Evals and data
Collection
Tests and open data for small language models. Scoring code: github.com/udgamlabs/udgam-evals • 6 items • Updated
Datasets:
id stringlengths 13 16 | source stringclasses 2
values | title stringlengths 5 179 | url stringlengths 34 131 | created stringdate 2026-09-01 01:13:27 2026-10-15 21:59:32 | licence stringclasses 2
values | attribution stringlengths 127 498 | text stringlengths 424 11.5k | n_bytes int64 424 11.5k | wiki_page_id int64 79.9M 84.5M ⌀ | wiki_revision_id int64 1.37B 1.38B ⌀ | arxiv_category stringclasses 113
values |
|---|---|---|---|---|---|---|---|---|---|---|---|
arxiv:2610.18880 | arxiv | Do Perturbation Predictors Learn Gene Interactions? Diagnosing Extrapolation in Agentic Screening | https://arxiv.org/abs/2610.18880v1 | 2026-10-01T02:29:02Z | CC0-1.0 | "Do Perturbation Predictors Learn Gene Interactions? Diagnosing Extrapolation in Agentic Screening", arXiv:2610.18880 (https://arxiv.org/abs/2610.18880v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Do Perturbation Predictors Learn Gene Interactions? Diagnosing Extrapolation in Agentic Screening
High-throughput perturbation screens increasingly use machine-learning models to predict the effects of gene combinations that were never measured. A difficult case arises when both genes in a test pair were seen individu... | 1,378 | null | null | cs.LG |
arxiv:2610.01078 | arxiv | Molecular architecture of TE-seeded chromosome-naive initiation networks of protein-protein interactions delineates the Human Protein Map of the assembly of multiprotein complexes | https://arxiv.org/abs/2610.01078v1 | 2026-10-01T05:19:42Z | CC0-1.0 | "Molecular architecture of TE-seeded chromosome-naive initiation networks of protein-protein interactions delineates the Human Protein Map of the assembly of multiprotein complexes", arXiv:2610.01078 (https://arxiv.org/abs/2610.01078v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Molecular architecture of TE-seeded chromosome-naive initiation networks of protein-protein interactions delineates the Human Protein Map of the assembly of multiprotein complexes
Upon binding to transposable elements seeded DNA, initiator transcription factors are poised to activate the hybrid assembly pathways (HAP)... | 2,098 | null | null | q-bio.BM |
arxiv:2610.01086 | arxiv | Multi-Scale Temporal Flows for Peptide Trajectory Generation | https://arxiv.org/abs/2610.01086v1 | 2026-10-01T05:29:38Z | CC0-1.0 | "Multi-Scale Temporal Flows for Peptide Trajectory Generation", arXiv:2610.01086 (https://arxiv.org/abs/2610.01086v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Multi-Scale Temporal Flows for Peptide Trajectory Generation
Peptides occupy a particularly challenging regime of biomolecular dynamics. Short peptides lack a stable folded core and populate broad conformational ensembles, while cyclization adds ring closure, non-local residue coupling, and stereochemical diversity. D... | 1,716 | null | null | q-bio.BM |
arxiv:2610.01142 | arxiv | Protocol for Discovering Convergent Molecular Networks across Multi-Omics Datasets Using Network Propagation | https://arxiv.org/abs/2610.01142v1 | 2026-10-01T06:23:09Z | CC0-1.0 | "Protocol for Discovering Convergent Molecular Networks across Multi-Omics Datasets Using Network Propagation", arXiv:2610.01142 (https://arxiv.org/abs/2610.01142v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Protocol for Discovering Convergent Molecular Networks across Multi-Omics Datasets Using Network Propagation
Phenotypic outcomes arise from coordinated changes across molecular layers. Although multi-omics data capture complementary molecular signals, integrating them into coherent biological mechanisms remains challe... | 739 | null | null | q-bio.MN |
arxiv:2610.01348 | arxiv | Verify Claims, Not Scores: Evidence-Based Verification of Modular Agents | https://arxiv.org/abs/2610.01348v1 | 2026-10-01T09:17:47Z | CC0-1.0 | "Verify Claims, Not Scores: Evidence-Based Verification of Modular Agents", arXiv:2610.01348 (https://arxiv.org/abs/2610.01348v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Verify Claims, Not Scores: Evidence-Based Verification of Modular Agents
When developers change one component of an agent, such as its controller, a learned model or its verifier, they usually judge the change by an aggregate task score. That score cannot tell whether improvement was attainable, which component lost v... | 1,730 | null | null | cs.AI |
arxiv:2610.01358 | arxiv | Fold'EM: Direct atomic structure inference from Cryo-EM particles | https://arxiv.org/abs/2610.01358v2 | 2026-10-01T09:28:37Z | CC0-1.0 | "Fold'EM: Direct atomic structure inference from Cryo-EM particles", arXiv:2610.01358 (https://arxiv.org/abs/2610.01358v2), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Fold'EM: Direct atomic structure inference from Cryo-EM particles
Single-particle cryo-electron microscopy (cryo-EM) has become a widely adopted technique for biomolecular structure determination. The conventional cryo-EM computational pipeline first combines many particle images to reconstruct an electrostatic potent... | 1,987 | null | null | q-bio.BM |
arxiv:2610.01636 | arxiv | Autocatalysis and Boundary Stability in Chemical Reaction Systems | https://arxiv.org/abs/2610.01636v1 | 2026-10-01T13:04:49Z | CC0-1.0 | "Autocatalysis and Boundary Stability in Chemical Reaction Systems", arXiv:2610.01636 (https://arxiv.org/abs/2610.01636v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Autocatalysis and Boundary Stability in Chemical Reaction Systems
The next-generation matrix method is a powerful tool for computing the basic reproduction number in compartmental mathematical models of infectious diseases. The method has been recently extended to mathematical biochemistry, where it can be used to est... | 1,331 | null | null | math.DS |
arxiv:2610.01891 | arxiv | RipplePLM: Structural and Property Decoupling for Protein Mutation Effect Generation | https://arxiv.org/abs/2610.01891v2 | 2026-10-01T15:41:24Z | CC0-1.0 | "RipplePLM: Structural and Property Decoupling for Protein Mutation Effect Generation", arXiv:2610.01891 (https://arxiv.org/abs/2610.01891v2), abstract by the paper's authors. arXiv metadata, CC0 1.0. | RipplePLM: Structural and Property Decoupling for Protein Mutation Effect Generation
Protein mutation effect generation asks a model to describe the functional consequence of a point mutation in natural language. Existing protein-to-text systems typically encode mutation information into undifferentiated representatio... | 1,660 | null | null | cs.CE |
arxiv:2610.02106 | arxiv | Densities for scalar-valued BSDEs via unique continuation and backward uniqueness | https://arxiv.org/abs/2610.02106v1 | 2026-10-01T17:29:40Z | CC0-1.0 | "Densities for scalar-valued BSDEs via unique continuation and backward uniqueness", arXiv:2610.02106 (https://arxiv.org/abs/2610.02106v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Densities for scalar-valued BSDEs via unique continuation and backward uniqueness
We give sufficient conditions ensuring that, at every fixed positive time, the scalar backward component of a Markovian forward-backward stochastic differential equation with multidimensional forward process admits a density with respect... | 1,432 | null | null | math.PR |
arxiv:2610.02175 | arxiv | Effective Resistance and Graph Neural Network Reliability in Tissue-Specific Interactomes | https://arxiv.org/abs/2610.02175v1 | 2026-10-01T17:57:10Z | CC0-1.0 | "Effective Resistance and Graph Neural Network Reliability in Tissue-Specific Interactomes", arXiv:2610.02175 (https://arxiv.org/abs/2610.02175v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Effective Resistance and Graph Neural Network Reliability in Tissue-Specific Interactomes
Protein function annotation needs to know which predictions to distrust, not only what a model predicts. We ask whether tissue-specific interaction structure carries that information. Our candidate signal is effective resistance,... | 1,247 | null | null | cs.LG |
arxiv:2610.03815 | arxiv | GCTAg: scalable mixed-model analysis for biobank-scale agricultural cohorts | https://arxiv.org/abs/2610.03815v1 | 2026-10-01T18:32:38Z | CC0-1.0 | "GCTAg: scalable mixed-model analysis for biobank-scale agricultural cohorts", arXiv:2610.03815 (https://arxiv.org/abs/2610.03815v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | GCTAg: scalable mixed-model analysis for biobank-scale agricultural cohorts
Genome-wide association studies identify genomic variants associated with traits. Mixed linear model association (MLMA) methods using a whole-genome relationship matrix, such as those implemented in GCTA, are powerful but computationally expen... | 911 | null | null | q-bio.GN |
arxiv:2610.03157 | arxiv | Uniqueness of Recursive Utility with Unbounded Consumption and Non-Expected Utility | https://arxiv.org/abs/2610.03157v1 | 2026-10-02T11:31:23Z | CC0-1.0 | "Uniqueness of Recursive Utility with Unbounded Consumption and Non-Expected Utility", arXiv:2610.03157 (https://arxiv.org/abs/2610.03157v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Uniqueness of Recursive Utility with Unbounded Consumption and Non-Expected Utility
We derive sufficient conditions for the existence, non-existence, uniqueness and global attractivity of stochastic recursive utilities. Our approach uses abstract order-theoretic fixed-point arguments, allowing us to accommodate a larg... | 1,200 | null | null | econ.TH |
arxiv:2610.03160 | arxiv | Multimodal reasoning for broadly neutralizing antibody discovery from label-free human B cell repertoires across virus families | https://arxiv.org/abs/2610.03160v1 | 2026-10-02T11:35:34Z | CC0-1.0 | "Multimodal reasoning for broadly neutralizing antibody discovery from label-free human B cell repertoires across virus families", arXiv:2610.03160 (https://arxiv.org/abs/2610.03160v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Multimodal reasoning for broadly neutralizing antibody discovery from label-free human B cell repertoires across virus families
Discovering broadly neutralizing antibodies (bnAbs) from human natural immune repertoires remains a fundamental challenge in immunology, hindered by: the extreme rarity of bnAb, incomplete un... | 1,726 | null | null | q-bio.QM |
arxiv:2610.03161 | arxiv | Landscape-Dependent Performance of Photonic Quantum Solvers in QUBO Feature Selection for Financial Risk Detection | https://arxiv.org/abs/2610.03161v1 | 2026-10-02T11:38:23Z | CC0-1.0 | "Landscape-Dependent Performance of Photonic Quantum Solvers in QUBO Feature Selection for Financial Risk Detection", arXiv:2610.03161 (https://arxiv.org/abs/2610.03161v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Landscape-Dependent Performance of Photonic Quantum Solvers in QUBO Feature Selection for Financial Risk Detection
Feature selection for imbalanced classification tasks such as credit card fraud and consumer default detection requires balancing predictive relevance, inter-feature redundancy, and computational feasibil... | 1,472 | null | null | quant-ph |
arxiv:2610.03314 | arxiv | DAWIS: Data Assimilation with Windowed Inverse Sampling via Multitask Interpolants | https://arxiv.org/abs/2610.03314v1 | 2026-10-02T13:51:00Z | CC0-1.0 | "DAWIS: Data Assimilation with Windowed Inverse Sampling via Multitask Interpolants", arXiv:2610.03314 (https://arxiv.org/abs/2610.03314v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | DAWIS: Data Assimilation with Windowed Inverse Sampling via Multitask Interpolants
Flow- and diffusion-based generative models have recently emerged as flexible and highly efficient forecasting models for dynamical systems. When combined with inference-time guidance, they offer a promising route to high-dimensional no... | 1,678 | null | null | stat.ML |
arxiv:2610.03355 | arxiv | Stochastic thermodynamics and predictability of Ornstein-Uhlenbeck processes: an application to the Madden-Julian Oscillation | https://arxiv.org/abs/2610.03355v1 | 2026-10-02T14:20:46Z | CC0-1.0 | "Stochastic thermodynamics and predictability of Ornstein-Uhlenbeck processes: an application to the Madden-Julian Oscillation", arXiv:2610.03355 (https://arxiv.org/abs/2610.03355v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Stochastic thermodynamics and predictability of Ornstein-Uhlenbeck processes: an application to the Madden-Julian Oscillation
Climate oscillations are intrinsically out-of-equilibrium phenomena, yet the connection between their predictability and nonequilibrium thermodynamic properties remains unexplored. We develop a... | 1,798 | null | null | physics.ao-ph |
arxiv:2610.03406 | arxiv | PreFER: Interactive Robo-Advisor with Scoring Mechanism | https://arxiv.org/abs/2610.03406v1 | 2026-10-02T14:56:22Z | CC0-1.0 | "PreFER: Interactive Robo-Advisor with Scoring Mechanism", arXiv:2610.03406 (https://arxiv.org/abs/2610.03406v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | PreFER: Interactive Robo-Advisor with Scoring Mechanism
We propose an interactive robo-advising framework that learns personalized risk preferences from scores provided by clients. The resulting preference-learning problem is closely related to inverse reinforcement learning (IRL), as the robo-advisor infers the clien... | 1,437 | null | null | q-fin.MF |
arxiv:2610.20640 | arxiv | KNOU-Flood: Advancing global streamflow forecasting with lead-aware probabilistic machine learning | https://arxiv.org/abs/2610.20640v1 | 2026-10-02T14:58:31Z | CC0-1.0 | "KNOU-Flood: Advancing global streamflow forecasting with lead-aware probabilistic machine learning", arXiv:2610.20640 (https://arxiv.org/abs/2610.20640v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | KNOU-Flood: Advancing global streamflow forecasting with lead-aware probabilistic machine learning
Global streamflow forecasting supports flood preparedness, yet data-driven models trained on reanalysis lose skill when driven by operational forecasts, increasingly so with lead time. AIFL addresses this shift with a tw... | 2,005 | null | null | cs.LG |
arxiv:2610.03573 | arxiv | Disentangling epidemics and percolation | https://arxiv.org/abs/2610.03573v1 | 2026-10-02T16:47:51Z | CC0-1.0 | "Disentangling epidemics and percolation", arXiv:2610.03573 (https://arxiv.org/abs/2610.03573v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Disentangling epidemics and percolation
Non-recurrent epidemics on complex networks, such as the paradigmatic susceptible-infected-recovered (SIR) model, are often studied by mapping the model to bond percolation. However, the conditions under which this identification is justified have not been fully understood, exce... | 1,028 | null | null | physics.soc-ph |
arxiv:2610.03856 | arxiv | Exact and asymptotic enumeration of unrestricted binary phylogenetic networks through automorphism weights | https://arxiv.org/abs/2610.03856v1 | 2026-10-02T17:16:05Z | CC0-1.0 | "Exact and asymptotic enumeration of unrestricted binary phylogenetic networks through automorphism weights", arXiv:2610.03856 (https://arxiv.org/abs/2610.03856v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Exact and asymptotic enumeration of unrestricted binary phylogenetic networks through automorphism weights
Let $\cP_{\ell,k}$ be the set of rooted binary phylogenetic networks with $\ell$ labelled leaves, $k$ reticulations and no parallel edges. We write $|\cP_{\ell,k}|=W_k(\ell)+D_k(\ell)$, where the weighted count $... | 1,734 | null | null | math.CO |
arxiv:2610.03685 | arxiv | Boundary inference for the Makeham term in mortality models | https://arxiv.org/abs/2610.03685v2 | 2026-10-02T17:50:37Z | CC0-1.0 | "Boundary inference for the Makeham term in mortality models", arXiv:2610.03685 (https://arxiv.org/abs/2610.03685v2), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Boundary inference for the Makeham term in mortality models
Whether a mortality model needs a Makeham term, the constant representing background mortality, is commonly decided with a likelihood-ratio test. The constant cannot be negative, so testing its absence places it on a boundary, where the usual chi-squared cali... | 1,341 | null | null | stat.ME |
arxiv:2610.03712 | arxiv | RNADyn: A Benchmark for Generating and Understanding RNA Dynamics | https://arxiv.org/abs/2610.03712v1 | 2026-10-02T17:58:08Z | CC0-1.0 | "RNADyn: A Benchmark for Generating and Understanding RNA Dynamics", arXiv:2610.03712 (https://arxiv.org/abs/2610.03712v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | RNADyn: A Benchmark for Generating and Understanding RNA Dynamics
Ribonucleic acid (RNA) functions through conformational changes that are not fully captured by static structures. However, large-scale standardized RNA dynamics data remain limited, and existing approaches typically treat trajectory generation and dynam... | 1,402 | null | null | cs.LG |
arxiv:2610.03978 | arxiv | Learning Latent Protein Languages for Autoregressive Generation | https://arxiv.org/abs/2610.03978v1 | 2026-10-02T19:44:35Z | CC0-1.0 | "Learning Latent Protein Languages for Autoregressive Generation", arXiv:2610.03978 (https://arxiv.org/abs/2610.03978v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Learning Latent Protein Languages for Autoregressive Generation
Autoregressive transformers remain comparatively weak for protein sequence and structure generation. We study the role of target representation: amino acid tokens encode residue identities without explicit contextual semantics, while backbone coordinates ... | 1,955 | null | null | cs.LG |
arxiv:2610.03987 | arxiv | On the sensitivity to hyperparameters of a Bayesian method to infer selection signatures | https://arxiv.org/abs/2610.03987v1 | 2026-10-02T19:50:05Z | CC0-1.0 | "On the sensitivity to hyperparameters of a Bayesian method to infer selection signatures", arXiv:2610.03987 (https://arxiv.org/abs/2610.03987v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | On the sensitivity to hyperparameters of a Bayesian method to infer selection signatures
Motivated by a real-life problem, namely, inferring genomic regions affected by selection, we studied some properties of a two-step method that performs this task based on the FST. This approach exhibits computational simplicity a... | 1,698 | null | null | stat.ME |
arxiv:2610.03992 | arxiv | MathAgent: Multi-Agent Optimization of Mathematical Invariants for Molecular Property Prediction | https://arxiv.org/abs/2610.03992v1 | 2026-10-02T19:55:27Z | CC0-1.0 | "MathAgent: Multi-Agent Optimization of Mathematical Invariants for Molecular Property Prediction", arXiv:2610.03992 (https://arxiv.org/abs/2610.03992v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | MathAgent: Multi-Agent Optimization of Mathematical Invariants for Molecular Property Prediction
Mathematical invariants provide complementary representations of molecular structure, but their predictive utility can vary across datasets and prediction tasks. For a new dataset, appropriate construction of invariant rep... | 1,659 | null | null | q-bio.BM |
arxiv:2610.04090 | arxiv | Fluctuation theorems and exact multi-allele fixation probabilities | https://arxiv.org/abs/2610.04090v1 | 2026-10-02T21:55:29Z | CC0-1.0 | "Fluctuation theorems and exact multi-allele fixation probabilities", arXiv:2610.04090 (https://arxiv.org/abs/2610.04090v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Fluctuation theorems and exact multi-allele fixation probabilities
A mutant gene that appears in a finite population will, in the long run, either disappear from it or become fixed, so that all members of the population will carry a copy of the mutant gene. Determining the fixation probability of a mutant gene is thus... | 1,987 | null | null | q-bio.PE |
arxiv:2610.04122 | arxiv | Cities of Signals: Compression, Separation, and the Geometry of Novelty | https://arxiv.org/abs/2610.04122v1 | 2026-10-02T22:52:37Z | CC0-1.0 | "Cities of Signals: Compression, Separation, and the Geometry of Novelty", arXiv:2610.04122 (https://arxiv.org/abs/2610.04122v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Cities of Signals: Compression, Separation, and the Geometry of Novelty
A cross-sectional signal is a forecast vector over $d$ assets at each date; demeaned and normalized, it is a point on a sphere. Its city is the direction of its time-averaged vector in a common target-aligned frame, a compressed summary. When is t... | 1,978 | null | null | q-fin.PM |
arxiv:2610.04145 | arxiv | Beyond Completion Time: A Multimodal Approach to Characterizing Eye-Hand Coordination During the Nine-Hole Peg Test in Multiple Sclerosis | https://arxiv.org/abs/2610.04145v1 | 2026-10-02T23:40:01Z | CC0-1.0 | "Beyond Completion Time: A Multimodal Approach to Characterizing Eye-Hand Coordination During the Nine-Hole Peg Test in Multiple Sclerosis", arXiv:2610.04145 (https://arxiv.org/abs/2610.04145v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Beyond Completion Time: A Multimodal Approach to Characterizing Eye-Hand Coordination During the Nine-Hole Peg Test in Multiple Sclerosis
Multiple sclerosis (MS) can impair upper-extremity function through motor, sensory, cognitive, and visual deficits. Although the Nine-Hole Peg Test (9-HPT) is widely used to assess ... | 1,497 | null | null | q-bio.NC |
arxiv:2610.04205 | arxiv | Origin of the inhomogeneous linewidth in a solid state $^{229}$Th:CaF$_2$ nuclear clock | https://arxiv.org/abs/2610.04205v1 | 2026-10-03T01:43:05Z | CC0-1.0 | "Origin of the inhomogeneous linewidth in a solid state $^{229}$Th:CaF$_2$ nuclear clock", arXiv:2610.04205 (https://arxiv.org/abs/2610.04205v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Origin of the inhomogeneous linewidth in a solid state $^{229}$Th:CaF$_2$ nuclear clock
Narrow linewidth (high-Q) quantum discriminators are essential for realizing solid-state $^{229}$Th nuclear clocks, yet the observed transition linewidths remain at the tens-to-hundreds-of-kilohertz level, far exceeding the sub-Her... | 1,618 | null | null | physics.atom-ph |
arxiv:2610.04222 | arxiv | Respiratory Mask Testing: Airway Inertance and System Dynamics | https://arxiv.org/abs/2610.04222v1 | 2026-10-03T02:23:15Z | CC0-1.0 | "Respiratory Mask Testing: Airway Inertance and System Dynamics", arXiv:2610.04222 (https://arxiv.org/abs/2610.04222v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Respiratory Mask Testing: Airway Inertance and System Dynamics
Life support equipment coupled to the human lungs forms a complex, nonlinear dynamic system. Current respiratory mask test standards rely on sinusoidal breathing machines and surrogate trachea geometries that substantially reduce physiologically relevant a... | 1,527 | null | null | eess.SY |
arxiv:2610.04223 | arxiv | Matched Triple-Differences: A Framework for Covariate Adjustment | https://arxiv.org/abs/2610.04223v1 | 2026-10-03T02:28:08Z | CC0-1.0 | "Matched Triple-Differences: A Framework for Covariate Adjustment", arXiv:2610.04223 (https://arxiv.org/abs/2610.04223v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Matched Triple-Differences: A Framework for Covariate Adjustment
A common empirical strategy in triple-differences (DDD) is to include either covariate trends or covariate levels to a three-way fixed effects (3WFE) regression. This strategy is typically motivated by the conditional parallel gaps assumption which assum... | 1,466 | null | null | econ.EM |
arxiv:2610.04258 | arxiv | FloVMos: Optical Flow-based Medical Video Mosaicking | https://arxiv.org/abs/2610.04258v1 | 2026-10-03T03:29:39Z | CC0-1.0 | "FloVMos: Optical Flow-based Medical Video Mosaicking", arXiv:2610.04258 (https://arxiv.org/abs/2610.04258v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | FloVMos: Optical Flow-based Medical Video Mosaicking
Biomedical imaging modalities often require a trade-off among resolution, field of view (FOV), and acquisition speed. Video mosaicking offers a strategy to overcome this limitation by computationally stitching sequential high-resolution frames into a wide-FOV compos... | 1,621 | null | null | eess.IV |
arxiv:2610.04337 | arxiv | FLAT: Smoothing the Rugged Landscape for Learnable, Sample-Efficient Traffic Calibration | https://arxiv.org/abs/2610.04337v1 | 2026-10-03T07:05:28Z | CC0-1.0 | "FLAT: Smoothing the Rugged Landscape for Learnable, Sample-Efficient Traffic Calibration", arXiv:2610.04337 (https://arxiv.org/abs/2610.04337v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | FLAT: Smoothing the Rugged Landscape for Learnable, Sample-Efficient Traffic Calibration
Calibrating microscopic traffic models for digital twins is an expensive black-box optimization problem: tuning car-following and lane-changing parameters requires a full simulation run, affording only a tight budget per recalibra... | 1,508 | null | null | physics.soc-ph |
arxiv:2610.04348 | arxiv | A Price-Based Framework for Stochastic Portfolio Theory | https://arxiv.org/abs/2610.04348v1 | 2026-10-03T07:18:09Z | CC0-1.0 | "A Price-Based Framework for Stochastic Portfolio Theory", arXiv:2610.04348 (https://arxiv.org/abs/2610.04348v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | A Price-Based Framework for Stochastic Portfolio Theory
We develop a price-based framework for stochastic portfolio theory in which trading strategies are generated from nominal price weights and evaluated relative to a price-weighted benchmark. Stock splits and reverse splits induce jumps in the generating weights wi... | 1,533 | null | null | q-fin.MF |
arxiv:2610.06947 | arxiv | FactorBench: A Portfolio-Aware Benchmark for Automated Factor Mining | https://arxiv.org/abs/2610.06947v1 | 2026-10-03T10:57:37Z | CC0-1.0 | "FactorBench: A Portfolio-Aware Benchmark for Automated Factor Mining", arXiv:2610.06947 (https://arxiv.org/abs/2610.06947v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | FactorBench: A Portfolio-Aware Benchmark for Automated Factor Mining
Factor mining seeks to discover signals from financial data that predict future asset returns and guide portfolio construction. Automated factor mining now spans genetic programming, reinforcement learning, generative models, and large language model... | 1,391 | null | null | q-fin.PM |
arxiv:2610.04558 | arxiv | ClimateBench v2.0: Probabilistic Climate Model Benchmarking | https://arxiv.org/abs/2610.04558v1 | 2026-10-03T14:49:16Z | CC0-1.0 | "ClimateBench v2.0: Probabilistic Climate Model Benchmarking", arXiv:2610.04558 (https://arxiv.org/abs/2610.04558v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | ClimateBench v2.0: Probabilistic Climate Model Benchmarking
We present ClimateBench v2, a standardized protocol for evaluating climate models on diagnostics expected to be informative for their skill in projecting mid-century regional temperature and precipitation changes. The protocol is designed to evaluate any phys... | 1,939 | null | null | physics.ao-ph |
arxiv:2610.04653 | arxiv | Phase-matching in high-order harmonic generation with a plasma-reshaped driving field | https://arxiv.org/abs/2610.04653v1 | 2026-10-03T16:54:04Z | CC0-1.0 | "Phase-matching in high-order harmonic generation with a plasma-reshaped driving field", arXiv:2610.04653 (https://arxiv.org/abs/2610.04653v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Phase-matching in high-order harmonic generation with a plasma-reshaped driving field
The transition from titanium-sapphire to ytterbium-based laser technology enables turnkey implementations of high-order harmonic generation (HHG) setups at high repetition rates, opening up new applications, e.g., in semiconductor me... | 1,407 | null | null | physics.optics |
arxiv:2610.04688 | arxiv | Submission quotas as self-screening when manuscript production is cheap | https://arxiv.org/abs/2610.04688v2 | 2026-10-03T18:04:38Z | CC0-1.0 | "Submission quotas as self-screening when manuscript production is cheap", arXiv:2610.04688 (https://arxiv.org/abs/2610.04688v2), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Submission quotas as self-screening when manuscript production is cheap
Generative AI enables authors to produce more manuscripts at lower cost. But evaluation by expert editors and referees remains scarce. In this context, the value of an author's submission rights can increase. Major research repositories and journa... | 1,171 | null | null | econ.TH |
arxiv:2610.04697 | arxiv | Efficiency and robustness partition the solution space for memory in recurrent networks | https://arxiv.org/abs/2610.04697v1 | 2026-10-03T18:19:05Z | CC0-1.0 | "Efficiency and robustness partition the solution space for memory in recurrent networks", arXiv:2610.04697 (https://arxiv.org/abs/2610.04697v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Efficiency and robustness partition the solution space for memory in recurrent networks
In computational neuroscience, task-trained recurrent neural networks (RNNs) are commonly used as a testbed for exploring the space of recurrent circuit solutions compatible with a particular function. However, these networks are s... | 1,433 | null | null | q-bio.NC |
arxiv:2610.04732 | arxiv | On the Spread of a Virus and Traveling Waves, Part I | https://arxiv.org/abs/2610.04732v1 | 2026-10-03T19:51:15Z | CC0-1.0 | "On the Spread of a Virus and Traveling Waves, Part I", arXiv:2610.04732 (https://arxiv.org/abs/2610.04732v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | On the Spread of a Virus and Traveling Waves, Part I
This paper develops a one-dimensional model for a mobile viral component propagating through a spatially distributed susceptible population. Random viral motion produces diffusion, whereas a response to the relative gradient of susceptible density generates directed... | 1,580 | null | null | math.DS |
arxiv:2610.04763 | arxiv | Tracing model-generated DNA with position-independent watermarking | https://arxiv.org/abs/2610.04763v1 | 2026-10-03T20:56:43Z | CC0-1.0 | "Tracing model-generated DNA with position-independent watermarking", arXiv:2610.04763 (https://arxiv.org/abs/2610.04763v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Tracing model-generated DNA with position-independent watermarking
Genomic language models can write synthetic DNA that carries no record of its origin. A generation-time watermark could provide a provenance signal, if a verifier can detect it using only the DNA sequence, a key, and published detector settings, withou... | 1,404 | null | null | q-bio.GN |
arxiv:2610.04787 | arxiv | Active-DiNTS: Active Differentiable Network Topology Search | https://arxiv.org/abs/2610.04787v1 | 2026-10-03T22:18:58Z | CC0-1.0 | "Active-DiNTS: Active Differentiable Network Topology Search", arXiv:2610.04787 (https://arxiv.org/abs/2610.04787v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Active-DiNTS: Active Differentiable Network Topology Search
Neural Architecture Search (NAS) has proved to be a strong alternative to manual network design, but applying it to 3D medical image segmentation is limited by two well-known costs, large annotation budgets and multi-GPU clusters. Thus, this paper introduces ... | 1,749 | null | null | cs.CV |
arxiv:2610.04797 | arxiv | Quantitative near-atomic-resolution oxidation-state mapping in complex oxide heterostructures by electron energy-loss spectroscopy | https://arxiv.org/abs/2610.04797v1 | 2026-10-03T22:54:10Z | CC0-1.0 | "Quantitative near-atomic-resolution oxidation-state mapping in complex oxide heterostructures by electron energy-loss spectroscopy", arXiv:2610.04797 (https://arxiv.org/abs/2610.04797v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Quantitative near-atomic-resolution oxidation-state mapping in complex oxide heterostructures by electron energy-loss spectroscopy
We present a theoretical framework for quantitative oxidation state mapping of transition-metal and rare-earth elements from near-atomic-resolution electron energy-loss spectroscopy (EELS)... | 1,845 | null | null | cond-mat.mtrl-sci |
arxiv:2610.04843 | arxiv | Multicatalyst reactions induce abrupt transition in a dense catalytic reaction network model | https://arxiv.org/abs/2610.04843v1 | 2026-10-04T00:55:55Z | CC0-1.0 | "Multicatalyst reactions induce abrupt transition in a dense catalytic reaction network model", arXiv:2610.04843 (https://arxiv.org/abs/2610.04843v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Multicatalyst reactions induce abrupt transition in a dense catalytic reaction network model
We generalize the Furusawa--Kaneko model, a simple model of intracellular catalytic reaction networks, to multicatalyst reactions and analyze the generalized model using dynamical mean-field theory. In the thermodynamic and de... | 1,150 | null | null | cond-mat.stat-mech |
arxiv:2610.04926 | arxiv | Reproducibility and Leakage-Controlled Evaluation of Machine-Learning Models for Chronic Kidney Disease Classification | https://arxiv.org/abs/2610.04926v1 | 2026-10-04T04:10:01Z | CC0-1.0 | "Reproducibility and Leakage-Controlled Evaluation of Machine-Learning Models for Chronic Kidney Disease Classification", arXiv:2610.04926 (https://arxiv.org/abs/2610.04926v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Reproducibility and Leakage-Controlled Evaluation of Machine-Learning Models for Chronic Kidney Disease Classification
Near-perfect classification of small clinical datasets can hinge on preprocessing, data leakage, missing-data handling, and the evaluation sample. We reproduced a published chronic kidney disease (CKD... | 2,021 | null | null | stat.AP |
arxiv:2610.04934 | arxiv | Effect of Added Salts on the Interfacial Dynamics of Monovalent Metal Ions and Model Water-Soluble Polymers | https://arxiv.org/abs/2610.04934v1 | 2026-10-04T04:25:57Z | CC0-1.0 | "Effect of Added Salts on the Interfacial Dynamics of Monovalent Metal Ions and Model Water-Soluble Polymers", arXiv:2610.04934 (https://arxiv.org/abs/2610.04934v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Effect of Added Salts on the Interfacial Dynamics of Monovalent Metal Ions and Model Water-Soluble Polymers
Traditional theories of electrolyte and polyelectrolyte solutions overlook ion and polymer hydration, yielding an overly idealized perspective of the thermodynamic and dynamic properties of these solutions. In p... | 1,818 | null | null | cond-mat.soft |
arxiv:2610.04945 | arxiv | TempoBridge: Source-Conditioned Flow Matching with Optimal Transport Couplings for Single-Cell Population Transitions | https://arxiv.org/abs/2610.04945v2 | 2026-10-04T04:39:07Z | CC0-1.0 | "TempoBridge: Source-Conditioned Flow Matching with Optimal Transport Couplings for Single-Cell Population Transitions", arXiv:2610.04945 (https://arxiv.org/abs/2610.04945v2), abstract by the paper's authors. arXiv metadata, CC0 1.0. | TempoBridge: Source-Conditioned Flow Matching with Optimal Transport Couplings for Single-Cell Population Transitions
Destructive single-cell measurements provide unpaired population snapshots rather than observations of the same cells across conditions. Local cell states and transition requests may also be insufficie... | 1,562 | null | null | cs.LG |
arxiv:2610.04970 | arxiv | Integrated omics reveals actionable drivers of bioactive variation in US milk | https://arxiv.org/abs/2610.04970v1 | 2026-10-04T05:35:46Z | CC0-1.0 | "Integrated omics reveals actionable drivers of bioactive variation in US milk", arXiv:2610.04970 (https://arxiv.org/abs/2610.04970v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Integrated omics reveals actionable drivers of bioactive variation in US milk
Despite milk being a global dietary staple, the molecular basis of its health-relevant bioactivity and the factors governing fine-scale compositional variation remain poorly understood. Here, we present an integrated seven-layer omics charac... | 1,860 | null | null | q-bio.QM |
arxiv:2610.04990 | arxiv | Buying at Support: The Entry Problem under Multi-Skew Brownian Motion | https://arxiv.org/abs/2610.04990v1 | 2026-10-04T06:20:15Z | CC0-1.0 | "Buying at Support: The Entry Problem under Multi-Skew Brownian Motion", arXiv:2610.04990 (https://arxiv.org/abs/2610.04990v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Buying at Support: The Entry Problem under Multi-Skew Brownian Motion
We study when to buy a share that will later be sold optimally, when the price follows a geometric multi-skew Brownian motion whose skew levels model support and resistance. The reward for buying is the exit premium of the liquidation problem solved... | 732 | null | null | q-fin.MF |
arxiv:2610.05145 | arxiv | Metastability and nucleation of bubbles in active chiral systems | https://arxiv.org/abs/2610.05145v1 | 2026-10-04T11:46:27Z | CC0-1.0 | "Metastability and nucleation of bubbles in active chiral systems", arXiv:2610.05145 (https://arxiv.org/abs/2610.05145v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Metastability and nucleation of bubbles in active chiral systems
Bubbles frequently emerge in dense chiral active fluids, yet it remains unclear whether they result from an instability of the homogeneous state or from rare nucleation events within a metastable phase. Here we develop a nucleation theory for chiral bubb... | 1,415 | null | null | cond-mat.soft |
arxiv:2610.05148 | arxiv | A cryogenic apparatus with a rotatable substrate for matrix isolation spectroscopy of BaF in neon crystals down to 3.2 K | https://arxiv.org/abs/2610.05148v1 | 2026-10-04T11:50:46Z | CC0-1.0 | "A cryogenic apparatus with a rotatable substrate for matrix isolation spectroscopy of BaF in neon crystals down to 3.2 K", arXiv:2610.05148 (https://arxiv.org/abs/2610.05148v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | A cryogenic apparatus with a rotatable substrate for matrix isolation spectroscopy of BaF in neon crystals down to 3.2 K
We present a new experimental apparatus for matrix isolation spectroscopy capable of growing cryogenic crystal matrices doped with barium monofluoride (BaF), a candidate for measurements of the elec... | 1,440 | null | null | physics.atom-ph |
arxiv:2610.07006 | arxiv | STOCK-JEPA: Prior-Anchored Latent Revision Representation Learning in Equity Markets | https://arxiv.org/abs/2610.07006v1 | 2026-10-04T12:42:27Z | CC0-1.0 | "STOCK-JEPA: Prior-Anchored Latent Revision Representation Learning in Equity Markets", arXiv:2610.07006 (https://arxiv.org/abs/2610.07006v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | STOCK-JEPA: Prior-Anchored Latent Revision Representation Learning in Equity Markets
Learning effective representations helps characterize the structure and dynamics of equity markets from financial data with a low signal-to-noise ratio. Black-box deep models can capture complex patterns but may overfit sample noise a... | 1,986 | null | null | cs.LG |
arxiv:2610.05182 | arxiv | TIRMamba: A Thermal-Prior-Modulated State-Space Network for Sub-Million-Parameter Infrared Image Super-Resolution | https://arxiv.org/abs/2610.05182v1 | 2026-10-04T12:45:13Z | CC0-1.0 | "TIRMamba: A Thermal-Prior-Modulated State-Space Network for Sub-Million-Parameter Infrared Image Super-Resolution", arXiv:2610.05182 (https://arxiv.org/abs/2610.05182v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | TIRMamba: A Thermal-Prior-Modulated State-Space Network for Sub-Million-Parameter Infrared Image Super-Resolution
Infrared image super-resolution is currently led by Mamba-based networks with 26 to 37 million parameters, which are difficult to deploy on the airborne and handheld platforms where thermal imaging is most... | 1,635 | null | null | eess.IV |
arxiv:2610.05196 | arxiv | Measuring Learned Monotone Temporal Aggregation at Matched Admissibility | https://arxiv.org/abs/2610.05196v2 | 2026-10-04T13:09:01Z | CC0-1.0 | "Measuring Learned Monotone Temporal Aggregation at Matched Admissibility", arXiv:2610.05196 (https://arxiv.org/abs/2610.05196v2), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Measuring Learned Monotone Temporal Aggregation at Matched Admissibility
Risk regulation imposes directional constraints on scores; we adopt their strict per-input form -- the score monotone non-decreasing in every exposure input -- as a normative commitment. Deployed pipelines -- monotone hand-crafted aggregates feed... | 1,984 | null | null | cs.LG |
arxiv:2610.05244 | arxiv | Neutrino Emissivity in $Δ$-direct-Urca Processes Inside Neutron Stars | https://arxiv.org/abs/2610.05244v1 | 2026-10-04T14:21:29Z | CC0-1.0 | "Neutrino Emissivity in $Δ$-direct-Urca Processes Inside Neutron Stars", arXiv:2610.05244 (https://arxiv.org/abs/2610.05244v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Neutrino Emissivity in $Δ$-direct-Urca Processes Inside Neutron Stars
Observed massive neutron stars may contain heavier baryons such as $Δ$ isobars and hyperons in their interiors. In the presence of $Δ$ isobars, the neutrino emitting nucleonic Urca processes can be accompanied by $Δ$-direct-Urca processes involving ... | 1,092 | null | null | nucl-th |
arxiv:2610.05251 | arxiv | What a Shorter Contest Really Costs: Serial Dependence, Effective Trials, and Evidence from a Rule Change in Professional Esports | https://arxiv.org/abs/2610.05251v1 | 2026-10-04T14:26:56Z | CC0-1.0 | "What a Shorter Contest Really Costs: Serial Dependence, Effective Trials, and Evidence from a Rule Change in Professional Esports", arXiv:2610.05251 (https://arxiv.org/abs/2610.05251v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | What a Shorter Contest Really Costs: Serial Dependence, Effective Trials, and Evidence from a Rule Change in Professional Esports
Operational research evaluates tournament formats by their efficacy, the probability that the design identifies the stronger competitor, and computes it almost exclusively by simulation und... | 1,527 | null | null | econ.GN |
arxiv:2610.05263 | arxiv | A Logic for Minor-Free Graph Classes: Model Checking, Dependence, and Combinatorial Reconfiguration | https://arxiv.org/abs/2610.05263v1 | 2026-10-04T14:34:24Z | CC0-1.0 | "A Logic for Minor-Free Graph Classes: Model Checking, Dependence, and Combinatorial Reconfiguration", arXiv:2610.05263 (https://arxiv.org/abs/2610.05263v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | A Logic for Minor-Free Graph Classes: Model Checking, Dependence, and Combinatorial Reconfiguration
We introduce \emph{sub-connectivity logic}, denoted by $\textsf{FOscon}$, an extension of first-order logic for graphs with a specified set $Z$ of admissible edges. Its additional atom $\textsf{scon}(x,y;\bar z)$ assert... | 1,942 | null | null | cs.LO |
arxiv:2610.05367 | arxiv | AIProver: Agentic Auto-Formalization of Mathematical Research via Certificate-Driven Evolving Harness | https://arxiv.org/abs/2610.05367v1 | 2026-10-04T16:48:12Z | CC0-1.0 | "AIProver: Agentic Auto-Formalization of Mathematical Research via Certificate-Driven Evolving Harness", arXiv:2610.05367 (https://arxiv.org/abs/2610.05367v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | AIProver: Agentic Auto-Formalization of Mathematical Research via Certificate-Driven Evolving Harness
Proof auto-formalization translates natural-language (NL) theorems and proofs into a formal language (FL) such as Lean, enabling mechanical verification. Despite rapid progress, research-level proofs often depend on c... | 2,020 | null | null | cs.LO |
arxiv:2610.05405 | arxiv | Impulse Response Inference for Matrix Autoregressions | https://arxiv.org/abs/2610.05405v1 | 2026-10-04T17:42:36Z | CC0-1.0 | "Impulse Response Inference for Matrix Autoregressions", arXiv:2610.05405 (https://arxiv.org/abs/2610.05405v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Impulse Response Inference for Matrix Autoregressions
Matrix autoregressive (MAR) models offer a parsimonious framework for modeling matrix-valued time series, yet tools for estimation and inference for their impulse response functions are lacking. We develop asymptotic and bootstrap-based inference for impulse respon... | 1,039 | null | null | econ.EM |
arxiv:2610.05419 | arxiv | Recurrent network dynamics explain the time course of perceptual grouping in natural scenes | https://arxiv.org/abs/2610.05419v1 | 2026-10-04T18:00:46Z | CC0-1.0 | "Recurrent network dynamics explain the time course of perceptual grouping in natural scenes", arXiv:2610.05419 (https://arxiv.org/abs/2610.05419v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Recurrent network dynamics explain the time course of perceptual grouping in natural scenes
How the brain groups image elements into coherent objects in natural scenes remains unclear. Existing models for grouping in biological vision rely on simplified stimuli with explicit boundaries and cannot explain human behavio... | 1,263 | null | null | q-bio.NC |
arxiv:2610.05426 | arxiv | Distortion-Free High-Resolution PROPELLER-DWI: Technical Advances and Initial Demonstration for the Prostate | https://arxiv.org/abs/2610.05426v1 | 2026-10-04T18:07:13Z | CC0-1.0 | "Distortion-Free High-Resolution PROPELLER-DWI: Technical Advances and Initial Demonstration for the Prostate", arXiv:2610.05426 (https://arxiv.org/abs/2610.05426v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Distortion-Free High-Resolution PROPELLER-DWI: Technical Advances and Initial Demonstration for the Prostate
Purpose: To develop and evaluate an integrated acquisition and reconstruction framework for distortion-free, high-resolution prostate diffusion-weighted imaging (DWI). Methods: PROPELLER-DWI combines pulsed-gra... | 1,811 | null | null | eess.IV |
arxiv:2610.05465 | arxiv | Tao-Travkin equivalence revisited | https://arxiv.org/abs/2610.05465v1 | 2026-10-04T19:12:24Z | CC0-1.0 | "Tao-Travkin equivalence revisited", arXiv:2610.05465 (https://arxiv.org/abs/2610.05465v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Tao-Travkin equivalence revisited
We give a simpler proof of the theorem of J.~Tao and R.~Travkin saying that (the neutral block of) the affine Hecke category is the monoidal colimit of finite Hecke categories. The argument is axiomatized to apply to various sheaf-theoretic and group-theoretic contexts. We also prove ... | 424 | null | null | math.RT |
arxiv:2610.05486 | arxiv | Surrogate-powered Causal Inference with Censored Outcomes | https://arxiv.org/abs/2610.05486v1 | 2026-10-04T19:49:01Z | CC0-1.0 | "Surrogate-powered Causal Inference with Censored Outcomes", arXiv:2610.05486 (https://arxiv.org/abs/2610.05486v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Surrogate-powered Causal Inference with Censored Outcomes
Clinical trials with survival endpoints lose information when participants are censored before death is observed. We develop target-preserving estimators that use posttreatment disease history, such as recurrence or progression, to recover information lost to c... | 1,613 | null | null | math.ST |
arxiv:2610.05556 | arxiv | How Simple Models Can Reproduce Multiple Epidemic Waves: Peak Topology, Minimal Dynamic Complexity, and Inference for Time-Varying Transmission | https://arxiv.org/abs/2610.05556v1 | 2026-10-04T21:44:53Z | CC0-1.0 | "How Simple Models Can Reproduce Multiple Epidemic Waves: Peak Topology, Minimal Dynamic Complexity, and Inference for Time-Varying Transmission", arXiv:2610.05556 (https://arxiv.org/abs/2610.05556v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | How Simple Models Can Reproduce Multiple Epidemic Waves: Peak Topology, Minimal Dynamic Complexity, and Inference for Time-Varying Transmission
Multiple epidemic waves are often taken as evidence that simple compartmental models are structurally inadequate. Yet the key question is not whether a simple model can reprod... | 2,021 | null | null | q-bio.PE |
arxiv:2610.05648 | arxiv | $^{86}$Ni: Possible doubly-magic nucleus with neutron halo | https://arxiv.org/abs/2610.05648v1 | 2026-10-05T00:52:16Z | CC0-1.0 | "$^{86}$Ni: Possible doubly-magic nucleus with neutron halo", arXiv:2610.05648 (https://arxiv.org/abs/2610.05648v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | $^{86}$Ni: Possible doubly-magic nucleus with neutron halo
Self-consistent mean-field calculations with semi-realistic interactions have predicted that $N=58$ becomes a magic number near Ni, where the $n2s_{1/2}$ orbit is the highest occupied state. As the last neutrons occupy the shallowly bound $s$-orbit, $^{86}$Ni ... | 794 | null | null | nucl-th |
arxiv:2610.05650 | arxiv | Payment Design under Belief Disagreement: Rent Placement, Bid Resolution, and Auction Choice | https://arxiv.org/abs/2610.05650v1 | 2026-10-05T00:53:05Z | CC0-1.0 | "Payment Design under Belief Disagreement: Rent Placement, Bid Resolution, and Auction Choice", arXiv:2610.05650 (https://arxiv.org/abs/2610.05650v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Payment Design under Belief Disagreement: Rent Placement, Bid Resolution, and Auction Choice
When auction bidders misperceive their rivals, a seller can change not only the amount of information rent but also the states in which it is delivered. We ask when this payment-design opportunity survives a finite bidding int... | 1,786 | null | null | cs.GT |
arxiv:2610.05675 | arxiv | A Superprocess-based Approach to Rough CIR Processes and Feller Random Measures | https://arxiv.org/abs/2610.05675v1 | 2026-10-05T01:42:42Z | CC0-1.0 | "A Superprocess-based Approach to Rough CIR Processes and Feller Random Measures", arXiv:2610.05675 (https://arxiv.org/abs/2610.05675v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | A Superprocess-based Approach to Rough CIR Processes and Feller Random Measures
Feller random measures generalize the Feller diffusion (the CIR process) by giving it memory. They arise as the scaling limits of nearly unstable Hawkes processes, and include the rough CIR process and its hyper-rough and discontinuous rel... | 1,158 | null | null | math.PR |
arxiv:2610.05676 | arxiv | An evolutionary origin of collective decision making in humans and machines | https://arxiv.org/abs/2610.05676v1 | 2026-10-05T01:42:46Z | CC0-1.0 | "An evolutionary origin of collective decision making in humans and machines", arXiv:2610.05676 (https://arxiv.org/abs/2610.05676v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | An evolutionary origin of collective decision making in humans and machines
Groups of individuals can solve collective problems more accurately than any single member, by aggregating their opinions. Recent theoretical work has identified individual-level reward schemes that allow uninformed individuals to evolve colle... | 1,816 | null | null | physics.soc-ph |
arxiv:2610.07050 | arxiv | Spontaneous Motion Generates Reversible Nonreciprocity in an Achiral Active Elastic Ring | https://arxiv.org/abs/2610.07050v1 | 2026-10-05T02:39:09Z | CC0-1.0 | "Spontaneous Motion Generates Reversible Nonreciprocity in an Achiral Active Elastic Ring", arXiv:2610.07050 (https://arxiv.org/abs/2610.07050v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Spontaneous Motion Generates Reversible Nonreciprocity in an Achiral Active Elastic Ring
Nonreciprocal mechanical response is usually associated with built-in directional couplings, structural chirality, or external driving. Whether an achiral active body can instead generate and reverse such directionality through it... | 1,001 | null | null | cond-mat.soft |
arxiv:2610.08856 | arxiv | Dynamical Insights into the Current Control Strategies for Chikungunya | https://arxiv.org/abs/2610.08856v2 | 2026-10-05T03:10:27Z | CC0-1.0 | "Dynamical Insights into the Current Control Strategies for Chikungunya", arXiv:2610.08856 (https://arxiv.org/abs/2610.08856v2), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Dynamical Insights into the Current Control Strategies for Chikungunya
This paper addresses the challenge that, under China's current mosquito-control policy for chikungunya---\textit{clearing stagnant water, killing adult mosquitoes, and preventing bites}---epidemics can still recur and continue to rise. By construct... | 1,441 | null | null | q-bio.PE |
arxiv:2610.05740 | arxiv | Latent Continuum of Regimes in Limit Order Book Dynamics | https://arxiv.org/abs/2610.05740v2 | 2026-10-05T03:44:47Z | CC0-1.0 | "Latent Continuum of Regimes in Limit Order Book Dynamics", arXiv:2610.05740 (https://arxiv.org/abs/2610.05740v2), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Latent Continuum of Regimes in Limit Order Book Dynamics
Market-regime models typically assume a finite set of discrete latent states. We examine whether high frequency limit-order-book dynamics exhibit distinct regime separation or apparent regimes result from discretising an underlying continuum, analysing deep limi... | 1,971 | null | null | q-fin.TR |
arxiv:2610.05741 | arxiv | A Dirichlet Mixed-Membership Model for Exact Multivariate Distributional Credibility | https://arxiv.org/abs/2610.05741v1 | 2026-10-05T03:45:00Z | CC0-1.0 | "A Dirichlet Mixed-Membership Model for Exact Multivariate Distributional Credibility", arXiv:2610.05741 (https://arxiv.org/abs/2610.05741v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | A Dirichlet Mixed-Membership Model for Exact Multivariate Distributional Credibility
Credibility theory combines individual experience with portfolio information for insurance pricing, but classical formulations focus primarily on conditional means and expected premiums. We propose a Dirichlet mixed-membership model (... | 1,614 | null | null | stat.ME |
arxiv:2610.05804 | arxiv | Multi-type branching diffusions with small mutation rates | https://arxiv.org/abs/2610.05804v1 | 2026-10-05T04:55:34Z | CC0-1.0 | "Multi-type branching diffusions with small mutation rates", arXiv:2610.05804 (https://arxiv.org/abs/2610.05804v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Multi-type branching diffusions with small mutation rates
Approximate solutions are found for Feller-like neutral multi-type branching diffusions $\big{(}\mathbf{X}(t)\big{)}_{t \in \mathbb{R}_{\ge 0}}$ in the limit of small mutation rates. The method employed involves solving approximations to the Laplace transformed... | 1,035 | null | null | math.PR |
arxiv:2610.05823 | arxiv | Who Keeps the Gains from Personal AI Assistants? Seller Adaptation and the Unassisted in a Language-Model Market Simulation | https://arxiv.org/abs/2610.05823v1 | 2026-10-05T05:20:51Z | CC0-1.0 | "Who Keeps the Gains from Personal AI Assistants? Seller Adaptation and the Unassisted in a Language-Model Market Simulation", arXiv:2610.05823 (https://arxiv.org/abs/2610.05823v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Who Keeps the Gains from Personal AI Assistants? Seller Adaptation and the Unassisted in a Language-Model Market Simulation
Personal AI assistants are beginning to transact for consumers, and early adopters capture real savings. Whether those savings survive, and what happens to consumers who have no assistant, depend... | 1,845 | null | null | cs.MA |
arxiv:2610.05836 | arxiv | Shape Transformations and Rupture of a Vesicle Driven by an Encapsulated Active Particle | https://arxiv.org/abs/2610.05836v1 | 2026-10-05T05:38:45Z | CC0-1.0 | "Shape Transformations and Rupture of a Vesicle Driven by an Encapsulated Active Particle", arXiv:2610.05836 (https://arxiv.org/abs/2610.05836v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Shape Transformations and Rupture of a Vesicle Driven by an Encapsulated Active Particle
Biological membranes are highly deformable structures that undergo shape transformations during numerous cellular processes. These transformations are often driven by active agents, including molecular motors, cytoskeletal machine... | 1,499 | null | null | cond-mat.soft |
arxiv:2610.05847 | arxiv | Classification of reductive monoids | https://arxiv.org/abs/2610.05847v1 | 2026-10-05T06:03:40Z | CC0-1.0 | "Classification of reductive monoids", arXiv:2610.05847 (https://arxiv.org/abs/2610.05847v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Classification of reductive monoids
We classify reductive monoids over arbitrary base schemes by descent from the split classification in arXiv:2607.00322. For a reductive group scheme $G$, we construct its sheaf of based root data, abstract Cartan torus, abstract Weyl group, and a scheme of Weyl cones. The last of th... | 634 | null | null | math.RT |
arxiv:2610.07060 | arxiv | Skillful Data-Driven Subseasonal Soil Moisture Forecasting: Prospects and Limits for Flash Drought Prediction | https://arxiv.org/abs/2610.07060v1 | 2026-10-05T06:34:10Z | CC0-1.0 | "Skillful Data-Driven Subseasonal Soil Moisture Forecasting: Prospects and Limits for Flash Drought Prediction", arXiv:2610.07060 (https://arxiv.org/abs/2610.07060v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Skillful Data-Driven Subseasonal Soil Moisture Forecasting: Prospects and Limits for Flash Drought Prediction
Despite substantial progress in short-to-medium-range weather forecasting, predicting high-impact events such as flash droughts remains a key challenge for both early warning operations and physically-based su... | 1,586 | null | null | cs.LG |
arxiv:2610.05895 | arxiv | Bifurcation of laser-dressed band extrema in solids under intense low-frequency fields | https://arxiv.org/abs/2610.05895v1 | 2026-10-05T07:12:20Z | CC0-1.0 | "Bifurcation of laser-dressed band extrema in solids under intense low-frequency fields", arXiv:2610.05895 (https://arxiv.org/abs/2610.05895v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Bifurcation of laser-dressed band extrema in solids under intense low-frequency fields
We show that an intense low-frequency laser field can qualitatively modify the local extremal structure of an interband energy gap in a crystal. Using an instantaneous laser-dressed basis obtained from the nonperturbative diagonaliz... | 1,458 | null | null | physics.atom-ph |
arxiv:2610.05913 | arxiv | CoHyFuse: Condition-wise Hypergraph Fusion with Global Connectome in Task-fMRI | https://arxiv.org/abs/2610.05913v1 | 2026-10-05T07:27:14Z | CC0-1.0 | "CoHyFuse: Condition-wise Hypergraph Fusion with Global Connectome in Task-fMRI", arXiv:2610.05913 (https://arxiv.org/abs/2610.05913v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | CoHyFuse: Condition-wise Hypergraph Fusion with Global Connectome in Task-fMRI
Task-fMRI connectomes reveal state-dependent neural reconfigurations, yet conventional methods marginalize these signals by aggregating distinct conditions into static pairwise graphs, thereby obscuring condition-specific multi-ROI organiza... | 1,908 | null | null | cs.LG |
arxiv:2610.05926 | arxiv | Comparing two approaches for modelling the loss given default of credit cards: Run-off triangles vs regression | https://arxiv.org/abs/2610.05926v1 | 2026-10-05T07:45:57Z | CC0-1.0 | "Comparing two approaches for modelling the loss given default of credit cards: Run-off triangles vs regression", arXiv:2610.05926 (https://arxiv.org/abs/2610.05926v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Comparing two approaches for modelling the loss given default of credit cards: Run-off triangles vs regression
The use of run-off triangles (ROTs) is a common industry practice in estimating the loss given default (LGD) risk parameter when predicting credit losses in banking. We benchmark this industry practice using ... | 1,552 | null | null | q-fin.RM |
arxiv:2610.06007 | arxiv | Seventy Years of Identifying the Phillips Curve and the Policy Rule | https://arxiv.org/abs/2610.06007v1 | 2026-10-05T09:04:21Z | CC0-1.0 | "Seventy Years of Identifying the Phillips Curve and the Policy Rule", arXiv:2610.06007 (https://arxiv.org/abs/2610.06007v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Seventy Years of Identifying the Phillips Curve and the Policy Rule
Two coefficients enter nearly every quantitative argument about monetary policy: the Phillips-curve slope and the policy rule's response to inflation. For seventy years, each generation that estimated them identified the coefficient conditional on ass... | 982 | null | null | econ.GN |
arxiv:2610.06022 | arxiv | Interplay between final-state nucleon distortion, nuclear transparency and intranuclear cascade models in neutrino event generators | https://arxiv.org/abs/2610.06022v1 | 2026-10-05T09:20:44Z | CC0-1.0 | "Interplay between final-state nucleon distortion, nuclear transparency and intranuclear cascade models in neutrino event generators", arXiv:2610.06022 (https://arxiv.org/abs/2610.06022v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Interplay between final-state nucleon distortion, nuclear transparency and intranuclear cascade models in neutrino event generators
We present an overview of the relativistic distorted-wave impulse approximation (RDWIA) for lepton-nucleus scattering, with emphasis on its use as an elementary-vertex model in event gene... | 1,449 | null | null | nucl-th |
arxiv:2610.06027 | arxiv | Unraveling phenotypic variation in multiscale behavioral dynamics | https://arxiv.org/abs/2610.06027v2 | 2026-10-05T09:26:23Z | CC0-1.0 | "Unraveling phenotypic variation in multiscale behavioral dynamics", arXiv:2610.06027 (https://arxiv.org/abs/2610.06027v2), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Unraveling phenotypic variation in multiscale behavioral dynamics
Behavioral variability reflects differences in internal and external parameters and provides a substrate for natural selection. However, uncovering its structure is challenging because behavior evolves nonlinearly across multiple timescales. Our goal is... | 1,557 | null | null | physics.bio-ph |
arxiv:2610.06034 | arxiv | How well do routinely collected demographic and clinical variables aid point-of-care lung ultrasound TB classification | https://arxiv.org/abs/2610.06034v1 | 2026-10-05T09:32:17Z | CC0-1.0 | "How well do routinely collected demographic and clinical variables aid point-of-care lung ultrasound TB classification", arXiv:2610.06034 (https://arxiv.org/abs/2610.06034v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | How well do routinely collected demographic and clinical variables aid point-of-care lung ultrasound TB classification
We consider the fusion of lung ultrasound images with routinely-collected clinical and demographic data for the purpose of automated tuberculosis (TB) screening using deep-learning. Such deep-learning... | 1,917 | null | null | eess.IV |
arxiv:2610.06065 | arxiv | From communication to computation in neurons-on-a-chip: an in silico study of neurotopomorphic computing | https://arxiv.org/abs/2610.06065v1 | 2026-10-05T10:01:31Z | CC0-1.0 | "From communication to computation in neurons-on-a-chip: an in silico study of neurotopomorphic computing", arXiv:2610.06065 (https://arxiv.org/abs/2610.06065v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | From communication to computation in neurons-on-a-chip: an in silico study of neurotopomorphic computing
Living neuronal networks transform inputs through recurrent cellular and population dynamics, yet it is unknown which network architecture supports which computation. Neurons-on-a-chip turn this question into a des... | 1,255 | null | null | q-bio.NC |
arxiv:2610.06074 | arxiv | Preemptive defense against re-identification attacks on voice anonymization via adversarial perturbation | https://arxiv.org/abs/2610.06074v1 | 2026-10-05T10:06:25Z | CC0-1.0 | "Preemptive defense against re-identification attacks on voice anonymization via adversarial perturbation", arXiv:2610.06074 (https://arxiv.org/abs/2610.06074v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Preemptive defense against re-identification attacks on voice anonymization via adversarial perturbation
Voice anonymization (VA) is used to conceal the voice identities in speech data. Performance is usually estimated using automatic speaker verification (ASV) as a proxy to judge the capability of an attacker to re-i... | 1,219 | null | null | eess.AS |
arxiv:2610.06099 | arxiv | A Renormalization-Group Hierarchy of Stochastic Effective Dynamics Learned through Path Integrals | https://arxiv.org/abs/2610.06099v1 | 2026-10-05T10:32:07Z | CC0-1.0 | "A Renormalization-Group Hierarchy of Stochastic Effective Dynamics Learned through Path Integrals", arXiv:2610.06099 (https://arxiv.org/abs/2610.06099v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | A Renormalization-Group Hierarchy of Stochastic Effective Dynamics Learned through Path Integrals
This study combines a stochastic renormalization group (RG) in space with a path-integral description of the time evolution, giving a spatial hierarchy of coarse-grained dynamics together with the distribution over spatio... | 1,669 | null | null | nlin.CD |
arxiv:2610.06107 | arxiv | Diffusion Meets Unrolling: Compressive SAR Image Reconstruction with Interleaved Learned Corrections | https://arxiv.org/abs/2610.06107v1 | 2026-10-05T10:40:03Z | CC0-1.0 | "Diffusion Meets Unrolling: Compressive SAR Image Reconstruction with Interleaved Learned Corrections", arXiv:2610.06107 (https://arxiv.org/abs/2610.06107v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Diffusion Meets Unrolling: Compressive SAR Image Reconstruction with Interleaved Learned Corrections
Compressive Synthetic Aperture Radar (SAR) imaging, typically formulated as an inverse problem and solved with traditional iterative optimisation methods, can be very computationally expensive. We investigate the use o... | 1,221 | null | null | eess.IV |
arxiv:2610.06118 | arxiv | Preferential centrality with a density penalty: formulation and an open-source implementation | https://arxiv.org/abs/2610.06118v1 | 2026-10-05T10:49:28Z | CC0-1.0 | "Preferential centrality with a density penalty: formulation and an open-source implementation", arXiv:2610.06118 (https://arxiv.org/abs/2610.06118v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Preferential centrality with a density penalty: formulation and an open-source implementation
Preferential centrality (Hellervik, Nilsson and Andersson, 2019) models spatial activity through feedback between activity and attraction. We extend it with a density penalty. The local response depends on activity per unit o... | 1,632 | null | null | physics.soc-ph |
arxiv:2610.06119 | arxiv | Power enhancement via cross-fit variance estimation: Applications to specification, overidentification, and many-restriction testing | https://arxiv.org/abs/2610.06119v1 | 2026-10-05T10:50:04Z | CC0-1.0 | "Power enhancement via cross-fit variance estimation: Applications to specification, overidentification, and many-restriction testing", arXiv:2610.06119 (https://arxiv.org/abs/2610.06119v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Power enhancement via cross-fit variance estimation: Applications to specification, overidentification, and many-restriction testing
Quadratic-form test statistics are widely used in econometrics, and their performance depends on accurate variance estimation. Conventional plug-in estimators are consistent under the nu... | 1,720 | null | null | econ.EM |
arxiv:2610.06146 | arxiv | Impact of Data Augmentation on Confidence Calibration in Melanoma Classification | https://arxiv.org/abs/2610.06146v1 | 2026-10-05T11:18:42Z | CC0-1.0 | "Impact of Data Augmentation on Confidence Calibration in Melanoma Classification", arXiv:2610.06146 (https://arxiv.org/abs/2610.06146v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Impact of Data Augmentation on Confidence Calibration in Melanoma Classification
Accurately quantifying the predictive uncertainty or improving model calibration plays an important role in medical image classification, in particular in melanoma diagnosis, where accurate uncertainty quantification can have significant ... | 1,131 | null | null | cs.AI |
arxiv:2610.06163 | arxiv | Where Did the Repair First Go Wrong? Localizing the Origins of Silent Failures in Agentic Vulnerability Repair | https://arxiv.org/abs/2610.06163v1 | 2026-10-05T11:31:16Z | CC0-1.0 | "Where Did the Repair First Go Wrong? Localizing the Origins of Silent Failures in Agentic Vulnerability Repair", arXiv:2610.06163 (https://arxiv.org/abs/2610.06163v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Where Did the Repair First Go Wrong? Localizing the Origins of Silent Failures in Agentic Vulnerability Repair
Localizing where an LLM-based agent first fails to uphold security during a repair can show which stage of its workflow needs an additional safeguard. This is difficult for silent failures, which are patches ... | 1,470 | null | null | cs.CR |
arxiv:2610.06164 | arxiv | Predicting Engagement with Sponsored Content Across Account Types on Instagram | https://arxiv.org/abs/2610.06164v1 | 2026-10-05T11:33:44Z | CC0-1.0 | "Predicting Engagement with Sponsored Content Across Account Types on Instagram", arXiv:2610.06164 (https://arxiv.org/abs/2610.06164v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Predicting Engagement with Sponsored Content Across Account Types on Instagram
In this work we explore how users interact with sponsored content on social media platforms by collecting and analyzing a large-scale dataset of sponsored Instagram posts. To maintain transparency, we favor robust statistical analysis and e... | 1,045 | null | null | cs.SI |
arxiv:2610.06168 | arxiv | Model Risk Analysis for Entropic Hedging Strategies | https://arxiv.org/abs/2610.06168v1 | 2026-10-05T11:39:06Z | CC0-1.0 | "Model Risk Analysis for Entropic Hedging Strategies", arXiv:2610.06168 (https://arxiv.org/abs/2610.06168v1), abstract by the paper's authors. arXiv metadata, CC0 1.0. | Model Risk Analysis for Entropic Hedging Strategies
Entropic risk optimisation is a general framework for pricing and hedging financial derivatives in incomplete markets that can be used to decompose P&L into market and model risk contributions. When the prices are quadratic Gaussian, the coupled equations for price a... | 864 | null | null | q-fin.MF |
End of preview. Expand in Data Studio
UDGAM LABS · Evaluation set · 1,670 documents written after 1 September 2026
Fresh Text, October 2026
1,670 English documents that did not exist before September 2026: 542 Wikipedia articles first created between 1 September and 15 October 2026, and 1,128 arXiv abstracts submitted between 1 and 15 October 2026, 3.9 MB of text in all. No model trained before September 2026 can have seen this text, so a model's bits per byte on it measures how well the model predicts new writing, not how much of a public test it has memorised. Scoring a 100–400M-parameter model takes two to five minutes on an Apple-silicon Mac.
Code: github.com/udgamlabs/udgam-evals · Companion test: Udgam Probe Suite · udgamlabs.com
Related work: UncheatableEval
This set follows the idea of UncheatableEval (Tan, Li and Shen, arXiv:2609.27510): score base models on text published after they were trained. UncheatableEval is broader. Its newest snapshot (July 2026, published 27 July) covers 15 categories, including code, news, fiction and non-English Wikipedia, and its leaderboard reaches far larger models. This set is a narrow companion for small models:
- A newer window. Text from 1 September to 15 October 2026.
- A licence on every row. Only sources whose licences allow redistribution (Wikipedia, CC BY-SA 4.0; arXiv metadata, CC0 1.0), with a revision permalink and a credit line per row. UncheatableEval's dataset cards do not state a licence.
- Documented provenance. Wikipedia pages are kept only if their first revision falls in the window, pages that were copied, split, merged, translated or restored are dropped, and pages Wikipedia editors later deleted or tagged as suspected AI-generated text are removed.
- Uncertainty with every comparison. A paired bootstrap gives a 95% interval for the gap between any two models.
- Results for small models, from 90M to 384M parameters, alongside the Udgam Probe Suite.
Quick start
Score any Hugging Face causal language model:
git clone https://github.com/udgamlabs/udgam-evals && cd udgam-evals
pip install -r requirements.txt
python fresh_text/score.py HuggingFaceTB/SmolLM2-135M --out results/SmolLM2-135M.json
Or load the documents:
from datasets import load_dataset
ds = load_dataset("udgamlabs/fresh-text-2026-10", split="test")
print(ds[0]["title"], ds[0]["url"])
The metric: bits per byte
The model reads each document one token at a time and gives a probability to the token that actually comes next. The cost of a token is −log₂ of that probability, in bits. Add up the bits over every document and divide by the total number of UTF-8 bytes of text. Lower is better. Because the score is per byte rather than per token, models with different tokenizers are compared fairly.
Settings used for every number on this page: each document is scored on its own; the model's start-of-text (BOS) token comes first when its tokenizer defines one; documents longer than 2,048 tokens are read with a 2,048-token window that slides by 1,024 tokens, so every token is scored exactly once; weights in float32.
Results
Bits per byte, lower is better. Measured on 17 October 2026 on an Apple-silicon Mac (PyTorch, MPS, float32) with the
settings above. The runs scored the 1,695 documents of the first build; the numbers below are recomputed from their
per-document results for the 1,670 published documents (see How it was made). Re-running fresh_text/score.py on
this dataset on 21 October 2026 gave SmolLM2-135M 0.944, the same as the table.
| Model | Parameters | All | Wikipedia | arXiv |
|---|---|---|---|---|
| SmolLM2-360M | 360M | 0.853 | 0.804 | 0.923 |
| Gargi-Base (Udgam Labs) | 384M | 0.916 | 0.882 | 0.966 |
| Udgam-384M-Base-4K (Udgam Labs) | 384M | 0.939 | 0.898 | 0.997 |
| SmolLM2-135M | 135M | 0.944 | 0.886 | 1.027 |
| Falcon-H1-Tiny-90M-Base | 90M | 0.981 | 0.955 | 1.017 |
Every one of the 10 pairwise gaps is significant in a paired bootstrap over documents (95% intervals of ±0.001 to ±0.003 bits per byte). The closest pair is Udgam-384M-Base-4K and SmolLM2-135M: 0.0049 apart, 95% interval [0.0025, 0.0072]. Falcon-H1-Tiny-90M-Base's tokenizer has no BOS token, so it starts without one. Exact model revisions are listed in the GitHub README.
Where it falls short
- Small and narrow. 1,670 documents in two genres (encyclopedia articles and scientific abstracts), English only. It says nothing about chat, following instructions, code or other languages.
- New wording, not always new facts. New Wikipedia articles often summarise older sources, and some arXiv abstracts appeared earlier with a conference or journal version. A model can know the facts without having seen this text.
- Some text may still be machine-written. By 21 October 2026 Wikipedia editors had deleted 11 of the 567 articles we first collected as presumed LLM-written text and one more for another reason, and had tagged 13 others as suspected AI-generated text. We removed all 25. For every model we tested, the 12 deleted articles were 0.12 to 0.14 bits per byte easier to predict than the articles we kept. More such articles may be found later.
- Freshness wears off. Once published, this text can end up in future training data. It is a fair test only for models whose training data stops before September 2026 (Wikipedia) or October 2026 (arXiv). Check a model's data cut-off before comparing.
- Settings matter. Context length, the BOS token and numeric precision all move the score. Compare models only under the same settings. Other hardware or bfloat16 may change the third decimal place.
- arXiv dates cluster at 5, 11 and 15 October, because the crawl took the newest abstracts in three date windows.
- Measured on small models only. We have scored five models from 90M to 384M parameters and have not checked how the set behaves for much larger ones.
How it was made
- Wikipedia. We listed every new page in the main article space between 1 September and 15 October 2026 (70,605 creation events). We dropped redirects, re-created pages, pages whose creation note says the text was copied, split, merged, translated or restored, lists, indexes, timelines and disambiguation pages, and kept only pages whose first revision falls in the window. For each page we took the plain-text version of its latest revision on 17 October 2026, kept prose paragraphs (at least 12 words, ending like a sentence, no leftover markup), dropped references and similar sections, and kept articles with at least 300 words of prose, cut at 1,500 words. Pages were sampled evenly across creation days.
- arXiv. The 10 newest abstracts in each of 75 categories (computer science, mathematics, statistics, physics, quantitative biology, economics, finance and electrical engineering) in each of three date windows, abstracts of 60 to 600 words, sampled evenly across categories.
- Screening. On 21 October 2026 we checked every Wikipedia page again and removed the 12 deleted since the crawl and the 13 that editors had tagged as suspected AI-generated text.
- Cleaning. Unicode normalised (NFC), control characters removed, exact duplicates removed. Each document is the title, a blank line, then the body.
The code is fresh_text/build.py;
build_summary.json in this repository records the counts and checksums.
Fields
| Field | Meaning |
|---|---|
id |
wiki:<page id> or arxiv:<arXiv id> |
source |
wikipedia or arxiv |
title |
article or paper title |
url |
permalink to the exact Wikipedia revision used, or the arXiv abstract page |
created |
UTC time the Wikipedia page was created (its first revision) or the arXiv paper was first submitted |
licence |
CC-BY-SA-4.0 or CC0-1.0 |
attribution |
a ready-to-use credit line for the row |
text |
title, blank line, body |
n_bytes |
UTF-8 bytes of text (the denominator of bits per byte) |
wiki_page_id, wiki_revision_id |
Wikipedia page and revision ids (empty for arXiv rows) |
arxiv_category |
arXiv primary category (empty for Wikipedia rows) |
Intended use
Evaluating pretrained (base) language models whose training data ends before September 2026; comparing architectures and tokenizers on unseen text; tracking a training run on held-out text. Training on it would remove its value as a test.
Refresh plan. We plan to build a new set each month with the same code, named fresh-text-YYYY-MM. This is a plan,
not a promise of a schedule.
Licence and attribution
- Wikipedia rows: CC BY-SA 4.0. The text is by Wikipedia contributors. Each row's
attributionnames the article, links the exact revision and its page history (the list of authors), and states the changes we made (plain text; references, headings, lists and short lines removed; long articles cut). Keep that credit and the share-alike licence when you reuse the text. - arXiv rows: CC0 1.0. Titles and abstracts are arXiv metadata, which arXiv releases under CC0. Each row still links the paper.
- The dataset as a whole is shared under CC BY-SA 4.0. The code that builds and scores it is Apache-2.0.
Citation
@misc{udgamlabs2026freshtext,
title = {Fresh Text, October 2026: new English text for contamination-free bits-per-byte evaluation},
author = {Prasad, Jitesh K and {Udgam Labs}},
year = {2026},
howpublished = {\url{https://huggingface.co/datasets/udgamlabs/fresh-text-2026-10}}
}
Please also credit the Wikipedia contributors and the arXiv authors, as in each row's attribution.
About Udgam Labs. Udgam Labs is an independent AI research lab. We work from first principles: before we build, we understand why things work, whether we are training a model, fine-tuning one or applying AI to a new problem. Udgam (उद्गम) is Sanskrit for origin. udgamlabs.com
- Downloads last month
- 92