of correlation lands on that cost. This cost reduces to the earlier one when inputs are uncorrelated and grows once they are positively correlated. Three findings follow. (1) Correlation reshapes the course of learning, not only its end. Memory builds, overshoots, and is partly removed, and the settled network keeps less of the past. (2) Memory switches off at a threshold set by one number, how much each input resembles the one just before it. Neither sequence length nor longer-range correlation moves this threshold. Memory is worth keeping only when the task needs the previous input more than the current input already supplies it through correlation with the past. (3) The best network changes too. Zero error demands a feedthrough, a path that passes the current input straight to the network's output and remembers nothing, and training builds it unprompted when given one spare hidden dimension. Our work turns one property of the input into a prediction of whether a network learns memory and explains why correlated data turns recurrent networks into change detectors.

Probing studies have established that syntactic information is decodable in early and middle transformer layers, but what happens to that information in later layers remains poorly understood. We apply a cross-layer generalisation analysis to three Greek-tuned large language models evaluated on tightly controlled minimal pairs: object-relative constructions in Modern Greek, where canonical (Subject-Verb-Object; SVO) and non-canonical (Verb-Subject-Object; VSO) orders differ only in within-clause word order, while preserving propositional meaning. When a probe trained on late layers (20-31) is tested on each early layer individually, it produces below-chance transfer (cluster-corrected, p<0.01), classifying 99.3% of non-canonical sentences as canonical. Probe coefficients reverse sign around layer 22, indicating a directional recoding toward the canonical form rather than simple information loss. These findings characterise a representational format change in late transformer layers that goes beyond the well-established decline in syntactic decodability, and they generate a directly testable prediction for human EEG and MEG decoding studies using the same stimuli. Code and stimuli are publicly available on OSF.

Modern face recognition (FR) owes much of its success to deep neural networks that learn to extract compact identity embeddings from face images. These models are typically trained for identity discrimination, producing embeddings that are highly effective for biometric matching but largely opaque to semantic interpretation. In contrast, foundation models, pretrained on broad visual or vision–language tasks, provide rich interfaces for describing, retrieving, generating, and organizing visual content. This contrast raises a natural question: what capabilities become available when face embeddings from domain-specific FR models are made interoperable with foundation models? Building on recent work on embedding compatibility across models, we use simple pre-computed linear transformations, estimated from paired embeddings alone, to connect existing FR models with off-the-shelf foundation models. Once aligned with a foundation model, a face embedding can be 'unmasked' in multiple ways, without training or modifying either model: it can be read in natural language, enabling free-form text queries over a gallery of FR embeddings; rendered into a face image that recovers a person's appearance, using an unmodified diffusion decoder; and converted to a name, enabling identification even in the absence of an enrolled face gallery. In effect, one linear transformation turns an identity embedding into a rich embedding for web-scale foundation models. This interoperability exposes face embeddings as semantically and visually rich biometric representations, with direct implications for interpretability, retrieval, reconstruction, and template security.

Mixture-of-Experts (MoE) models enable efficient scaling of large language model (LLM) inference but suffer from substantial data-movement overhead when deployed on neural processing unit (NPU)-based systems. Near-Data Processing (NDP) provides a promising way to mitigate this bottleneck via cooperative NPU-NDP execution. However, existing NPU-NDP MoE systems do not fully account for hardware heterogeneity, dynamic expert-level concurrency, and temporal expert reuse during batched inference. This paper presents DynaNDE, a dynamic near-data expert scheduling framework that exploits NPU-NDP collaboration to accelerate batched MoE inference. DynaNDE introduces an analytical performance model that captures hardware heterogeneity, data-movement costs, and communication-computation overlap in cooperative NPU-NDP execution. Guided by this model, DynaNDE determines per-layer expert scheduling across the NPU and NDP while accounting for expert-level concurrency. DynaNDE also incorporates a reuse-aware runtime that avoids redundant parameter movement when experts reside in NPU memory. Experimental results show that DynaNDE achieves substantial throughput improvements over the state-of-the-art NPU-NDP MoE serving framework, with average speedups of 2.6$\times$ and 2.2$\times$ for the prefill and decoding stages, respectively.

This paper presents a multi-branch fusion framework for detecting and characterising the propagation of health misinformation in online social networks (OSNs). Grounded in the Elaboration Likelihood Model (ELM) and the Theory of Planned Behaviour (TPB), the model fuses transformer-based semantics with rhetorical cues, stance representations, and psychologically motivated proxies in a unified multi-task architecture. In addition to binary classification, we introduce the Cognitive Propagation Score (CPS), an interpretable post-hoc auxiliary score computed from psychologically motivated, text-derived cues capturing argument complexity, emotional intensity, and content-derived virality potential, to support diffusion-risk reasoning when engagement ground truth is incomplete or unavailable. Experiments on three benchmark datasets, Constraint, COVID–19\_FNIR, and Monkeypox, show strong classification performance, achieving ROC–AUC up to 0.9999 on COVID–19\_FNIR, while propagation-oriented ranking achieves near-perfect agreement when engagement-derived supervision is available (Monkeypox, Spearman's $\rho = 0.9952$) and similarly high ranking alignment under proxy-based supervision on COVID–19\_FNIR ($\rho = 0.9954$). Compared with representative literature baselines, the fusion model improves detection on Constraint and COVID–19\_FNIR, while Monkeypox remains more challenging, reflecting domain- and signal-specific differences. Ablation analysis further indicates that psychological and rhetorical branches provide complementary gains beyond semantic embeddings. Overall, the framework bridges cognitive theory and neural modelling to improve transparency and to support scalable misinformation monitoring, with future work required to validate CPS against human-centred diffusion judgements.

This paper develops a general convolutional neural network (CNN) framework for detecting heterogeneous event-driven dynamics in univariate time series windows. We show that the induced CNN class exactly represents classifiers based on range, maximum drawup, maximum drawdown and slope change, and uniformly approximates realised volatility and autoregressive explosiveness on compact domains. We further establish error bounds for representative rules in finite samples and an oracle inequality for learning across them. Simulations show that the proposed model can match or outperform classifiers based on individual statistics as the training sample grows. In an application to six daily energy price series, a hierarchical CNN distinguishes event windows and event families. Applied without retraining to observations withheld after 20 February 2026, the fitted model identifies predominantly geopolitical dynamics in several oil and refined product series around the outbreak of the 2026 Iran war, while distinguishing a contemporaneous natural gas spike associated with weather.

High-energy physics analyses often process the same data as physicists refine algorithms and test new ideas. With data increasingly read from remote storage, each iteration is subject to network latency and depends on network bandwidth and shared-storage throughput, which can vary substantially under load. We present uCache (xrd-ucache), a transparent client-side cache implemented as an XRootD client plugin that requires neither server-side deployment nor changes to analysis code. It uses local storage on the analysis machine as a cache layer between the network and memory. The cache stores only the data actually read by an analysis. It can also rebuild cached data into a branch-aligned, recompressed form, eliminating most of the input/output and decompression costs of subsequent passes. We benchmark the cache using the Analysis Grand Challenge top quark pair analysis on public CMS Open Data compressed with zlib and LZMA. Filling the cache adds essentially no overhead compared with a direct read. Subsequent passes are 1.6-8.8 times faster from the byte cache and 2.1-15.7 times faster from the recompressed cache. For a typical analysis, a 1 TB cache suffices for datasets of 10-20 TB. The largest improvements occur when the remote data source is heavily loaded or geographically distant.

Let $X_1,X_2,\ldots$ be i.i.d. finitely supported random variables in a torsion-free abelian group, and write $S_k=X_1+\cdots+X_k$, and $H(S_k)$ is the Shannon entropy $S_k$, for all $k \ge 1$. We prove that, for every fixed $n\geq1$, \[ H(S_{n+1})-H(S_n) \geq \frac12\log\frac{n+1}{n} -o_{H(X_1)\to\infty}(1), \] uniformly over the ambient group and the input law. This proves a conjecture of Tao [29] in 2010.

Histopathological whole slide images (WSIs) are central to cancer diagnosis, but their gigapixel scale, tissue heterogeneity, weak slide-level supervision, sparse diagnostic regions, and multi-scale evidence make robust automated analysis challenging. Multiple instance learning (MIL) is widely used to aggregate tile-level features into slide-level predictions, yet existing augmentation strategies often perturb tissue regions without preserving diagnostic relevance, slide context, or cross-scale structure. We propose SlideMix, a model-agnostic multimodal augmentation framework for MIL-based WSI analysis. SlideMix uses a retrieval-augmented vision-language model (VLM)-based Visual-Language Adaptive Region selector to identify diagnostically relevant regions and reduce weak-label noise. It then performs In-place Tile Shuffling within meaningful tissue regions to mix feature embeddings while preserving slide-level context. A VLM-based soft-labeling module supervises mixed samples, while a multi-factor, loss-driven online Curriculum-Learning Feedback scheme adaptively controls shuffle granularity, feature similarity, and shuffle ratio to promote cross-scale representation learning. Across 11 WSI datasets comprising 20,523 slides, 8 diagnostic tasks, and 10 WSI backbones, SlideMix improves accuracy and generalization in most settings and compares favorably with established augmentation baselines, providing a simple plug-and-play approach for more robust and scalable digital pathology models. Source code: https://github.com/Xia-Research-Lab/SlideMix

Pseudosimilar vertices are vertices in distinct automorphism orbits whose deletions produce isomorphic graphs. Classical work has studied the existence, group-theoretic origin, and construction of large sets of such vertices. We ask a different recursive question: how long can one repeatedly delete a vertex that is pseudosimilar at the moment of deletion? We define the pseudosimilarity depth of a graph and construct connected graphs in which this process continues through all but a sublinear number of vertices. A two-clock construction gives a square-root deficit uniformly in the order, while a Chinese-remainder construction with many cyclic clocks yields an infinite family of asymmetric graphs with only a polylogarithmic number of vertices left outside the active chain. The mechanism realizes pseudosimilarity by breaking a long hidden automorphism orbit and enlarging the break one vertex at a time. Thus pseudosimilarity can persist through an asymptotically full sequence of vertex deletions, even though every graph encountered in the main construction is asymmetric.

Let $P$ be a polygonal curve with $n$ vertices in the plane. We construct a data structure of size $O(n \log n)$ suited for simplification queries of the following kind. Given a query line $\ell$ and an integer $k\ge1$, find a curve $Q$ on $\ell$ with at most $k$ vertices that minimizes the discrete Fréchet distance to $P$, among all such curves. Using our data structure, a query can be handled in $O(k^2 \log^3 n + k\log^4 n)$ time. More generally, a geometric tree $T$ on $n$ vertices in the plane can be preprocessed into a near-linear-size structure so that, given a pair $u$, $v$ of its vertices, a line $\ell$, and an integer $k\ge1$, one can find a curve $Q$ on $\ell$ with at most $k$ vertices that minimizes the discrete Fréchet distance to the path from $u$ to $v$ in $T$, in time $O(k^2 \mathop{polylog} n)$. For the general dimension-reduction problem, where $P$ is a curve in $\mathbb{R}^d$ ($d \ge 3$), $0 < \varepsilon_0 < 1$ is a real parameter, and a query specifies a $g$-flat $h$ ($1 \le g \le d-1$) and an integer $k \ge 1$, we construct a data structure of size $O(n\log n + f(\varepsilon_0) n)$, where $f(\varepsilon_0)=(1+1/\varepsilon_0)^{(d-1)/2}$, that allows us to find a curve $Q$ on $h$ with at most $k$ vertices, whose discrete Fréchet distance to $P$ is at most $1+\varepsilon_0$ times the distance of $Q^*$ to $P$, where $Q^*$ is such a curve that minimizes the distance to $P$. The query handling time is $O(f(\varepsilon_0) k^2 \log^2 n)$.

Wearable EEG systems may expose sensitive information beyond their intended health function, creating substantial risks to neuroprivacy. In this work, we show that commonly used EEG features can reveal participant identity and demographic attributes in addition to supporting the intended cognitive task. Wearable EEG is increasingly being explored for cognitive monitoring, neurological assessment, and longitudinal digital-health applications, yet many systems assume that transmitting compact spectral or spatial features instead of raw EEG provides sufficient privacy protection. Using EEGMAT as a motivating case study, we find that compact EEG features achieve a balanced accuracy of 0.788 for cognitive-state classification while enabling gender, age, and subject-identity inference with balanced accuracies of 0.858, 0.789, and 0.692, respectively. We further show that privacy-aware representation learning preserves task performance at 0.781 while reducing these inference accuracies to 0.563, 0.467, and 0.206. These findings motivate purpose-limited representations and explicit privacy auditing in wearable neurohealth systems.

Open synthetic test systems are essential for reproducible power system research, yet the available cases represent almost exclusively North American and European grids. No open transmission test system exists for any Middle Eastern country, whose grids raise different questions such as single corridor fuel-supply resilience and high-renewable operation within small synchronous systems, and established benchmarks cannot demonstrate. This paper presents a synthetic test system for the Jordanian transmission grid, assembled entirely from public sources, where the real topology is reconstructed from a published diagram and restored to its 2018 energized state, with plant-level generation and renewable fleets, per-site hourly profiles, and loads calibrated to the values published by Jordan's grid operators. The case is validated through structural statistics against real-grid criteria, power flow and $N\!-\!1$ screening, an energy-weighted loss decomposition, in addition to a full-year production-cost run compared against the published per-plant energy, and cross-solver verification. The model reproduces the annual energy mix at technology level (largest producers within 4%, system total within 0.1%), and the released dataset includes the bus-identity key, all scripts, and a post-2019 scenario variant. This test case is designed to benchmark system-level resource adequacy, time-series dispatch and unit commitment, renewable integration, fuel-supply resilience scenarios, interconnection studies and steady-state studies.

Large language models extracting knowledge graphs from text capture only explicitly stated facts, often leaving semantically related entities disconnected across documents. We present an additive, engine-neutral second pass that discovers these latent ties without altering extracted facts. Each document is chunked and embedded once; top-k nearest- neighbor queries across existing chunks yield candidate node pairs via entity membership maps. Candidate pairs are scored using Shepard inverse-distance weighting with a rescaled chord distance metric, avoiding the threshold-collapsing flaw of affine cosine scoring behind a k-NN gate. Un-gated per-pair accumulators form a commutative monoid, ensuring the pipeline is strictly order-independent and scales incrementally without recomputing prior documents. Implemented across FalkorDB, Kinetica, ArangoDB, and Neo4j, our method shows that 768- and 240-dimensional embeddings retain 92% and 72% edge fidelity against a 3072-D baseline while achieving a 25x faster top-k formulation.

Large Language Model (LLM) agents increasingly automate multi-step engineering workflows through tool use, interpretation of intermediate results, and iterative planning. Diagnosing and resolving non-convergent power flow cases is a promising yet largely unexplored application, as it requires engineering judgment, experimentation, and decision-making within constrained action spaces. We introduce a benchmark that evaluates these capabilities across multiple LLMs and three architectures: chatbot, single agent, and multi-agent systems. The evaluation covers two power grids and 46 cases per grid, each requiring one or more corrective actions to restore convergence. The benchmark defines the simulation environment, observation and action spaces, and evaluation metrics, providing a reproducible foundation for developing agentic AI systems for power system planning and operation. The code is available at https://github.com/Mansutti081/RestoreBench

LTLf, a finite word variant of LTL, can be reduced to LTL by introducing a new atomic proposition indicating the prefix of the infinite words that correspond to the finite words that the original LTLf formula was considering. Such a reduction was originally proposed by De Giacomo and Vardi (IJCAI'13). However, while any LTL formula reduced from LTLf describes an obligation property in the hierarchy of Manna and Pnueli (PODC'90), the aforementioned reduction does not provide an LTL formula that belongs to the syntactic obligation fragment of LTL. This note shows how the reduction was fixed in Spot in order to ensure that the resulting LTL formula is always a syntactic obligation. Doing so allows algorithms specialized to syntactic obligation to be used on LTLf formulas. For instance, in previous work (CAV'26) we described a specialized translation from syntactic obligations to minimal, weak, deterministic Büchi automata that would not be usable with the original reduction.

We present FoldingAgent, an agentic framework for inferring explicit parametric folding programs directly from origami demonstration videos. Our framework leverages the reasoning power of a pre-trained Vision-Language Model (VLM) equipped with a suite of specialized tools that enable the agent to simulate geometric transitions, verify physical plausibility, retrieve and compare visual content, and evaluate its own predictions. To translate visual content into folding programs, we define a parametric space that consists of the paper's geometry and a set of parametric folding actions. Unlike models that predict static crease patterns, our agent operates sequentially and possesses the ability to re-plan its actions, effectively mitigating the compounding errors inherent in multi-step folding. Our approach takes a step toward closing the gap between human origami knowledge, which is primarily shared through unstructured visual demonstrations, and computational methods, which typically rely on structured, parametric representations such as a crease pattern or an executable parametric plan. We evaluate our approach on PurelandFold, a newly curated benchmark of diverse Pureland origami videos with ground-truth geometry and action labels. Our results demonstrate that by combining VLM reasoning with a set of specialized tools and physical simulation, we can successfully transform unstructured visual demonstrations into executable, physically plausible folding procedures.

Test-time adaptation (TTA) typically assumes that model parameters can be updated at inference time. This assumption is restrictive for inference-only accelerators, frozen or third-party models, and memory-constrained deployments, and standard BatchNorm-based TTA configurations may also become inactive on architectures without BatchNorm. We study adaptation when the learned model must remain frozen. We introduce CASTER, a gradient-free method that stores source class statistics in a discriminative subspace, estimates a class-shared affine transformation from target-batch moments, and analytically transports the source class distributions before classification. CASTER requires no backward pass, optimizer state, or stored source feature bank. Across four backbones and seven datasets, it outperforms k-NN on identical frozen features in 27 of 28 backbone-dataset settings while retaining a median of 18x less state. Affine transport is not always reliable. On ImageNet-C, where batches contain only 64 samples for 1000 classes, unconditional transport loses 21.2 top-1 points. We therefore introduce an empirical residual-to-margin transportability certificate. Across 307 evaluation cells, every transport losing more than 10 points has certificate value above 3.9, although benign and destructive regimes are not perfectly separated. Gating converts an average $-3.35$-point effect of unconditional transport into a +1.69-point gain, and performance remains within 0.3 points of the best threshold over a broad threshold range. Finally, we show that this certificate is mechanism-specific: when applied to Tent, it accepts only $4.3\%$ of updates and preserves 0.6% of Tent's available gain. These results position CASTER as a lightweight adaptation mechanism for frozen-model deployment, together with an explicit account of when its safety signal is informative and when it is not.

With quantum sensors, simulators and networks emerging, a future of quantum technology may produce quantum states as data—that is, coherently rather than as classical measurement records—thus motivating the study of suitable quantum generalisations of modern machine learning, including the automated, unsupervised extraction of useful representations. Two ingredients are central to the latter: inference, mapping observations to latent representations, and generation, mapping latent states back to synthetic data. Both are related to each other and to joint distributions for training models by the chain-rule of classical probability theory. The fact that quantum states however lack such universal, standard factorisation property thus poses a challenge. Here we develop a conceptual and mathematical framework for unsupervised representation learning from quantum data. Models are joint quantum states over visible and latent systems; state-over-time maps provide a notion of factorisation into a marginal state and inference (generation) channel; models with inference (generation) are ambiguous states—states for which such factorisation obtains—subject to a further consistency condition on extended inference maps as data extension. These stipulations are restrictive: we show that non-trivial models must feature non-linear such maps to the extended space. For three representative state-over-time maps, we completely characterise the ambiguous states, uncovering a hierarchy tied to the positive-partial-transpose (PPT) criterion from entanglement theory. Notably, the Leifer-Spekkens construction supports inference and generation exactly for model classes of PPT states, thus allowing genuinely quantum visible-latent correlations. We also formulate quantum counterparts of exact and approximate inference training, explore weaker notions of data extension and sketch a future research programme.

Generating co-speech gestures that are temporally coherent, semantically aligned with speech, and grounded with surrounding objects remains challenging. Prior speech-driven gesture models emphasize audio-gesture alignment but do not explicitly account for posture constraints or surrounding objects, failing to capture the inherent correlation between body gestures and the physical space. We present Puppeteer, a posture-aware, object-grounded co-speech gesture diffusion model operating in a causal latent space. We decompose long gestures into structured primitives and learn a causal variational autoencoder that encodes them into temporally ordered latent tokens, each depending only on the past. We then perform conditional diffusion directly in the causal latent space, conditioning on speech signals, motion history, an initial posture reference, and object geometry to synthesize physically consistent gestures. This temporally ordered latent formulation enables explicit temporal control and supports tasks such as gesture in-betweening and gesture completion. To better assess co-speech gesture synthesis beyond existing measures, we introduce new evaluation metrics tailored to this task. We also created SceneGes, the first curated synthetic 3D dataset of embodied co-speech gestures and corresponding 3D objects, enabling object-grounded gesture generation. Experiments show that Puppeteer generates more diverse and temporally synchronized gestures than prior methods, while enabling object-grounded gesture synthesis.

Command-line coding agents (e.g., Claude Code, Gemini CLI) can already read and write files and sustain long sessions, yet end-to-end research still fragments across chat tools, IDEs, terminals, and writing environments, and the decisions that make it auditable are rarely preserved. We present Dr. Claw, an open-source workspace that wraps existing coding-agent executors in a controllable and auditable human-in-the-loop workflow rather than introducing another autonomous agent. Persistent state objects, a reusable skill library, and multi-executor coordination link human decisions to AI execution, turning planning, execution, and writing into one traceable, recoverable loop. We demonstrate Dr. Claw through an interactive three-view scenario and a failure-recovery walkthrough, and evaluate it against a bare command-line agent sharing the same backend executor, so the comparison contrasts the whole orchestration layer (task graph, state objects, and skill library) with the agent it wraps. Holding the executor fixed, Dr. Claw scores higher on research completeness while persisting an auditable, recoverable process trail. Demo access: repository https://github.com/OpenLAIR/dr-claw, released under AGPL-3.0 with GPL-3.0 upstream components.

Embedding models are trained and evaluated as if retrieval were exact; in production they serve behind approximate indexes – HNSW, IVF, product quantization, or the fixed-dimensional encodings (FDEs) of late-interaction models – whose behavior the encoder's benchmarks never see: one modern encoder recovers just 14% of its exact top-10 through its raw FDE index. Such failures surface only after an index is built, and the standard patches – corpus-fitted transforms such as whitening – must be fitted, stored, and refit as the corpus changes, and can silently rewrite what the encoder returns. This paper shows that index behavior is predictable before anything is built, from label-free statistics of the raw embeddings, through a ladder of instruments matched to what each index family consumes: (1) closed-form moment statistics for the fixed-grid quantizers (PQ, FDE); (2) simulation on a synthetic twin corpus – cluster statistics made generative, on which any index, composed production systems included, can be built and tested – for partition indexes; (3) size-extrapolated, lightly calibrated twins for graph indexes at million-document scale. Predictions land within 0.03 of measured recall on an unseen million-document corpus. The same geometry is trainable: targeting the one statistic no post-hoc transform can move – the score margin – lifts recall for every index family at once, at a small measured task cost. The result: index choice, correction pricing, and production recall forecast from one cheap measurement pass, on new corpora and new indexes alike; serving without per-corpus transform machinery, suited to continuously changing corpora; and a recall-compute frontier pushed by adapting encoders to geometry rather than coupling them to any single index.

Since the advent of Large Language Models (LLMs), practitioners have increasingly leveraged them to support their software engineering tasks, including automated code repair, showing promising results. Yet, concerns regarding reproducibility and generalizability remain largely unexplored. To further evaluate these concerns and associated impacts, we partially reproduce and conduct an exploratory Java extension of the FeedbackEval benchmark [1], which evaluates how LLMs leverage different feedback types for Python code repair. First, we partially replicate the original study on 394 repair tasks using GPT-4o and Claude 3.5 Sonnet, reproducing and observing the main qualitative trends reported in the original work. Second, we conduct an exploratory Java extension by constructing 100 erroneous repair instances from 50 Java tasks and evaluating feedback effectiveness. Our results show that previous conclusions from Python may be sensitive to benchmark construction, feedback representation, and tooling ecosystem, motivating more controlled multilingual benchmarks. Specifically, while test feedback remains the strongest feedback type in our Python replication, the same ranking is not observed in our Java extension, as simple and JUnit-based test feedback do not differ significantly. We hypothesize that differences in feedback informativeness and tooling ecosystems, such as the verbosity of test frameworks, may partly explain such a difference. Finally, lighter prompts reduce cost without significant differences in repair effectiveness. Overall, our findings confirm key trends under a partially controlled replication and highlight the need for more rigorous multilingual evaluation and careful feedback design in LLMbased repair systems.