Peer Revenue Play credits. Real papers.

Papers

Fresh from arXiv, typeset properly. Highlighted words trade on the market.

All of arXiv

Thursday 10 September

Computation and Language

Break Step: Recursive Training Resonates with Replayed Sampling Noise

How fast does a language model degrade when trained on its own outputs? Theory traces it to gradually accumulating errors, while experiments report repeated phrases within ten generations. Under a fixed sampling seed in vLLM, the fast loss of lexical diversity comes from the sampler. When vLLM serves a batch from one seeded sampling configuration, every request receives the same random draws, and a fixed seed replays them every generation. Fine-tuning raises the tokens that won, and the replayed draws let them win by more. Sharing across requests and replay across generations matter only together. Remove either one, by changing the shared seed every generation or by giving each request its own seed that repeats every generation, and the unique-4-gram fraction of two StableLM checkpoints stays near its starting value of about 0.98 through generation 3. Keep both, and the replayed shared seed takes seven checkpoints from five families to between 0.045 and 0.38 by then. Three generations of replay write the favoured phrases into the weights: decoded with one seed per request, the generation-3 weights of the replayed StableLM-2-1.6B chain recover most of their diversity, yet the phrase that filled every sample under the shared seed still opens 46% of them. Without replay, five checkpoints drift slowly, consistent with the gradual accumulation that theory describes, and three turn incoherent though their diversity scores stay high. One peer-reviewed model-collapse pipeline that fine-tunes Gemma-2-27B samples identical prompts under one seeded configuration, and three quarters of the rows it released for one iteration repeat nearly as often as one such batch copies them. A seed per request restores the fresh sample that stability analyses assume.
Read the full abstract

Databases

OmniTable: A Unified Wide-Table System for Petabyte-Scale LLM Data Curation and Exploration

Data curation is a critical bottleneck in industrial-grade LLM development, where petabyte-scale unstructured corpora are scattered across hundreds of physical tables, feature engineering relies on manual, table-centric pipeline orchestration, and data lineage is largely absent. We present OmniTable as an architecture blueprint for a unified wide-table layer built on Logical Unification, Physical Separation, targeting petabyte-scale LLM data curation and exploration. OmniTable makes four contributions: (1) a unified wide-table abstraction that consolidates multi-source heterogeneous data and thousands of derived features under a single logical schema via logical-physical mapping; (2) declarative feature lifecycle management that automates dependency resolution, execution planning, operator fusion, and lineage tracking, replacing manual pipeline orchestration with a "declare-and-execute" paradigm; (3) an adaptive execution engine with autonomous governance that achieves stable PB-scale feature backfill through heterogeneous compute routing (CPU/GPU), adaptive tuning, UDF-level fault tolerance, and automated storage layout optimization; and (4) hybrid-accelerated data exploration combining a global ID index, transparent OLAP offloading, and background materialized views to deliver second-level point lookups and filtered exports exceeding 20 TB/hour. In production, OmniTable manages over 35 PB of training data across web, code, PDF, and SFT domains, reducing the human-in-the-loop curation cycle from approximately 14 days to approximately 2.5 days (5.6x over the pre-OmniTable production workflow), with consistent feature versioning, auditable lineage, and minimal manual intervention.
Read the full abstract

All authors Yuzhuo Fu, Xiangchun Wang, Chao Huang, Liyi Wang, Binwei Zeng, Yuhan Wang, Taotao Nie, Dongke Hu, Wang Hong, Jiayi Wang, Wenwen Cui, Zhuyan Zhou, Yushun Guo, Yuhan Xing, Jiaxin Lian, Peng Lin, Qing Cui, Wenhui Shi, Jun Zhou

Artificial Intelligence

Autonomous Chemical Mechanistic Discovery through Agentic Reasoning and Validation

Unraveling reaction mechanisms is central to modern chemistry, yet automating these investigations remains challenging because computational workflows still rely heavily on expert intervention. Here we introduce ARCHE, an autonomous agentic system that integrates a general-purpose reasoning model, a domain-specialized computational chemistry model, and a structured tool registry to transform mechanistic inquiry into a scalable, self-validating process. ARCHE interprets scientific questions, generates and prioritizes mechanistic hypotheses, orchestrates computational workflows, and iteratively refines conclusions based on computed evidence within a closed loop. We validate its capabilities across three increasingly demanding scenarios: reconstructing stereocontrolling transition states and validating the corresponding reaction mechanism in a previously reported asymmetric catalytic reaction; proposing and validating a plausible radical pathway through iterative hypothesis refinement for a recently discovered but unpublished $\alpha$-iodoboronate C-I cleavage reaction; and identifying a chemically interpretable descriptor that governs selectivity in nickel-catalysed migratory cross-coupling reactions. By coupling agentic reasoning with rigorous computational validation, ARCHE advances autonomous mechanistic discovery and establishes a foundation for broader machine-assisted chemical research. The code for ARCHE is publicly available at https://github.com/JetAstra/Arche-Harness.
Read the full abstract

All authors Dong Li, Sixuan Mi, Zihao Ye, Huan Xiong, Tao XU, Tong Zhu, Aijia Zhang, Junqi Gao, Kaiyan Zhang, Shijie Wang, Bowen Zhou, Yuqiang Li, Biqing Qi

Artificial Intelligence

A Dominant Supplier Slows Recursive Drift More Than It Steers It

More and more of the text future language models learn from is written by a few of today's models. If one supplier writes most of a shared corpus, does it pull the models trained on it toward its own writing, or change how fast they drift? We retrain eight open models from their base weights on a shared pool of each other's text for five generations, varying the part written by one model, Phi-2, from an equal share to 90%. The models drift together toward a style with fewer function words, and none starts repeating itself. No share of Phi-2 brings the other models closer to its text than the equal share does. We split each ecosystem's separation from the equal-share one into a delay along its route and a departure from that route, both counted beyond the difference between two equal-share runs. With Phi-2 at 90%, delay outweighs departure 72 to 28 and 64 to 36 in two runs, and the ecosystem falls 2.7 and 2.5 generations behind. With Phi-2 at half the pool the two parts are about equal. When SmolLM2 or Qwen3-1.7B writes half instead, the ecosystem slows less or not at all. The departure leans toward Phi-2 more as its share grows, but more than toward every other model only at 90%. Human text filling a quarter or half of the pool slows the models along the same route.
Read the full abstract

Distributed Computing

The Computing Channel: How Modulation Programs the Airwaves

Distributed computing and distributed artificial intelligence require frequent exchanges of intermediate results, although many applications need only an aggregate rather than messages from individual devices. Conventional systems recover each message before computing the aggregate, whereas over-the-air computation (OAC) exploits simultaneous transmission to obtain it directly. However, dominant OAC implementations rely on analog signaling, creating a mismatch with finite-precision data and digital communication procedures. This article presents digital function-oriented communication, in which finite-alphabet symbol representations and receiver decisions are jointly designed so that multiple-access superposition encodes the desired function without recovering individual inputs. We introduce its computational-constellation principle, main design approaches, extensions, and implementation challenges. Federated edge learning illustrates how the framework can reduce user-dependent data-bearing resources while operating directly on quantized model updates.
Read the full abstract

Artificial Intelligence

Human Agreement and Return Association Are Not Interchangeable Criteria

Financial NLP has a standard workflow: validate a sentiment tool against human labels, then trust it to extract market signal. This assumes the two evaluations measure the same thing. We test that assumption in a setting where both can be measured at once: a corpus of securities class actions (2002-2025) linking 70,500 X messages to abnormal stock returns, with a single-annotator human labelled gold sample. Running five instruments (VADER, Loughran-McDonald, FinBERT, Twitter-RoBERTa, and an LLM annotator) through one identical pipeline, we find that the relationship between construct and predictive validity depends on the sampling convention and score representation. Under conventional method-specific sampling, human agreement aligns more closely with graded same-day associations than with one-day leads. On a fixed-n panel, however, agreement has similar graded rank correlations at both horizons, while the coarse ordering remains weak. Benchmark agreement therefore establishes semantic validity but does not by itself determine predictive rankings. In a conversation that is 17.6% spam, message volume predicts neither market damage nor settlement size.
Read the full abstract

Solar and Stellar Astrophysics

Statistical evidence for linear combination frequencies in {\gamma} Doradus stars being nonlinear resonant modes

The quest for the origin of linear combination frequencies in pulsating stars is critical for mode identification and subsequent seismic modeling. Hybrid {\gamma} Doradus (Dor) stars provide a crucial testbed for such investigations due to their rich frequency content in both gravity (g) and pressure (p) modes. Here, we performed a comprehensive survey of 608 Kepler {\gamma} Dor stars, detecting 81,265 frequencies, and identifying 1,895 harmonics and 45,968 combinations involving two parent frequencies. Our results suggest that the proportion of linear combinations increases with the frequency number in each {\gamma} Dor star, leveling off at approximately 75%. The majority of the parent-to-child amplitude ratios are distributed between 100 and 10,000, peaking between 2,000 and 4,000. Moreover, the amplitude ratios, involving two g-mode parents, are found to escalate with increasing combination frequency. These values are at least an order of magnitude higher than those predicted by nonlinearity distortion of the light curve, whereas they align well with those of nonlinear resonant coupling. Our findings suggest that a significant portion of combination frequencies are likely intrinsic modes under resonant conditions rather than products of surface geometric distortions, offering new insights that warrant systematic investigations across other types of pulsating star.
Read the full abstract

All authors Yanqi Mo, Weikai Zong, Xuan Wang, Jian-Ning Fu, Xiao-Yu Ma, Mingfeng Qin, Stéphane Charpinet

Signal Processing

Radio Map Construction with Post-Hoc Location Calibration under Quasi-Static Positioning Errors: Joint Estimation, Performance Bounds, and GNSS-Based Evaluation

Radio maps enable environment-aware wireless and Internet-of-Things applications and can be constructed from location-tagged received signal strength (RSS) measurements collected by mobile devices. In urban environments, temporally correlated GNSS errors can shift an entire sensing trajectory, causing systematic spatial misregistration that is not mitigated by collecting more measurements. This paper presents a radio-map construction framework that uses the radio measurements themselves to calibrate erroneous location tags after data collection. The dominant positioning error is modeled as a sensor-specific quasi-static offset, which is jointly estimated with radio-propagation parameters in a Gaussian process regression (GPR) framework by exploiting complementary spatial information from distance-dependent path loss and spatially correlated shadowing. We establish lower and upper bounds on the conditional Bayes risk and show that, under a translation-invariant trajectory model, trajectory information alone cannot identify the quasi-static offset, thereby motivating the use of RSS-derived spatial information for calibration. Numerical evaluations across propagation conditions show that the proposed method reduces the mean squared error (MSE) gap from ideal GPR to approximately $3.26\mathrm{dB}^2$, compared with about $10\mathrm{dB}^2$ for position-error-agnostic and noisy-input GPR baselines. Evaluation using positioning-error models derived from smartphone GNSS measurements shows that the proposed method outperforms a KF–RTS trajectory-smoothing baseline despite unmodeled time-varying positioning errors, remaining within approximately $5\mathrm{dB}^2$ of ideal GPR at the median MSE. These results demonstrate that RSS measurements can serve not only as observations for radio-map reconstruction but also as spatial cues for post-hoc calibration of imperfectly geotagged sensing data.
Read the full abstract

Computation and Language

Can LLMs Normalize Databases? A Benchmark and Multi-Agent Framework for Schema Normalization

Large Language Models (LLMs) are increasingly used to generate structured outputs, but their reliability remains unclear when those outputs must satisfy database-level constraints. We study this issue through database normalization, involving reasoning about functional dependencies, lossless join decompositions, and inter-table constraints. We introduce a Database Normalization Benchmark (DNBENCH), comprising 3,275 samples for evaluating LLM-driven database normalization from 1NF to BCNF. DNBENCH uses a three-axis protocol to measure semantic equivalence, structural accuracy, and logical validity. Across Single, Complex, and Real World levels, DNBENCH uncovers recurring failures in dependency inference, schema decomposition, and inter-table constraint reconstruction. We further propose Multi-Agent Reasoning for Schemas (MARS), which separates evidence extraction, violation diagnosis, and decomposition planning from schema generation and verification. MARS improves the DNB-SCORE by 82.0% over the single-prompt baseline. All artifacts will be released upon acceptance.
Read the full abstract

All authors Dong-Jae Koh, Huisu Kim, SeongHwan Yoon, Lasse M. Jantsch, Chun-Hee Lee, Seonghyeon Lee, Young-Kyoon Suh

General Relativity and Quantum Cosmology

Electrovacuum Black Hole Uniqueness

We prove the black hole uniqueness conjecture in the axially symmetric, stationary, electrovacuum setting, subject to the refined asymptotic analysis of the associated singular harmonic maps, which includes an analyticity hypothesis at the axes. More precisely, it is shown that any asymptotically flat solution of the Einstein–Maxwell equations in this class, with more than one black hole horizon component is either: Majumdar–Papapetrou, up to a duality rotation, in which case all logarithmic angle defects vanish, or every finite axis rod logarithmic angle defect is strictly negative and hence every interaction force is strictly attractive. The proof extends the singular harmonic map method used for vacuum Kerr uniqueness in [18].
Read the full abstract

All authors Qing Han, Marcus Khuri, Gilbert Weinstein, Jingang Xiong

Systems and Control

A Counterexample to Two Representative Unit Aggregation Formulations for Unit Commitment

Unit aggregation removes symmetry among identical generators in unit commitment, but an aggregate formulation is feasible-region exact only if every aggregate trajectory it admits has a feasible unit-level realization. This letter shows that two commonly used aggregation formulations for slow-ramping units, i.e., p-clustered unit commitment (PCUC) and tight unit aggregation (TUA), cannot satisfy this requirement. We construct a counterexample that satisfies all aggregate constraints of both formulations yet admits no feasible disaggregation. This counterexample reveals a limitation common to both formulations: they do not guarantee intertemporal consistency of unit-level output allocations. Experiments on the replication cases reported in published literature further verify that such infeasibility of disaggregation can even occur in optimal solutions of aggregate models. This letter demonstrates that PCUC and TUA can still overestimate ramping flexibility. Developing exact aggregation models that can be solved efficiently for identical slow-ramping units remains a challenge.
Read the full abstract

Mesoscale and Nanoscale Physics

Fingerprints of Excitonic Collective Modes in the Two-Dimensional Electron Gas

The two-dimensional homogeneous electron gas (2D HEG) is a prototype system of fundamental interest that can be realized experimentally over a wide range of densities. Here, we investigate its collective charge excitations at low densities, using time-dependent density functional theory. We demonstrate that, beyond traditional plasmons, new collective excitonic modes emerge for a Wigner-Seitz radius larger than $r_{\mathrm s} \approx 1$. These excitonic modes leave characteristic fingerprints in experimentally accessible quantities, namely, asymmetric peak structures in the loss function and very strong Friedel-like oscillations in the static linear density response that increase when the system approaches a regime of instability. Indeed, at low enough densities the collective modes cross the zero energy axis, indicating an instability of the paramagnetic 2D HEG towards the formation of a charge-density-wave phase with excitonic origin. These findings provide valuable insights for the experimental detection of excitonic collective modes in tunable 2D electron systems and contribute to the fundamental understanding of many-body effects in low-density electron gases.
Read the full abstract

All authors Jakob Wolff, Silvana Botti, Lucia Reining, Matteo Gatti

Cryptography and Security

The Machines Are Calling: Measuring Automated and Synthetic Voices in Unwanted Inbound Calls

In February 2024 the U.S. Federal Communications Commission (FCC) placed AI-generated voices under the Telephone Consumer Protection Act (TCPA). Yet no peer-reviewed measurement says how much unwanted call traffic is placed by a machine, or how much of that machine speech is synthesized rather than played from a recording. We report both with a disclosed pipeline. An interactive voice honeypot (language-model personas on real U.S. numbers, the caller recorded on its own track) recorded 10,987 calls over 66 days. Three instruments read each opening: an audio fingerprint that finds the same recording played on other calls, a commercial synthetic-speech detector on the caller's first ten seconds, and blinded listeners who check what it flags. Of the 7,233 greeted calls we analyze, 13.8% open with a recording we also heard on another call, and 13.1% with fresh audio the detector labels synthetic. A further 9.9% open with a caller who never spoke after our greeting, 54.2% with fresh audio the detector labels human, and 9.0% could not be scored. Machine-voiced openings are therefore at least 26.9%, a further tenth of calls are silent connections we read as machine-placed, and replays of a recording make up 45% of the detector's own rate (29.3% of 6,192 scored openings). The same waveform played on two calls lands on opposite sides of the detector's threshold 13.6% of the time, and eleven listeners confirm 54.4% of what it flags. Synthetic openings concentrate in lead-generation spam (33.8%), not fraud (21.1%); 0.44% disclose automation. Prevalence tracks how long a bait number has circulated (59% against 19% in the same weeks): seeding history, not calendar time, explains the trend. Campaigns outlast their numbers: one recorded compliance notice opens calls in six campaigns, and one synthetic voice serves nine.
Read the full abstract

All authors Xingyu Shen, Tommy Duong, Muduo Xu, Xiaodong An, Jiaqi Gan, Haoyuan Tang, Jamey Z. Liang, Siyu Zhang, Yan Zhang, Ethan Traister, Simiao Ren

Statistical Methodology

Beyond Tweedie's Formula: Conditional Score Modeling for Empirical Bayes Inference

We propose conditional f-modeling (Cf-modeling), a framework for empirical Bayes inference with covariates. A central identity shows that the conditional marginal score function determines not only the posterior mean through Tweedie's formula, but also the posterior moment-generating function, providing a basis for recovering posterior quantities without explicit prior modeling. Motivated by this observation, we treat the conditional marginal score as the primary object of inference and estimate it directly using an energy-based representation and Hyvärinen score matching, thereby avoiding potentially intractable covariate-dependent normalizing constants. The resulting framework flexibly accommodates covariate effects and heteroscedasticity and provides a practical approach to posterior moment estimation and uncertainty quantification. We demonstrate the effectiveness of the proposed method through simulations and an RNA-seq application.
Read the full abstract

Machine Learning

Bidirectional Multimodal Fusion of Sky Images and Time-Series for Solar Forecasting with Large Language Models

Short-term photovoltaic (PV) power and global horizontal irradiance (GHI) forecasts are essential for effective dispatch, reserve scheduling, and grid operations. At these forecasting horizons, errors are predominantly driven by cloud induced ramps: relying solely on historical numerical data may struggle to anticipate an incoming cloud, making ground-based sky images a crucial complementary physical signal. Furthermore, forecast performance is highly sensitive to location and local observing conditions, creating a strong need for site-specific data that are often scarce. Recently, large language models (LLMs) have demonstrated competitive performance and high data efficiency in time-series forecasting. Despite their success, existing LLM-based forecasting methods remain predominantly unimodal, relying primarily on historical numerical time-series data. Effectively incorporating sky imagery into an LLM-based forecasting framework remains under-explored and an open challenge. In this paper, we propose SolCloudLLM, an LLM-based multimodal forecasting framework. SolCloudLLM aligns sky-image patches with time-series patches and fuses their corresponding representations through bidirectional multimodal fusion, yielding a unified representation that is subsequently mapped into the embedding space of an LLM. Extensive experiments on the SIRTA and SKIPP'D datasets demonstrate that SolCloudLLM consistently outperforms the best baseline methods in MSE across all forecasting horizons, achieving a maximum relative MSE reduction of 25.4%. Stratified analysis further indicates that the benefits of multimodal fusion are concentrated primarily under cloudy conditions. Notably, SolCloudLLM achieves the best performance in nearly all few-shot settings, whereas other deep learning baselines experience substantial performance degradation and are frequently outperformed by the non-learning physical method.
Read the full abstract

All authors Ken Chen, Maneesha Perera, Wei Wang, Sachith Seneviratne, Hansani Weeratunge, Saman Halgamuge

Computer Vision

LAION-Mobile: Evaluating Deepfake Detectors On One Million Smartphone Photos

Most Deepfake detectors report near-perfect AUC scores on their reference benchmarks. However, a recent ICML position paper argues that these evaluations collectively neglect the impact of modern smartphone photography: the widely used on-device neural image-signal processing pipelines (like multi-sensor fusion or noise and motion-blur suppression) increasingly shift the imaging paradigm from simple lens projections towards computational photography. Hence, devices actually generate, rather than record photos. This increases the risk that deepfake detectors may flag ordinary phone photos as fake. Due to the lack of large-scale datasets containing images from modern smartphones, this hypothesis has so far only been tested in small proof-of-concept studies. The aim of this paper is to close this gap. We introduce LAION-Mobile, an open dataset containing about 1 million smartphone images with EXIF metadata distilled from re-LAION-5B. Evaluating twelve state-of-the-art deepfake detectors with their original paper checkpoints on a 9,115-image evaluation sample of this pool (DIRE on 738), we report three key findings: (i) On modern AI content no detector exceeds AUC 0.624, and five of twelve fall below chance. (ii) Real-photo false-alarm rates are an artefact of threshold calibration: thresholds fitted on legacy GAN data make several detectors look deployable (less than 11 percent FPR), yet the same detectors flag 17-91 percent of real photos once the identical criterion is refit on modern content. (iii) Consequently, no detector both beats chance on modern AI content and keeps a deployable real-photo false-alarm rate. Mirroring the device mix of web collections, the corpus probes the first neural-ISP generation (2018-2020); current flagships are essentially absent, leaving the modern-ISP regime as the open gap.
Read the full abstract

Machine Learning

Phase-Decoupled, Model-Calibrated Power Control for Disaggregated LLM Serving

Datacenter GPU power is the binding constraint on LLM serving capacity, and production serving has shifted to prefill/decode (PD) disaggregation. Deploying NVIDIA's Max-Q inference profile on a disaggregated B200 system, we found its realized gain modest (+8.6% tokens/J), model-dependent, and carrying a mean end-to-end latency cost (+5.2%) that throughput-only evaluation does not surface; the profile also applies one setting to prefill and decode GPUs that operate in opposite hardware regimes. We hypothesize that the optimal power setting is a property of the deployed (model, quantization, engine, hardware) combination rather than of the GPU class, that each lane warrants its own profile, and that converting SLO headroom into energy safely requires latency-gated calibration under a runtime SLO guard rather than a fixed recipe. We present a phase-decoupled, model-calibrated controller: the prefill lane runs under an SM-clock window whose floor is a latency guarantee by construction, and the decode lane under a power cap placed by automatic calibration just above a measured throughput/latency cliff. Because a disaggregated decode lane draws flat, memory-bound power, the cap binds continuously, the reactive-overshoot weakness that led POLCA to reject capping is absent, and the GPU's own power manager retains throughput under the cap. On an 8x B200 node serving Qwen3-Coder-480B (FP8) under agentic load, our balanced mode delivers +20.4% tokens/J at +3.5% mean e2e versus +8.6% at +5.2% for Max-Q, a Pareto improvement on both axes. On Qwen3-235B-A22B (NVFP4) every operating mode meets the ITL-p99 SLO in every repetition; both vendor profiles miss it. A decode-actuator A/B shows the calibrated cap beats static clock locks, and a three-day sustained run saves 32.3% of a lane pair's electricity. Both models are MoE; a dense model recovers roughly 5x less, so we scope our claims to MoE serving.
Read the full abstract

All authors Jae Gon Kim, Donghoon Yoo, Hanyul Ryu, Sungho Ha, Juyeon Lee, Soojung Ryu

Machine Learning

How Wrong Can a Good Predictor Be? Diverging Updates with Vanishing Predictive KL

Accurate posterior prediction need not require accurate approximation of Bayesian updates. We prove that an unbounded gap between the update maps can coexist with vanishing predictive KL for every fixed finite $K\ge2$ in a stationary symmetric Gaussian HMM. Exact Bayesian mixing and an explicit deterministic radial filter act on the same $K-1$ belief coordinates. As $q\to0^+$, their separation in centered logits in the worst case grows at least linearly in the natural confidence scale $L_K(q)$, while their categorical $D_{\mathrm{KL}}(\mathrm{exact}\|\mathrm{radial})$ vanishes at the same explicit witness. Along stationary HMM trajectories, the expected terminal KL between filtered posteriors also converges to zero at $H(q)=\lceil-\log(q)/c\rceil+1$. Typical blocks without switches drive both filters into a common confidence cone, where softmax curvature suppresses their disagreement; a single Gaussian maximal event controls adaptive noise. A sweep with equally spaced Gaussians over $K\in\{2,4,8\}$ illustrates the opposing trends, and binary controls at long horizons compare saturating and nonsaturating recurrences. The result isolates two missing links between internal update gaps and predictive cost: the contribution of separating states to expected loss and decoder sensitivity. Thus even an unbounded internal update gap does not by itself certify predictive failure. The construction is fixed in $K$ and does not provide a universal criterion for when compression is harmless or characterize when internal gaps must incur task loss.
Read the full abstract

All authors Qifu Wen, Shuaijun Liu, Zihan Zhou, Xi Zeng, Ningxin Su

Computation and Language

Rubric-Aligned Disentangled Evaluation of Human Simultaneous Interpreting

Human simultaneous interpreting (SI) is commonly assessed with analytic rubrics separating meaning transfer, delivery quality, and temporal synchrony, yet no automatic metric is designed for rubric-aligned segment-level SI evaluation. We construct a professionally annotated corpus of 1,101 SI segments with scores for meaning transfer (LQ), delivery quality (EXP), and perceived latency (LAT). We show that structured LLM prompting and scalar supervision collapse rubric dimensions, yielding near-zero correlation with human ratings and strong cross-dimension coupling. To isolate supervision structure under identical backbone capacity, we introduce dual regression heads on a LoRA-adapted COMET-KIWI encoder. On a held-out talk-level test set, the model achieves Pearson correlations of 0.388 (LQ) and 0.301 (EXP), improving over frozen COMET-KIWI. Given low absolute rater agreement, we interpret results relative to human consistency and target stable ranking signals for formative assessment.
Read the full abstract

Systems and Control

Exact PID and PI Gain Regions for Uncertain Non-Affine MIMO Systems

Characterizing all PID gains that guarantee uniform exponential regulation of uncertain nonlinear systems remains a fundamental problem. Most existing results provide only sufficient conditions derived from a particular Lyapunov construction. This paper gives exact gain regions, defined directly by uniform regulation over the entire uncertainty class, for two classes of nonaffine MIMO systems with $m\ge2$ controlled coordinates and input channels. For systems of second order under PID control, we independently characterize a common quadratic inner region and an outer region based on a linear subclass. The two regions coincide and therefore characterize the PID gain region exactly at the system level. For systems of first order under PI control, a lossless reduction to a linear family on the uncertainty boundary, combined with a common quadratic certificate, yields the exact PI region. Both regions strictly enlarge existing sufficient regions, and the PID result also provides a sequential tuning rule. A compact linear family with one uncertain parameter shows that common quadratic certification is not generally lossless. These results establish endpoint coincidence and boundary family equivalence as two routes to exact gain regions and motivate the broader question of which nonlinear uncertainty classes admit such characterizations.
Read the full abstract

Computer Vision

ReconPlusGen: Injecting Reconstruction Prior into Multi-view 3D Generation through Noise Inversion and Modulation

Qualitative results and an illustration of our core idea. Top left: reconstruction results on benchmark images. Top right: reconstruction results on real-world images. Bottom: illustration of reconstruction-guided noise initialization and modulation. Given multiple input images, we predict a point cloud in canonical space, deterministically inject the predicted geometry into the diffusion process through noise inversion, and modulate the resulting noise to preserve the generative flexibility required to complete unobserved regions and refine visible geometry.
Read the full abstract

All authors Jiarui Liu, Heng Li, Weiyu Li, Keng Deng, Junyuan Deng, Zheng Zhongxing, Junyu Huang, Jiahao Chang, Xiaoguang Han, Ping Tan

Computation and Language

From Repetition to Recognition: Inductive Discovery of Disinformation Narratives

In disinformation datasets, narratives are often understood as recurring interpretive patterns that group texts under narrative labels. Recent work formalized narrative mining as inductively inferring narrative labels from corpora, but its evaluation stays tied to predefined taxonomies, a closed-world setting that cannot capture narratives absent from the reference labels. We introduce a three-tier evaluation framework for unsupervised narrative label generation: recovery (against a corpus's own taxonomy), mining (against external label sets), and discovery (without predefined labels). Applying it, we compare clustering-based and graph-community-based pipelines across seven disinformation datasets, with human validation of discovery on two. The two families are complementary under automated metrics, but in a corpus with two prominent topics, clustering can reduce one topic to 2% of generated labels while graph-based pipelines stay balanced. Discovery validation also reveals many singletons (narrative labels derived from single claims, 30-62% of graph outputs), which clustering cannot produce. Annotators confirm many as recognizable disinformation narratives, suggesting that in open-world discovery the repetition assumed by narrative mining may be recognized outside the corpus, not within it. We release human-validated narrative candidate labels for the Climate Obstruction and PolyNarrative datasets to support taxonomy development and dataset extension.
Read the full abstract

All authors Max Upravitelev, Veronika Solopova, Jing Yang, Charlott Jakob, Alexandra Tsiakalou, Neda Foroutan, Vera Schmitt

Artificial Intelligence

KuaiRP Series Role-playing Models Technical Report

This paper introduces the complete technical solution for the KuaiRP series of role-playing models. We aim to achieve four core objectives for a dedicated role-playing model: simplified prompt engineering, highly stable output quality, built-in domain world knowledge, and high-efficiency deployment with a small parameter size. However, effectively injecting deep domain knowledge often leads to a severe catastrophic forgetting of the model's general agent capabilities. To overcome this trade-off, we propose a multi-stage training pipeline. First, we design a standardized character template and construct an SFT data pipeline based on user behavior simulation and reverse profile filtering. Next, we utilize a rule-based composite reward function during the Reinforcement Learning (RL) phase to eliminate common degradation phenomena like length expansion and repetitive generation. Finally, to recover the general capabilities compromised during SFT and RL, we propose a novel self-distillation paradigm using Two-stage On-Policy Distillation (OPD) equipped with Cumulative-Divergence Decay (CDD). By using the domain-adapted model as the teacher and the original base model as the student, we effectively balance deep domain knowledge injection with the preservation of general agent capabilities. Experimental results demonstrate that the KuaiRP models not only match the current state-of-the-art proprietary models in role-playing fidelity within our target domains, but also successfully recover general agent capabilities, maintaining extremely low deployment costs.
Read the full abstract

All authors Yipeng Wang, Ziwei Zhang, Jiahui Zhang, Qi Gan, Kai Sheng

Computer Vision

Beyond Benchmarks: Using VLMs to Reveal Systematic Classification Failures Under Real World Conditions

Verification and validation (V&V) of classification models is crucial to enable a wide range of sensor processing applications. Currently, the V&V process relies on time-consuming manual inspection of erroneous samples to find meaningful patterns. This work explores the use of Vision Language Models (VLMs) to speed up this laborious process. VLMs are trained to embed images into a semantically meaningful vector representation, from which human-interpretable systematic errors can be distilled. Deploying such VLM-based methods in a defence context introduces two major challenges: (1) the defence domain is underrepresented in the training data of VLMs, and (2) surroundings and context are less diverse than for other domains. This study provides an initial assessment of the suitability of VLM-based methods for V&V of defence applications. We propose a VLM-based error slice detection (ESD) method that independently groups and labels systematic errors made by a classification model. We demonstrate that this method is able to identify operationally-relevant artificially added perturbations in a non-military dataset. In a military context, our method clusters and describes images based on their surroundings, but also exhibits overlap between cluster descriptions. We further investigate the difference in embedding variation between our military and non-military dataset, which remains a topic of interest. Although the results do not yet warrant fully automated V&V through VLM-based ESD, they show that VLMs could be used to accelerate V&V processes in the future.
Read the full abstract

All authors Dieuwertje Alblas, Alma M. Liezenga, Jan Erik van Woerden, Fedor Taggenbrock, Dalia Aljawaheri, Klamer Schutte

Computers and Society

From Digital Accountability to Accountable Digitality Through Needs-Aware Information Systems: The Case of Auditable Child-Welfare Judgments

Digital accountability research asks how digital systems can, among other aims, be made transparent, explainable, auditable, contestable, and supportive of ongoing learning and improvement. This paper reverses the question: how can digital transformation make established human institutions more accountable? It theorizes this reversal as accountable digitality and specifies needs-aware information systems as the mediating mechanism. The hard and paradigmatic case is child-welfare judgment, where best-interest procedures must protect children, preserve confidentiality, and respect judicial independence while enabling aggregate learning about needs, reasons, exceptions, and disparities. The case is used diagnostically and illustratively to derive and examine the design logic, not as empirical evidence or validation. Conceptual design-oriented analysis decomposes and recombines digital and legal accountability under child-rights constraints, deriving a canonical theory-to-design chain, contingent mechanisms, implications, and safeguards. It advances IS responsibility and ethics research by showing how privacy-preserving, co-created, needs-aware information systems can support institutional self-knowledge and auditable justice.
Read the full abstract

Medical Physics

Phase-contrast micro-CT for intra-operative breast tumour margin assessment using a microfocus x-ray source and photon-counting detector

Objective: Intra-operative tumour margin assessment during breast-conserving surgery requires rapid, high-resolution imaging of excised tissue, allowing the surgical team to take appropriate action within a single operation. This study evaluates a custom propagation-based phase-contrast micro-computed tomography (micro-CT) system designed to meet these clinical constraints without specialised optical elements. Methods: The experimental setup pairs a microfocus x-ray source with a photon-counting detector in a cone-beam geometry. We explore how the spatial coherence of the source can provide propagation-based phase contrast – with no additional specialised optical elements – and balance this against maximising the x-ray flux of the cone-beam geometry. System performance was evaluated across two anode target materials and filtration configurations at various tube power settings. Imaging capabilities were validated using anthropomorphic breast tissue phantoms and a formalin-fixed paraffin-embedded (FFPE) breast tissue specimen, with reconstructions compared against gold-standard histology. Results: An unfiltered tungsten target operated at 40 kVp yielded optimal image quality. The optimised system achieved high-resolution CT reconstructions of a 5 cm diameter sample with an isotropic voxel size of 40.7 $\upmu\text{m}$ in a scan time of 12 minutes. Reconstructed volumes demonstrated strong visual correlation with corresponding histology slides. Conclusion: Combining a microfocus source with a photon-counting detector enables high-resolution, phase-contrast micro-CT within a clinically viable timeframe, demonstrating strong potential for intra-operative margin assessment.
Read the full abstract

All authors Michelle K. Croughan, Timur E. Gureyev, Jane Fox, Mikkaela McCormack, James A. Pollock, Dominic Jurkschat, Stephanie A. Harker, Marcus J. Kitchen

Machine Learning

HERALD: High-Fidelity Exemplar Retrieval with Adaptive Landmark Distillation for Heterophily-Aware Graph Condensation

Graph condensation aims to produce a small surrogate graph that preserves the downstream node-classification performance of a much larger original graph. Existing methods rely on Weisfeiler-Lehman neighbourhood aggregation or gradient-based distribution matching, both of which assume that adjacent nodes share the same label, an assumption that breaks down under heterophily. We propose HERALD (High-fidelity Exemplar Retrieval with Adaptive Landmark Distillation), a gradient-free graph condensation framework that adapts the node scoring and feature selection in the condensation pipeline to the graph's measured heterophily. HERALD selects features via a joint Fisher-discriminability and activation-density criterion that down-weights aggregated representations on heterophilic graphs, and scores nodes by a weighted combination of prototype representativeness, decision-boundary proximity, and Local Intrinsic Dimensionality (LID), where the weights are driven by a smooth sigmoid function of the heterophily ratio. Nodes are then assembled into a condensed subgraph through score-ordered BFS expansion, Personalised PageRank pruning, and class rebalancing, all at an identical storage budget to BONSAI, enabling direct comparison. Experiments on eight benchmark datasets spanning homophilic and heterophilic settings show that HERALD matches or outperforms state-of-the-art condensers on heterophilic graphs and remains competitive on homophilic ones across four GNN architectures.
Read the full abstract

Software Engineering

A Model-Centric DevOps Architecture for DEVS-Based Digital Twin Simulation Services

Digital twin simulation models evolve and are redeployed like software. Engines based on the Discrete Event System Specification (DEVS) have a sound formal basis but offer little support for versioning, automated validation, or continuous delivery in cloud-native environments. This paper proposes a model-centric DevOps architecture that deploys DEVS-based digital twin simulations as managed services. Models are first-class DevOps artefacts written in a declarative language based on YAML (YAML Ain't Markup Language) with a tuple-level mapping to the multiPDEVS formalism. Repository continuous integration (CI) validates publishing manifests, the engine loader performs syntactic, structural, and static semantic checks before every simulation, and published model artefacts are versioned and immutable, so that an earlier validated version can be selected by pinning its identifier. The platform runs as containerised microservices on Kubernetes with engine adaptations that externalise inputs and results and expose lifecycle control. An initial case study on bus route 22 in Riga exercises the lifecycle on an offline model with roughly 115,000 DEVS atomic components and reports single-container throughput and concurrent job execution on one Kubernetes node. Behavioural validation and multi-node evaluation remain future work.
Read the full abstract

All authors Arnis Lektauers, Gusts Linkevičs, Guntis Mosāns, Arina Fokina, Rasa Gulbe

2609.11122 · 10 Sep · v2: revised version. Corrects the case study of v1 (bus route 22 only; the v1 scenario had also included the trolleybus route of the same number) and all reported measurements; adds a Kubernetes concurrency experiment, a SysML v2 comparison and a validation walk-through. Submitted to Software and Systems Modeling (Springer). Under review

Systems and Control

Quantifying the Reality Gap for RL-Based UAV Placement at mmWave and Sub-THz

Reinforcement learning (RL) policies for unmanned aerial vehicle (UAV) placement in mmWave and sub-terahertz networks are typically trained on simplified analytical channels. We quantify the resulting sim-to-real gap on a real urban map of Doha, Qatar, at carriers {28, 140, 183, 300} GHz and altitudes {50, 75, 100, 125} m, evaluating three channel pipelines: an analytical model (FSPL + atmospheric absorption + cuboid LoS), full Monte-Carlo ray tracing in Sionna RT with ITU-R P.676-13 absorption, and a deterministic-LoS hybrid that reuses Sionna's mesh under a closed-form path-gain expression. We formalize the gap on the spatial SNR distribution via four metrics, namely bias, RMSE, Jensen-Shannon divergence, and optimum-deployment displacement. Three findings emerge: at 28/140 GHz, $\sim$70% of the apparent -5.6/-4.8 dB Sionna bias is Monte-Carlo undersampling and shrinks to -1.7/-1.5 dB after mitigation; at 183 GHz a -9.2 dB residual isolates the atmospheric absorption / ITU-R P.676 line-shape disagreement; at 300 GHz the stochastic ray tracer agrees with the analytical model only coincidentally, with a +3.8 dB structural offset exposed by the deterministic-LoS pipeline. Across all carriers the linear-domain regret of the analytical-trained policy stays $\geq$ 0.93, indicating practical near-optimality but with a carrier-resolved SNR bias that warrants explicit reporting.
Read the full abstract

All authors Abdullateef Almohamad, Mostafa Ibrahim, Sabit Ekin, Khalid Qaraqe

Numerical Analysis

Dirichlet-Neumann Waveform Relaxation Method for Hyperbolic PDE with Time Delay in Multiple Subdomains

Hyperbolic partial differential equations (PDEs) with time delay are essential mathematical tools used to model numerous physical systems where wave propagation or oscillations depend heavily on their historical states. As these applications scale in physical complexity, efficient parallel computing techniques become essential; however, developing highly scalable parallel solvers for delayed PDEs remains a significant computational challenge. To address this gap, this study extends the Dirichlet-Neumann Waveform Relaxation (DNWR) method to a multi-subdomain framework explicitly designed for solving time-delayed hyperbolic PDEs in parallel. We advance the underlying mathematical framework and provide a rigorous convergence analysis for both one-dimensional and two-dimensional spatial configurations. A central theoretical contribution of this work is the establishment of finite-step convergence for the multi-domain DNWR algorithm. These theoretical guarantees are firmly corroborated through comprehensive numerical experiments. Furthermore, a detailed comparative analysis against alternative domain decomposition techniques namely NNWR, OSWR, and Classical SWR, highlight the advantages of the proposed method. Finally, this approach provides a robust, highly parallelizable solver capable of efficiently handling complex hyperbolic PDEs with time delay.
Read the full abstract