davanstrien/ultrabert-380m-curation
Zero-Shot Classification • 0.4B • Updated • 71 • 10
id large_stringlengths 7 126 | config large_stringclasses 9
values | src large_stringclasses 195
values | row_type large_stringclasses 2
values | teacher large_stringclasses 18
values | licence_class large_stringclasses 4
values | licence large_stringclasses 16
values | state large_stringlengths 0 40.6k ⌀ | state_is_record bool 2
classes | question large_stringlengths 102 20.1k ⌀ | target large_stringlengths 10 1.27k ⌀ | text large_stringlengths 4 116k ⌀ | all_labels large listlengths 1 102 ⌀ | true_labels large_stringlengths 2 5.3k ⌀ | prompt large_stringlengths 0 2.21k ⌀ | multi_label bool 2
classes | source_dataset large_stringclasses 3
values | source_row int64 0 467k ⌀ |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
fh-arxiv-title-2609.15148 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Transfer learning is an effective technique for addressing data scarcity in deep learning for time series classification, but its success depends on the selection of source datasets. Conventional transferability estimation methods are often computationally expensive, as they require fully pre-training a model on each p... | [
"Video-STLayout Pre-training",
"Multi-source Transfer Learning of Time Series with a Shapelet-based Distance Measure",
"Design of a Deep Learning Credit Risk Early Warning System Integrating Multi-source Heterogeneous Data",
"UltraBERT: Beyond the Limit of BERT Pre-Training",
"AutoData: Agentic Search for P... | {"Video-STLayout Pre-training": 0.0, "Multi-source Transfer Learning of Time Series with a Shapelet-based Distance Measure": 1.0, "Design of a Deep Learning Credit Risk Early Warning System Integrating Multi-source Heterogeneous Data": 0.0, "UltraBERT: Beyond the Limit of BERT Pre-Training": 0.0, "AutoData: Agentic Sea... | Pick the paper title that matches this abstract. | false | null | null |
fh-arxiv-title-2609.28160 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Existing link-based combinatorial preprocessing methods speed up the computation of persistent homology by removing a vertex or edge only when its link remains a cone. We replace this condition with a quantitative homological certificate. The reduced homology of the filtered link of a generator (a vertex or edge) defin... | [
"Homological Trimming and Regularity of Filtrations via Local Obstruction Modules",
"The forced colouring function of a graph",
"Vertex-Coloring Edge-Weighting: Kernelization and Generalization",
"Empirical Auditing of Edge-Private Graph Generators",
"One Color Preprocessing Improves DSATUR"
] | {"Homological Trimming and Regularity of Filtrations via Local Obstruction Modules": 1.0, "The forced colouring function of a graph": 0.0, "Vertex-Coloring Edge-Weighting: Kernelization and Generalization": 0.0, "Empirical Auditing of Edge-Private Graph Generators": 0.0, "One Color Preprocessing Improves DSATUR": 0.0} | Which of these is the title of the paper described? | false | null | null |
fh-arxiv-title-2609.29856 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | This paper sets out a computational workflow that reconstructs the phylogeny of human Y-chromosome populations from a Variant Call Format (VCF) file of biallelic Single Nucleotide Polymorphisms (SNP). The two classical phylogenetic decisions - topology selection and root placement - are cast as Quadratic Unconstrained ... | [
"Best Matches in Phylogenetic Networks",
"Classical, quantum, and general probabilistic state-discrimination profiles",
"A Block Decomposed QUBO Workflow for Chromosome-Y Phylogeny Reconstruction",
"Fast and Robust Temporal Logic Planning via ADMM-based Trajectory Optimization",
"A rooted tree framework for... | {"Best Matches in Phylogenetic Networks": 0.0, "Classical, quantum, and general probabilistic state-discrimination profiles": 0.0, "A Block Decomposed QUBO Workflow for Chromosome-Y Phylogeny Reconstruction": 1.0, "Fast and Robust Temporal Logic Planning via ADMM-based Trajectory Optimization": 0.0, "A rooted tree fram... | Which title belongs to this abstract? | false | null | null |
fh-arxiv-cat-2609.18726 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Calmables: Demonstrating Closed-Loop Infrared Earables for Thermal Biofeedback and Relaxation Support
We present Calmables, a walk-up demo of a closed-loop infrared earable that uses smart-ring heart rate to create subtle, ear-localized warming cues. Building on work on thermal comfort and in-ear infrared stimulation,... | [
"Human-Computer Interaction",
"Computation and Language",
"Artificial Intelligence",
"Data Structures and Algorithms",
"Machine Learning",
"Computer Vision and Pattern Recognition"
] | {"Human-Computer Interaction": 1.0, "Computation and Language": 0.0, "Artificial Intelligence": 0.0, "Data Structures and Algorithms": 0.0, "Machine Learning": 0.0, "Computer Vision and Pattern Recognition": 0.0} | Which arXiv category is this paper's primary category? | false | null | null |
fh-arxiv-title-2609.28391 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Score matching has driven major advances in classical generative learning by enabling models to learn from data without evaluating intractable normalization constants, or partition functions. Yet, extending this principle to quantum learning requires rethinking its foundations, as quantum states are described by noncom... | [
"Weakly Supervised Quantum Error Mitigation",
"Quantum score matching with applications to learning thermal states",
"Learning Sparse Quantum States",
"Noise-Robust Quantum State Characterization for Remote State Preparation with Deep Learning",
"QEMScore: How Much Does the Measurement Add to Learned Quantu... | {"Weakly Supervised Quantum Error Mitigation": 0.0, "Quantum score matching with applications to learning thermal states": 1.0, "Learning Sparse Quantum States": 0.0, "Noise-Robust Quantum State Characterization for Remote State Preparation with Deep Learning": 0.0, "QEMScore: How Much Does the Measurement Add to Learn... | Which of these is the title of the paper described? | false | null | null |
fh-arxiv-cat-2609.13895 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Influence Enhancement in Opinion Dynamics Using Edge Modification: A Kron Reduction-Based Approach
With the emergence of online social networks as a primary platform for advertising and advocacy, enhancing a user's influence has become of significant interest. In this paper, we investigate this problem under the Fried... | [
"Software Engineering (cs.SE)",
"Sound (cs.SD)",
"Artificial Intelligence (cs.AI)",
"Optimization and Control (math.OC)",
"Human-Computer Interaction (cs.HC)",
"Systems and Control (EESS) (eess.SY)"
] | {"Software Engineering (cs.SE)": 0.0, "Sound (cs.SD)": 0.0, "Artificial Intelligence (cs.AI)": 0.0, "Optimization and Control (math.OC)": 0.0, "Human-Computer Interaction (cs.HC)": 0.0, "Systems and Control (EESS) (eess.SY)": 1.0} | What is the primary subject area of this paper? | false | null | null |
fh-arxiv-cat-2609.12231 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Learning to adapt GR(1) specifications through degradation
Reactive synthesis is a powerful tool for generating correct-by-construction controllers from formal specifications. GR(1) is an assume-guarantee specification framework that enables efficient synthesis, allowing synthesised controllers to be used in a wide ar... | [
"Cryptography and Security (cs.CR)",
"Computation and Language (cs.CL)",
"Robotics (cs.RO)",
"Audio and Speech Processing (eess.AS)",
"Optimization and Control (math.OC)",
"Software Engineering (cs.SE)"
] | {"Cryptography and Security (cs.CR)": 0.0, "Computation and Language (cs.CL)": 0.0, "Robotics (cs.RO)": 0.0, "Audio and Speech Processing (eess.AS)": 0.0, "Optimization and Control (math.OC)": 0.0, "Software Engineering (cs.SE)": 1.0} | Classify this paper into its main arXiv category. | false | null | null |
fh-arxiv-cat-2609.25364 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Quantum ROP: Using Quantum Algorithms for ROP Chain Selection in Exploit Construction
The quantum computing threat to cybersecurity is nowadays predominantly framed around Shor's algorithm and its eventual capacity to break asymmetric cryptography. Beyond cryptanalysis, however, quantum computing may also enable other... | [
"cs.CR",
"cs.LG",
"cs.AI",
"cs.CV",
"cs.IT",
"cs.NI"
] | {"cs.CR": 1.0, "cs.LG": 0.0, "cs.AI": 0.0, "cs.CV": 0.0, "cs.IT": 0.0, "cs.NI": 0.0} | Classify this paper into its main arXiv category. | false | null | null |
fh-arxiv-cat-2609.27238 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Strip Convolution and Direction-Aware Exclusion Loss for Oriented Ship Detection
Oriented ship detection in very high resolution (VHR) remote sensing imagery remains challenging due to elongated hull geometry and dense target distributions in complex port scenes. Existing methods typically address geometric representa... | [
"Robotics",
"Computer Vision and Pattern Recognition",
"Artificial Intelligence",
"Information Theory",
"Cryptography and Security",
"Sound"
] | {"Robotics": 0.0, "Computer Vision and Pattern Recognition": 1.0, "Artificial Intelligence": 0.0, "Information Theory": 0.0, "Cryptography and Security": 0.0, "Sound": 0.0} | Route this paper to the arXiv listing it belongs to. | false | null | null |
fh-arxiv-cat-2609.15643 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Principal-timestep Restricted Init via Sparse Matrix-decomposition in Flow-matching
Flow-matching diffusion models have recently emerged as a strong paradigm for high-fidelity visual generation. However, their prohibitively high fine-tuning cost limits scalability to downstream tasks. While Low-Rank Adaptation (LoRA) ... | [
"Robotics (cs.RO)",
"Human-Computer Interaction (cs.HC)",
"Audio and Speech Processing (eess.AS)",
"Machine Learning (cs.LG)",
"Artificial Intelligence (cs.AI)",
"Computer Vision and Pattern Recognition (cs.CV)"
] | {"Robotics (cs.RO)": 0.0, "Human-Computer Interaction (cs.HC)": 0.0, "Audio and Speech Processing (eess.AS)": 0.0, "Machine Learning (cs.LG)": 1.0, "Artificial Intelligence (cs.AI)": 0.0, "Computer Vision and Pattern Recognition (cs.CV)": 0.0} | Which arXiv category is this paper's primary category? | false | null | null |
fh-arxiv-cat-2609.13523 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Positive Topology and Feasible Refinement: Forcing Matrices, Positivity, and Information
We develop a conceptual and operational account of Positive Topology starting from a basic relation between points or models and observable properties. From this relation, two complementary structures emerge. The first captures un... | [
"Human-Computer Interaction",
"Networking and Internet Architecture",
"Logic in Computer Science",
"Computation and Language",
"Multiagent Systems",
"Sound"
] | {"Human-Computer Interaction": 0.0, "Networking and Internet Architecture": 0.0, "Logic in Computer Science": 1.0, "Computation and Language": 0.0, "Multiagent Systems": 0.0, "Sound": 0.0} | Classify this paper into its main arXiv category. | false | null | null |
fh-arxiv-title-2610.02331 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Interactive world models are increasingly capable of generating environments and acting within them, yet deliberately editing an existing executable world remains underexplored. We formulate world editing as intervening on an existing world while preserving properties that should remain unchanged, and introduce interve... | [
"RAR: Rate-Adaptive Rollout for Time-Consistent Interactive World Models",
"Code2Games: Enabling Coding Agents for Gaming World Generation",
"WorldPlay2: Extending Real-Time Interactive World Models in Control and Horizon",
"World Editing: Intervening on Executable Worlds at Increasing Depth",
"Astronex-Wor... | {"RAR: Rate-Adaptive Rollout for Time-Consistent Interactive World Models": 0.0, "Code2Games: Enabling Coding Agents for Gaming World Generation": 0.0, "WorldPlay2: Extending Real-Time Interactive World Models in Control and Horizon": 0.0, "World Editing: Intervening on Executable Worlds at Increasing Depth": 1.0, "Ast... | Which of these is the title of the paper described? | false | null | null |
fh-arxiv-title-2610.07609 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Neural simulation of high-fidelity deformable bodies is a foundational challenge in computer graphics and physical AI. Long-horizon prediction for high-resolution 3D volumetric meshes is hard: autoregressive methods are prone to error accumulation, while direct multi-frame prediction at native resolution is computation... | [
"PhysLDM: Latent Diffusion for High-Fidelity Deformable Simulation",
"ZoomDiff: A High-Fidelity Diffusion Model for Dual-Camera Smooth Zooming",
"Uni-LaDiR: Latent Diffusion Unifies Multimodal Reasoning",
"RoAR-VAE: Improving Latent Diffusability through Robust Prediction",
"MedDiME: Efficient Latent Diffus... | {"PhysLDM: Latent Diffusion for High-Fidelity Deformable Simulation": 1.0, "ZoomDiff: A High-Fidelity Diffusion Model for Dual-Camera Smooth Zooming": 0.0, "Uni-LaDiR: Latent Diffusion Unifies Multimodal Reasoning": 0.0, "RoAR-VAE: Improving Latent Diffusability through Robust Prediction": 0.0, "MedDiME: Efficient Late... | Which of these is the title of the paper described? | false | null | null |
fh-arxiv-title-2609.24444 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Accurate, reliable, and deployable wind power forecasting is critical for power system dispatch, renewable energy integration, and electricity market operations. Progress in this field hinges on the ability to empirically and comprehensively benchmark forecasting methods. Yet existing benchmarks fall short of supportin... | [
"Horizon-specific Expert Fusion for Photovoltaic Power Forecasting",
"Informed Masking: Structure-Aware Perturbation for Reinforcement Learning in Diffusion Large Language Models",
"WPBench: A Comprehensive Benchmark for Wind Power Forecasting",
"Offline Reinforcement Learning for Wind Farm Control: A Wind Tu... | {"Horizon-specific Expert Fusion for Photovoltaic Power Forecasting": 0.0, "Informed Masking: Structure-Aware Perturbation for Reinforcement Learning in Diffusion Large Language Models": 0.0, "WPBench: A Comprehensive Benchmark for Wind Power Forecasting": 1.0, "Offline Reinforcement Learning for Wind Farm Control: A W... | Which title belongs to this abstract? | false | null | null |
fh-arxiv-title-2610.28435 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Multimodal rerankers are usually dedicated models: a vision-language model (VLM) is trained on retrieval data to score the candidates of a query. Scoring options is not specific to reranking. System One decision models such as Jev map a state and a typed question to one probability per option without generating text, a... | [
"Multimodal Reranking Is a System One Decision",
"Replacing Large Language Models with Jev Decision Models for Low-Latency Edge Service Orchestration",
"Decision-Oriented Recommendation Reranking: An Empirical Study of Jev",
"Adaptive Cortically Constrained EEG-Vision Alignment for Zero-Shot Brain-to-Image Re... | {"Multimodal Reranking Is a System One Decision": 1.0, "Replacing Large Language Models with Jev Decision Models for Low-Latency Edge Service Orchestration": 0.0, "Decision-Oriented Recommendation Reranking: An Empirical Study of Jev": 0.0, "Adaptive Cortically Constrained EEG-Vision Alignment for Zero-Shot Brain-to-Im... | Pick the paper title that matches this abstract. | false | null | null |
fh-arxiv-cat-2609.29268 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | BridgeMem: Causal Dyadic Transition Residuals for Temporal Knowledge Graph Forecasting
Temporal knowledge graph forecasting aims to infer future relational facts from the temporal structure of observed events. Existing forecasters mainly summarize history through entity states, relation states, paths, or exact recurre... | [
"Machine Learning (cs.LG)",
"Cryptography and Security (cs.CR)",
"Information Retrieval (cs.IR)",
"Networking and Internet Architecture (cs.NI)",
"Artificial Intelligence (cs.AI)",
"Information Theory (cs.IT)"
] | {"Machine Learning (cs.LG)": 1.0, "Cryptography and Security (cs.CR)": 0.0, "Information Retrieval (cs.IR)": 0.0, "Networking and Internet Architecture (cs.NI)": 0.0, "Artificial Intelligence (cs.AI)": 0.0, "Information Theory (cs.IT)": 0.0} | What is the primary subject area of this paper? | false | null | null |
fh-arxiv-title-2609.23604 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Influence maximization asks for $k$ seed vertices that maximize the expected spread of a diffusion process in a network. Standard near-optimal-time algorithms based on reverse-reachable sampling achieve a $(1-1/e-\varepsilon)$ approximation, but their expected running-time bounds grow linearly with the seed budget $k$.... | [
"Budget-Independent Influence Maximization in Nearly Linear Time",
"A Strongly Subquadratic $(3+\\varepsilon)$-Approximation for Weighted Edit Distance over Arbitrary Metrics",
"Expected Hypervolume Maximization for Multiobjective Optimization under Uncertainties",
"Exact Greedy Influence Maximization in Line... | {"Budget-Independent Influence Maximization in Nearly Linear Time": 1.0, "A Strongly Subquadratic $(3+\\varepsilon)$-Approximation for Weighted Edit Distance over Arbitrary Metrics": 0.0, "Expected Hypervolume Maximization for Multiobjective Optimization under Uncertainties": 0.0, "Exact Greedy Influence Maximization i... | Which title belongs to this abstract? | false | null | null |
fh-arxiv-cat-2609.27489 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Passing: An Endless Journey through Reconstructed Spacetime with AI-Generated Sound
This paper introduces Passing, an interactive audiovisual installation that generates an endless journey from a single continuous monorail-window recording by reconstructing it as a spatiotemporal volume. Rather than replaying the foot... | [
"Computational Complexity (cs.CC)",
"Sound (cs.SD)",
"Information Retrieval (cs.IR)",
"Cryptography and Security (cs.CR)",
"Audio and Speech Processing (eess.AS)",
"Robotics (cs.RO)"
] | {"Computational Complexity (cs.CC)": 0.0, "Sound (cs.SD)": 1.0, "Information Retrieval (cs.IR)": 0.0, "Cryptography and Security (cs.CR)": 0.0, "Audio and Speech Processing (eess.AS)": 0.0, "Robotics (cs.RO)": 0.0} | Which arXiv category is this paper's primary category? | false | null | null |
fh-arxiv-title-2609.12035 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Cardiovascular diagnosis and treatment rest on integrating complementary modalities, such as electrocardiogram, echocardiography, and chest X-rays, each capturing distinct but complementary aspects of cardiac pathophysiology. Yet most medical foundation models remain modality-specific, combining modalities only for fin... | [
"MiTHras: Task-specific Hierarchical Semi-supervised Contrastive Masked Autoencoder for Mitotic Figure Analysis",
"Reading the Whole Heart: Latent-Attention Masked Autoencoders for Multimodal Cardiac Representation Learning",
"EMGBlend: Heterogeneity-Aware Self-Supervised Pretraining for Gesture and Force Decod... | {"MiTHras: Task-specific Hierarchical Semi-supervised Contrastive Masked Autoencoder for Mitotic Figure Analysis": 0.0, "Reading the Whole Heart: Latent-Attention Masked Autoencoders for Multimodal Cardiac Representation Learning": 1.0, "EMGBlend: Heterogeneity-Aware Self-Supervised Pretraining for Gesture and Force De... | Pick the paper title that matches this abstract. | false | null | null |
fh-arxiv-title-2609.13353 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Agent skills are reusable units for language-model agents, but their risks emerge through model decisions, user context, tool calls, and execution feedback rather than through stable signatures or a single sandbox run. Existing static, dynamic, and benchmark-style evaluations rarely preserve public evidence that can be... | [
"Beyond Maliciousness: Static and Runtime Attack-Surface Profiling of Agent Skills",
"SkillDRE: Dual-Stage Red-Team Evolution of Agent Skills via Pre-Execution and Runtime Feedback",
"SkillAtlas: An Attack Trace Library for Agent Skills",
"Skill Constellations: Tracing the Supply Chain of Agent Skills on GitH... | {"Beyond Maliciousness: Static and Runtime Attack-Surface Profiling of Agent Skills": 0.0, "SkillDRE: Dual-Stage Red-Team Evolution of Agent Skills via Pre-Execution and Runtime Feedback": 0.0, "SkillAtlas: An Attack Trace Library for Agent Skills": 1.0, "Skill Constellations: Tracing the Supply Chain of Agent Skills o... | Pick the paper title that matches this abstract. | false | null | null |
fh-arxiv-title-2609.12489 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Test case synthesis is crucial for evaluating and ranking programs generated by large language models (LLMs). However, constructing high-quality test cases remains challenging because reliable expected outputs are often difficult to obtain. We propose Confidence-Gated Transductive Test Generation (CoTT), which first us... | [
"Even Sharper Bounds for Transductive Learning and Its Applications",
"Enhancing Automated Unit Test Generation for NLP Libraries Using Large Language Models",
"RobotUse: Allocating Computation, Context, and Decisions",
"NEUROTESTGEN: Neuro-Symbolic Guided Test Generation with Large Language Models",
"Confi... | {"Even Sharper Bounds for Transductive Learning and Its Applications": 0.0, "Enhancing Automated Unit Test Generation for NLP Libraries Using Large Language Models": 0.0, "RobotUse: Allocating Computation, Context, and Decisions": 0.0, "NEUROTESTGEN: Neuro-Symbolic Guided Test Generation with Large Language Models": 0.... | Which of these is the title of the paper described? | false | null | null |
fh-arxiv-title-2609.21729 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | As robots transition from performing repetitive tasks to collaborating with humans, understanding human intent becomes crucial to effective interaction. Anticipation enables robots to predict human actions, while proactivity allows them to take initiative and guide human behavior toward optimal outcomes. Although resea... | [
"Robots That Take Initiative: A Framework for Building and Evaluating Proactive Robots",
"Learning from Humans for Proactive Assistance in Human-Robot Collaborative Transport",
"Visual Proactivity: Enhancing Human-Robot Collaboration Through Intent Communication",
"Toward Human-in-the-Loop Robot Failure Recov... | {"Robots That Take Initiative: A Framework for Building and Evaluating Proactive Robots": 0.0, "Learning from Humans for Proactive Assistance in Human-Robot Collaborative Transport": 0.0, "Visual Proactivity: Enhancing Human-Robot Collaboration Through Intent Communication": 1.0, "Toward Human-in-the-Loop Robot Failure... | Pick the paper title that matches this abstract. | false | null | null |
fh-arxiv-cat-2609.23547 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Preserving Geometric Integrity in Graph Prompting via Measure-Constrained Optimal Transport
Graph prompt learning enables parameter-efficient adaptation of frozen Graph Neural Networks to downstream tasks through lightweight prompt parameters. As routing becomes increasingly node-adaptive, however, independently optim... | [
"Software Engineering",
"Sound",
"Machine Learning (Statistics)",
"Computation and Language",
"Machine Learning",
"Artificial Intelligence"
] | {"Software Engineering": 0.0, "Sound": 0.0, "Machine Learning (Statistics)": 0.0, "Computation and Language": 0.0, "Machine Learning": 1.0, "Artificial Intelligence": 0.0} | Which arXiv category is this paper's primary category? | false | null | null |
fh-arxiv-title-2609.26905 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Stacking heterogeneous vision backbones (CNNs, ViTs, and hybrids) is the de facto recipe for accuracy, calibration, and robustness, yet two coupled pathologies limit its returns. Prediction-space multicollinearity ill-conditions the meta-learner's Gram matrix, inflating weight variance and producing brittle solutions o... | [
"PAC-Bayesian Meta-Learning for Few-Shot Identification of Linear Dynamical Systems",
"Linear Ensemble Sampling with Smaller Ensembles",
"RPyForth: Exposing a Call-Shared Data Stack to a Meta-Tracing JIT Compiler",
"Learning to Move Cities: Deep Meta-Models and Reinforcement Policies for Calibration and Contr... | {"PAC-Bayesian Meta-Learning for Few-Shot Identification of Linear Dynamical Systems": 0.0, "Linear Ensemble Sampling with Smaller Ensembles": 0.0, "RPyForth: Exposing a Call-Shared Data Stack to a Meta-Tracing JIT Compiler": 0.0, "Learning to Move Cities: Deep Meta-Models and Reinforcement Policies for Calibration and... | Which of these is the title of the paper described? | false | null | null |
fh-arxiv-title-2609.20026 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Urban traffic forecasting often relies on information distributed across stakeholders who may be unable to share raw data due to privacy or commercial constraints, motivating federated spatial-temporal approaches. In such federated settings, each client observes traffic over a distinct sensor subgraph with its own spat... | [
"Learning Local Heterogeneity and Cross-Region Context for Large-Scale Traffic Forecasting",
"A Multi-Resolution Multi-Domain Pre-Training Framework for Universal Traffic Forecasting",
"Disentangling Heterogeneous Traffic Dynamics for Multi-Step Traffic Forecasting via Adaptive Spectral Decomposition",
"FedeR... | {"Learning Local Heterogeneity and Cross-Region Context for Large-Scale Traffic Forecasting": 0.0, "A Multi-Resolution Multi-Domain Pre-Training Framework for Universal Traffic Forecasting": 0.0, "Disentangling Heterogeneous Traffic Dynamics for Multi-Step Traffic Forecasting via Adaptive Spectral Decomposition": 0.0, ... | Which of these is the title of the paper described? | false | null | null |
fh-arxiv-title-2609.12816 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Multistable systems admit distinct stable equilibria for the same loading and boundary conditions. Shape optimization of multistable structures offers a promising route to engineer the mechanical response of advanced materials and metamaterials. In this work, we develop a formulation and algorithm for controlling the e... | [
"The shape of quark flavors",
"When Connected Does Not Mean Similar: Charting the Homophily Boundary of SNAP-KG for Streaming Entity Integration",
"Embodied Snap: Octopus-Inspired Distributed Reach-and-Attach with a Speed-Limited Soft Arm",
"Efficient Geometry Representation Strategies for the Shape Optimizat... | {"The shape of quark flavors": 0.0, "When Connected Does Not Mean Similar: Charting the Homophily Boundary of SNAP-KG for Streaming Entity Integration": 0.0, "Embodied Snap: Octopus-Inspired Distributed Reach-and-Attach with a Speed-Limited Soft Arm": 0.0, "Efficient Geometry Representation Strategies for the Shape Opt... | Which of these is the title of the paper described? | false | null | null |
fh-arxiv-title-2609.17775 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Enterprise guideline documents mix narrative text, complex tables, and embedded images, and converting them into structured work artifacts still takes two to three days of manual effort each. Current language and vision-language models extract from such documents but offer no governed workflow beyond extraction: no val... | [
"SAGE: Governed Artifact Generation from Enterprise Guidelines",
"Spectra: A Rules-Driven LLM Pipeline for Automated KYC Document Processing",
"From the Task Boundaries of Narrative Text to Structural Anchoring, Uncertainty Triggers, and Cross-Calibration",
"A Multi-Stage Agentic Framework for Effective Count... | {"SAGE: Governed Artifact Generation from Enterprise Guidelines": 1.0, "Spectra: A Rules-Driven LLM Pipeline for Automated KYC Document Processing": 0.0, "From the Task Boundaries of Narrative Text to Structural Anchoring, Uncertainty Triggers, and Cross-Calibration": 0.0, "A Multi-Stage Agentic Framework for Effective... | Which title belongs to this abstract? | false | null | null |
fh-arxiv-title-2609.22038 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | We introduce QuranicMMLU, a benchmark for evaluating generative AI on Quranic Arabic across multiple dimensions of linguistic complexity. Existing Quranic benchmarks center on general question answering and semantic retrieval, without probing specific linguistic competencies or stratifying by cognitive demand and verse... | [
"An Evolutionary Agentic Approach for Open-ended Image Quality Perception",
"The syntax and semantics of goals",
"QuranicMMLU: A Cognitively-Aware Benchmark for Evaluating Generative AI Solutions on Quranic Linguistic Knowledge",
"A Corpus-Aligned Uthmani-to-Standard Quranic Word Mapping and a Deterministic R... | {"An Evolutionary Agentic Approach for Open-ended Image Quality Perception": 0.0, "The syntax and semantics of goals": 0.0, "QuranicMMLU: A Cognitively-Aware Benchmark for Evaluating Generative AI Solutions on Quranic Linguistic Knowledge": 1.0, "A Corpus-Aligned Uthmani-to-Standard Quranic Word Mapping and a Determini... | Which title belongs to this abstract? | false | null | null |
fh-arxiv-cat-2609.26624 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | On Basis Function Selection for Sparse Gaussian Process Regression
Sparse Gaussian processes achieve $O(N)$ inference by replacing the kernel with an appropriate expansion in a fixed basis $\{φ_j\}$ on the input space. Given a compute budget $M \ll N$, practitioners conventionally truncate the basis to its first $M$ e... | [
"Multiagent Systems",
"Logic in Computer Science",
"Robotics",
"Human-Computer Interaction",
"Artificial Intelligence",
"Machine Learning (Statistics)"
] | {"Multiagent Systems": 0.0, "Logic in Computer Science": 0.0, "Robotics": 0.0, "Human-Computer Interaction": 0.0, "Artificial Intelligence": 0.0, "Machine Learning (Statistics)": 1.0} | Classify this paper into its main arXiv category. | false | null | null |
fh-arxiv-title-2609.23442 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Diffusion models generate high-quality images, yet often violate the physical laws governing mirror reflections. Reflections often suffer from geometric aberrations, including positional offsets, directional misalignment, proportional imbalance, and structural distortion. These failures remain evident even in contempor... | [
"Iris-3B: Going Beyond the Latent with Pixel-Space Diffusion Training, Conversion and Fine-Tuning",
"PhysReflect: Geometry and Perception Guided Diffusion for Physically-Plausible Mirror Reflections",
"Mirror, Mirror on the Wall: Prompt Echoing in Small Instruct Language Models",
"Boundary transmission and ex... | {"Iris-3B: Going Beyond the Latent with Pixel-Space Diffusion Training, Conversion and Fine-Tuning": 0.0, "PhysReflect: Geometry and Perception Guided Diffusion for Physically-Plausible Mirror Reflections": 1.0, "Mirror, Mirror on the Wall: Prompt Echoing in Small Instruct Language Models": 0.0, "Boundary transmission ... | Which of these is the title of the paper described? | false | null | null |
fh-arxiv-cat-2609.15044 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Data Attribution at Scale via Influence Matrix Estimation
Data attribution seeks to quantify how individual training examples shape a model's predictions and underpins problems including data valuation, machine unlearning, and model interpretability. Despite having a long line of work, computationally scalable methods... | [
"Audio and Speech Processing (eess.AS)",
"Computer Vision and Pattern Recognition (cs.CV)",
"Artificial Intelligence (cs.AI)",
"Machine Learning (Statistics) (stat.ML)",
"Cryptography and Security (cs.CR)",
"Software Engineering (cs.SE)"
] | {"Audio and Speech Processing (eess.AS)": 0.0, "Computer Vision and Pattern Recognition (cs.CV)": 0.0, "Artificial Intelligence (cs.AI)": 0.0, "Machine Learning (Statistics) (stat.ML)": 1.0, "Cryptography and Security (cs.CR)": 0.0, "Software Engineering (cs.SE)": 0.0} | What is the primary subject area of this paper? | false | null | null |
fh-arxiv-title-2609.27014 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Legal contracts are structurally complex documents in which contradictions may emerge across distant and interconnected provisions. Although large language models (LLMs) improve legal language understanding, contradiction analysis remains a human-centered and evidence-grounded review task. We present ContraVis, a visua... | [
"PaperDoctor: Evidence-Grounded and Actionable Feedback for Scientific Papers in Progress",
"ContraVis: Evidence-Grounded Visual Analytics for Contradiction Review in Legal Contracts",
"CALICO: A Human-Centered, Codebook-Aligned System for Annotation",
"EvaOmni: An Evidence-Grounded Benchmark for Audio-Visual... | {"PaperDoctor: Evidence-Grounded and Actionable Feedback for Scientific Papers in Progress": 0.0, "ContraVis: Evidence-Grounded Visual Analytics for Contradiction Review in Legal Contracts": 1.0, "CALICO: A Human-Centered, Codebook-Aligned System for Annotation": 0.0, "EvaOmni: An Evidence-Grounded Benchmark for Audio-... | Which title belongs to this abstract? | false | null | null |
fh-arxiv-cat-2609.28736 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Towards a Platform for Mastering Personal Sovereignty
The ongoing digital transformation of work, administration, health, mobility, and social interaction is profoundly reshaping everyday life, steadily shifting control from individuals to large platform providers. Although data is often labeled the "gold of the 21st ... | [
"Computers and Society",
"Artificial Intelligence",
"Machine Learning (Statistics)",
"Distributed, Parallel, and Cluster Computing",
"Cryptography and Security",
"Computer Science and Game Theory"
] | {"Computers and Society": 1.0, "Artificial Intelligence": 0.0, "Machine Learning (Statistics)": 0.0, "Distributed, Parallel, and Cluster Computing": 0.0, "Cryptography and Security": 0.0, "Computer Science and Game Theory": 0.0} | Which arXiv category is this paper's primary category? | false | null | null |
fh-arxiv-title-2609.26445 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | This manuscript formalizes the most popular model validation tools used in general insurance actuarial modeling. These include graphical tools like calibration plots, actual-vs-expected plots, lift charts, Murphy diagrams, as well as classical statistical tools such as Bregman losses, deviance losses, elementary losses... | [
"ggaction: A Grammar of Graphical Actions",
"A Practical Guide on Graphical Model Validation",
"String Diagrams for Process Mining",
"LWCal: Loss-Weighted Calibration for Tabular Classifiers with Noisy Calibration Labels",
"Restructuring Tree Decision Diagrams"
] | {"ggaction: A Grammar of Graphical Actions": 0.0, "A Practical Guide on Graphical Model Validation": 1.0, "String Diagrams for Process Mining": 0.0, "LWCal: Loss-Weighted Calibration for Tabular Classifiers with Noisy Calibration Labels": 0.0, "Restructuring Tree Decision Diagrams": 0.0} | Which title belongs to this abstract? | false | null | null |
fh-arxiv-cat-2609.25186 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | From Pattern Recognizers to Personalized Companions: A Survey of Large Language Models in Mental Health
The rising global prevalence of mental health conditions, together with longstanding barriers in traditional healthcare, such as limited resources, high cost, stigma, and privacy concerns, has created an urgent need... | [
"cs.CV",
"stat.ML",
"cs.CY",
"cs.IR",
"cs.HC",
"cs.SD"
] | {"cs.CV": 0.0, "stat.ML": 0.0, "cs.CY": 1.0, "cs.IR": 0.0, "cs.HC": 0.0, "cs.SD": 0.0} | Classify this paper into its main arXiv category. | false | null | null |
fh-arxiv-title-2609.18124 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Visible-infrared object detection (VIOD) detects objects with oriented bounding boxes from paired visible and infrared images. Existing methods depend on costly dual-modality annotations. Semi-supervised learning can reduce this burden, but extending it from single-modal detection to VIOD is challenging. In the practic... | [
"Efficient Quantization-Aware Distillation with Cross-Modal Alignment for Edge Vision-Language Models",
"A Practical Recipe for Semi-Supervised Federated ASR: Online Pseudo-Labels with Server Update Stabilization",
"Aligned Consensus Teaching for Label-Efficient Oriented Object Detection in Weakly-Aligned Visib... | {"Efficient Quantization-Aware Distillation with Cross-Modal Alignment for Edge Vision-Language Models": 0.0, "A Practical Recipe for Semi-Supervised Federated ASR: Online Pseudo-Labels with Server Update Stabilization": 0.0, "Aligned Consensus Teaching for Label-Efficient Oriented Object Detection in Weakly-Aligned Vi... | Pick the paper title that matches this abstract. | false | null | null |
fh-arxiv-title-2609.27294 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Scaling a language model is not only a question of final quality: the architectural choice determines how much computation is spent during training, prompt processing, and autoregressive decoding to achieve certain model quality. An ideal model architecture should lower all above computation costs to facilitate scaling... | [
"Enterprise Process Control Tower for IT-Led Improvement",
"KITE: Scaling Jev Population Experiments with Sparse Flagship Calibration",
"MeshKV: A Network-on-Chip KV Cache Fabric for Scalable Transformer Decoding Accelerators",
"KITE: KV-Invariant Transformer Expansion for Efficient Agentic LLM Scaling",
"A... | {"Enterprise Process Control Tower for IT-Led Improvement": 0.0, "KITE: Scaling Jev Population Experiments with Sparse Flagship Calibration": 0.0, "MeshKV: A Network-on-Chip KV Cache Fabric for Scalable Transformer Decoding Accelerators": 0.0, "KITE: KV-Invariant Transformer Expansion for Efficient Agentic LLM Scaling"... | Pick the paper title that matches this abstract. | false | null | null |
fh-arxiv-cat-2609.21511 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | 2nd Place Solution to the HANDS 2026 Workshop Challenge-Dexterous Grasp Motion Track: Single-Shot Trajectory Warping for Grasp Motion Generation
This report describes our 2nd place solution to the HANDS 2026 workshop challenge (Dexterous Grasp Motion track) in conjunction with ECCV 2026. In this challenge, we address ... | [
"Machine Learning",
"Robotics",
"Cryptography and Security",
"Hardware Architecture",
"Audio and Speech Processing",
"Computers and Society"
] | {"Machine Learning": 0.0, "Robotics": 1.0, "Cryptography and Security": 0.0, "Hardware Architecture": 0.0, "Audio and Speech Processing": 0.0, "Computers and Society": 0.0} | Classify this paper into its main arXiv category. | false | null | null |
fh-arxiv-title-2609.22781 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | We describe the integration of io_uring into Oracle Database's storage layer and the architectural decisions required to deploy it in a production multi-process RDBMS. Our design uses per-process ring contexts that eliminate inter-process synchronization, a shared buffer registration mechanism now part of the mainline ... | [
"io_uring in Oracle Database: A Hybrid Storage I/O Architecture at Production Scale",
"Hybrid GPU-CPU Retrieval for Personalized Search at Ultra-Large Scale",
"ASPIRE: Asynchronous Batched Self-Speculative Decoding for Long-Context LLM Inference",
"Invisible Yet Dominant: Big Stalls of Kernel I/O Mechanisms i... | {"io_uring in Oracle Database: A Hybrid Storage I/O Architecture at Production Scale": 1.0, "Hybrid GPU-CPU Retrieval for Personalized Search at Ultra-Large Scale": 0.0, "ASPIRE: Asynchronous Batched Self-Speculative Decoding for Long-Context LLM Inference": 0.0, "Invisible Yet Dominant: Big Stalls of Kernel I/O Mechan... | Which title belongs to this abstract? | false | null | null |
fh-arxiv-title-2609.19528 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | We consider the question of designing a universal family of sets $F \subset 2^{[n]}$ such that for any function $f:2^{[n]} \to R_{\geq 0}$ in a certain class, we have $$\max_{S \in F} f(S) \geq c(n) \cdot \max_{S \subset [n]} f(S).$$ We prove that there is a family of subpolynomial size such that for any nonnegative su... | [
"Submodular Maximization over Bipartite Perfect Matchings and Matroid Intersection Bases",
"Universal set families for maximization of nonnegative submodular and XOS functions",
"Strong and Compact Policies for Submodular Markov Decision Processes via LP-Based Submodular Orienteering",
"Breaking the 1/3 Barri... | {"Submodular Maximization over Bipartite Perfect Matchings and Matroid Intersection Bases": 0.0, "Universal set families for maximization of nonnegative submodular and XOS functions": 1.0, "Strong and Compact Policies for Submodular Markov Decision Processes via LP-Based Submodular Orienteering": 0.0, "Breaking the 1/3... | Which of these is the title of the paper described? | false | null | null |
fh-arxiv-title-2609.12788 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Aeroelastic ?utter is normally a failure mode to be designed against, yet the limit-cycle oscillations (LCOs) that follow it convert flow energy into sustained structural motion that a piezoelectric transducer can turn into electrical power. A transducer is embedded in a three-degree-of-freedom pitch-plunge aerofoil wi... | [
"Floquet Fibre Geometry and Higher-Order Reduced Coordinates for Off-Manifold Transients near Nonlinear Aeroelastic Flutter",
"Ladder of information limits on prediction for reduced-order models",
"Transducer Placement and the Limits of a Four-State Reduced Model in Post-Flutter Piezoelectric Energy Harvesting ... | {"Floquet Fibre Geometry and Higher-Order Reduced Coordinates for Off-Manifold Transients near Nonlinear Aeroelastic Flutter": 0.0, "Ladder of information limits on prediction for reduced-order models": 0.0, "Transducer Placement and the Limits of a Four-State Reduced Model in Post-Flutter Piezoelectric Energy Harvesti... | Which title belongs to this abstract? | false | null | null |
fh-arxiv-cat-2609.20480 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Worst-Case Hidden-Vehicle Trajectory Search in Spatiotemporal Occlusion Regions
Occlusion creates fundamental uncertainty in autonomous driving. Existing methods often propagate frame-wise hypotheses or optimize ego behavior against prescribed hidden-agent predictions, leaving the worst history-consistent interaction ... | [
"Computer Vision and Pattern Recognition (cs.CV)",
"Robotics (cs.RO)",
"Artificial Intelligence (cs.AI)",
"Data Structures and Algorithms (cs.DS)",
"Systems and Control (EESS) (eess.SY)",
"Computation and Language (cs.CL)"
] | {"Computer Vision and Pattern Recognition (cs.CV)": 0.0, "Robotics (cs.RO)": 1.0, "Artificial Intelligence (cs.AI)": 0.0, "Data Structures and Algorithms (cs.DS)": 0.0, "Systems and Control (EESS) (eess.SY)": 0.0, "Computation and Language (cs.CL)": 0.0} | Classify this paper into its main arXiv category. | false | null | null |
fh-arxiv-title-2609.34677 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | World models predict future observations from current experience and actions, yet prediction can depend on observations seen far in the past. Episodic memory preserves past observations for later recall; however, as memory accumulates, it raises a fundamental question: which memories are useful for the current predicti... | [
"The Past Frames the Future: Memory for Autoregressive Video Generation",
"CapMem: A Benchmark for Caption-Based Episodic Memory in Egocentric Video",
"World Models with Configurable Perception",
"Agent Memory with Episodic Retrieval for Financial Decision-Making",
"Learning What to Recall: Adaptive Multi-C... | {"The Past Frames the Future: Memory for Autoregressive Video Generation": 0.0, "CapMem: A Benchmark for Caption-Based Episodic Memory in Egocentric Video": 0.0, "World Models with Configurable Perception": 0.0, "Agent Memory with Episodic Retrieval for Financial Decision-Making": 0.0, "Learning What to Recall: Adaptiv... | Pick the paper title that matches this abstract. | false | null | null |
fh-arxiv-title-2609.25892 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | In recent years, the threat of application-layer (L7) distributed denial-of-service (DDoS) attacks is ever increasing. To defend against them, network operators deploy web application firewalls (WAFs). WAFs are stateful scoring systems which are configured with a rule set that specifies what malicious traffic looks lik... | [
"Rethinking Web Application Firewalls",
"Human Driver Temperament and the Safety Impact of a C-V2X Denial-of-Service Flooding Attack in Mixed-Autonomy Traffic",
"NeuroRule: Making Black-Box Neural Networks Explainable through Rule-set Evolution",
"What Looks Like a Capability Limit in Vision-Language Models I... | {"Rethinking Web Application Firewalls": 1.0, "Human Driver Temperament and the Safety Impact of a C-V2X Denial-of-Service Flooding Attack in Mixed-Autonomy Traffic": 0.0, "NeuroRule: Making Black-Box Neural Networks Explainable through Rule-set Evolution": 0.0, "What Looks Like a Capability Limit in Vision-Language Mo... | Pick the paper title that matches this abstract. | false | null | null |
fh-arxiv-title-2609.23548 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Generating sewing patterns from images and text requires modeling a heterogeneous representation composed of discrete topology and continuous geometry. Existing methods mainly follow two paradigms: diffusion-based methods enable holistic geometry generation by converting the entire pattern into a continuous representat... | [
"DAGFormer: Global Topology Prediction for Transformer Cross-Layer Connections",
"RoAR-VAE: Improving Latent Diffusability through Robust Prediction",
"Diffusion-Based Generation of Gait Trajectories",
"FlowATC: Aircraft Trajectory Prediction via Flow Matching",
"SewFusion: Tailored Generation of Topology a... | {"DAGFormer: Global Topology Prediction for Transformer Cross-Layer Connections": 0.0, "RoAR-VAE: Improving Latent Diffusability through Robust Prediction": 0.0, "Diffusion-Based Generation of Gait Trajectories": 0.0, "FlowATC: Aircraft Trajectory Prediction via Flow Matching": 0.0, "SewFusion: Tailored Generation of T... | Pick the paper title that matches this abstract. | false | null | null |
fh-arxiv-title-2609.19939 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Digital Educational Systems (DES) enable millions of individuals to learn or to acquire new skills. Their promises vary from one platform to another: some grant genuine degrees, others pledge rapid progress, drawing for this purpose on a range of techniques, from the recommendation of learning resources through complex... | [
"ALIBI: Adversarial Legitimacy Injection in Binary Input against LLM Malware Analyzers",
"A Flexible Recommendation System for Individuals and Groups",
"On-Policy Distillation Teaches New Skills but Not New Knowledge",
"Brown-Gerver-Ramsey Theorems in Small Dimensions",
"Who Decides? Agency and Legitimacy i... | {"ALIBI: Adversarial Legitimacy Injection in Binary Input against LLM Malware Analyzers": 0.0, "A Flexible Recommendation System for Individuals and Groups": 0.0, "On-Policy Distillation Teaches New Skills but Not New Knowledge": 0.0, "Brown-Gerver-Ramsey Theorems in Small Dimensions": 0.0, "Who Decides? Agency and Leg... | Which of these is the title of the paper described? | false | null | null |
fh-arxiv-cat-2609.28344 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Mizar: A 159M-Parameter Audio-Language Model for Audio Understanding
Audio-language models (ALMs) integrate acoustic perception with the knowledge encoded in language models, enabling contextual understanding of auditory events. Making these capabilities practical on devices with limited memory and computation motivat... | [
"cs.HC",
"cs.AI",
"cs.CC",
"cs.CV",
"eess.AS",
"cs.SD"
] | {"cs.HC": 0.0, "cs.AI": 0.0, "cs.CC": 0.0, "cs.CV": 0.0, "eess.AS": 0.0, "cs.SD": 1.0} | Classify this paper into its main arXiv category. | false | null | null |
fh-arxiv-title-2610.01509 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | An emerging hypothesis about reinforcement learning (RL) post-training of large language models (LLMs) is that it merely sharpens existing behaviors of a base model, improving single-shot accuracy at the cost of solution coverage. Although this trade-off has been observed in math and coding tasks, it need not extend to... | [
"Sharpening Tax in Post-Training",
"Verifiable Hidden Dynamics Play: Generating Agentic RL Environments from Solved Mechanisms",
"AgenticRoboRL: Bringing Coding-Agent Post-Training to Agentic Robotics",
"SupportCal: Label-Free Calibration of Post-Trained LLMs via Reference Support and Corroboration",
"Hill ... | {"Sharpening Tax in Post-Training": 1.0, "Verifiable Hidden Dynamics Play: Generating Agentic RL Environments from Solved Mechanisms": 0.0, "AgenticRoboRL: Bringing Coding-Agent Post-Training to Agentic Robotics": 0.0, "SupportCal: Label-Free Calibration of Post-Trained LLMs via Reference Support and Corroboration": 0.... | Pick the paper title that matches this abstract. | false | null | null |
fh-arxiv-cat-2609.16423 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | No Bit Left Behind: Using Brute-Force Lifting to Achieve Fully Static Binary Recompilation
Binary recompilation is a technique for operating directly on executable code. It promises to automate two important tasks: retrofitting security mitigations onto legacy binaries, and migrating binaries across instruction set ar... | [
"cs.CV",
"cs.IR",
"cs.CR",
"cs.AI",
"cs.SE",
"eess.SY"
] | {"cs.CV": 0.0, "cs.IR": 0.0, "cs.CR": 1.0, "cs.AI": 0.0, "cs.SE": 0.0, "eess.SY": 0.0} | Route this paper to the arXiv listing it belongs to. | false | null | null |
fh-arxiv-title-2609.19138 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Enabling robots to adapt to unfamiliar environments as readily as humans remains a moonshot goal of embodied AI. No finite collection of demonstrations can cover every task and situation a robot will encounter, making the ability to learn from context at deployment essential for generalization. Such in-context learning... | [
"In-Context Learning for Robots: Methods and Applications",
"In-context Robot Learning Made Simple: A Democratized Recipe for Manipulation Tasks",
"Has GPT-6-Astra Achieved Embodied Intelligence?",
"In-Context Robot Learning with VLM Agents",
"An Unexpected Robot Policy: Early Evaluations of GPT-6 Astra on ... | {"In-Context Learning for Robots: Methods and Applications": 0.0, "In-context Robot Learning Made Simple: A Democratized Recipe for Manipulation Tasks": 0.0, "Has GPT-6-Astra Achieved Embodied Intelligence?": 0.0, "In-Context Robot Learning with VLM Agents": 1.0, "An Unexpected Robot Policy: Early Evaluations of GPT-6 ... | Pick the paper title that matches this abstract. | false | null | null |
fh-arxiv-title-2609.19647 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | A neural turbulence closure defines a new boundary-value problem, $R(U)=N(U)+F(U)=0$, with a coupled Jacobian $J(U)=N'(U)+F'(U)$, where $N$ is the original mean-flow operator and $F$ the learned closure. We establish two consequences of global tangent dissipation. For a monotone original operator, a positive uniform ma... | [
"TicTacBench: Benchmarking Timing Closure Capabilities of Coding Agents",
"Regularization and compression of integral operator kernels with tensor paraproducts",
"Factorisability of Low Dimensional Non-Negative Integer Matrices",
"Well-posedness of neural turbulence closures and tangent dissipation",
"Monot... | {"TicTacBench: Benchmarking Timing Closure Capabilities of Coding Agents": 0.0, "Regularization and compression of integral operator kernels with tensor paraproducts": 0.0, "Factorisability of Low Dimensional Non-Negative Integer Matrices": 0.0, "Well-posedness of neural turbulence closures and tangent dissipation": 1.... | Pick the paper title that matches this abstract. | false | null | null |
fh-arxiv-cat-2609.24393 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Passive Hybrid Network-Based Intrusion Detection System (Hybrid-NIDS) Combining Suricata and Random Forest
This paper evaluates a passive Hybrid Network-based Intrusion Detection System (Hybrid-NIDS) prototype that combines Suricata with Random Forest flow classification and centralized ELK-based alert handling. The s... | [
"Sound (cs.SD)",
"Computation and Language (cs.CL)",
"Robotics (cs.RO)",
"Multiagent Systems (cs.MA)",
"Artificial Intelligence (cs.AI)",
"Cryptography and Security (cs.CR)"
] | {"Sound (cs.SD)": 0.0, "Computation and Language (cs.CL)": 0.0, "Robotics (cs.RO)": 0.0, "Multiagent Systems (cs.MA)": 0.0, "Artificial Intelligence (cs.AI)": 0.0, "Cryptography and Security (cs.CR)": 1.0} | Which arXiv category is this paper's primary category? | false | null | null |
fh-arxiv-title-2609.14789 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Artificial intelligence (AI) systems in education are developing on timescales that sit uneasily with conventional evaluation. By the time a large-scale trial has been designed, delivered, analysed and published, the technology under study may have changed materially. This creates a temporal problem for evidence-inform... | [
"Evaluating AI Tutoring at the Speed of Innovation: Practitioner-Led Micro-Randomised Trials of an AI Tutoring Platform in GCSE Science",
"Can AI systems have free will?",
"Teaching AI, Robotics, & Community: A Hubs-Based K-12 Education Framework for Reaching Rural Schools",
"Will It Teach as Intended? How Te... | {"Evaluating AI Tutoring at the Speed of Innovation: Practitioner-Led Micro-Randomised Trials of an AI Tutoring Platform in GCSE Science": 1.0, "Can AI systems have free will?": 0.0, "Teaching AI, Robotics, & Community: A Hubs-Based K-12 Education Framework for Reaching Rural Schools": 0.0, "Will It Teach as Intended? ... | Pick the paper title that matches this abstract. | false | null | null |
fh-arxiv-cat-2609.22028 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Spherical Harmonic Sliced Wasserstein Displacement Interpolation for Acoustic Source and Reflection Density Modeling
Spatial room impulse responses (SRIRs) capture directional distributions of acoustic sound-sources and their reflections. However, collecting SRIRs of moving sound-sources remains a challenge, requiring... | [
"Sound (cs.SD)",
"Systems and Control (EESS) (eess.SY)",
"Robotics (cs.RO)",
"Computer Vision and Pattern Recognition (cs.CV)",
"Artificial Intelligence (cs.AI)",
"Software Engineering (cs.SE)"
] | {"Sound (cs.SD)": 1.0, "Systems and Control (EESS) (eess.SY)": 0.0, "Robotics (cs.RO)": 0.0, "Computer Vision and Pattern Recognition (cs.CV)": 0.0, "Artificial Intelligence (cs.AI)": 0.0, "Software Engineering (cs.SE)": 0.0} | Which arXiv category is this paper's primary category? | false | null | null |
fh-arxiv-title-2609.28715 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Creative and artistic data visualizations communicate stories and invite engagement, yet how designers develop their expressive forms remains poorly understood. We investigate the design process of data artists through three complementary studies: an analysis of 40 public project accounts, artifact-anchored interviews ... | [
"One Model, Two Physical Stories: Auditing Misalignment in Multi-Modal World Modeling",
"Understanding Creative Design Practices among Data Artists",
"The Gold in Bias: Maturing the AI Design Process through Verification",
"AI-Assisted Social Story Intervention for Special Education: The Design of AdaptED Sto... | {"One Model, Two Physical Stories: Auditing Misalignment in Multi-Modal World Modeling": 0.0, "Understanding Creative Design Practices among Data Artists": 1.0, "The Gold in Bias: Maturing the AI Design Process through Verification": 0.0, "AI-Assisted Social Story Intervention for Special Education: The Design of Adapt... | Which title belongs to this abstract? | false | null | null |
fh-arxiv-title-2609.28878 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Sim-to-real transfer has made substantial progress, but can still produce controllers that remain stable and functional on hardware while suffering from degraded tracking accuracy due to residual dynamics mismatch. Correcting these errors typically requires identifying the underlying system dynamics, adapting the contr... | [
"Sim-to-Real Aware End-to-End Learning Environment for Micromobility",
"Online Sim-to-Real Adaptation via Closed-Loop System Modeling",
"ZYT-World: A Real-Time Controllable World Model for Closed-Loop Autonomous-Driving Simulation",
"Learning a Speed-adaptive Hip Exoskeleton Control Policy Via Sim-to-real Rei... | {"Sim-to-Real Aware End-to-End Learning Environment for Micromobility": 0.0, "Online Sim-to-Real Adaptation via Closed-Loop System Modeling": 1.0, "ZYT-World: A Real-Time Controllable World Model for Closed-Loop Autonomous-Driving Simulation": 0.0, "Learning a Speed-adaptive Hip Exoskeleton Control Policy Via Sim-to-re... | Pick the paper title that matches this abstract. | false | null | null |
fh-arxiv-title-2609.13755 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | LP-Newton solves a linear program through a sequence of nearest-point problems. Given an interior feasible point for an inequality-form LP, we construct the compact hull formed by the origin and its normalized constraint rows. Polar LP-Newton (P-LPN) follows the objective ray to the boundary of this row-polar hull, whe... | [
"Faster Verification of PJR$^+$ via Mincuts",
"Row-Polar LP-Newton for Linear Programming with Corral Repair",
"A $59/33$ Cut-LP Guarantee for Matching Augmentation",
"Presolve heuristics in HiGHS: implementation and computational study",
"Equilibrium bias and convergence in augmented primal--dual dynamics ... | {"Faster Verification of PJR$^+$ via Mincuts": 0.0, "Row-Polar LP-Newton for Linear Programming with Corral Repair": 1.0, "A $59/33$ Cut-LP Guarantee for Matching Augmentation": 0.0, "Presolve heuristics in HiGHS: implementation and computational study": 0.0, "Equilibrium bias and convergence in augmented primal--dual ... | Pick the paper title that matches this abstract. | false | null | null |
fh-arxiv-title-2609.26333 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Prefill and decode reward different approaches to quantization: low-precision arithmetic accelerates prompt processing, while compact weights reduce memory traffic during generation. We propose "disaggregated quantization" (DQ), which specializes computation formats, weights and storage placement to both of these phase... | [
"SlimWise: Decoupling Expert Pruning Across Prefill and Decode for Efficient MoE Serving",
"GGUF-Metadata Prediction of Single-Sequence llama.cpp Throughput Across Three Systems",
"Disaggregated Quantization: Specializing LLM Prefill and Decode",
"Crossflow: Prefill-Decode Elasticity for Agentic LLM Serving",... | {"SlimWise: Decoupling Expert Pruning Across Prefill and Decode for Efficient MoE Serving": 0.0, "GGUF-Metadata Prediction of Single-Sequence llama.cpp Throughput Across Three Systems": 0.0, "Disaggregated Quantization: Specializing LLM Prefill and Decode": 1.0, "Crossflow: Prefill-Decode Elasticity for Agentic LLM Ser... | Pick the paper title that matches this abstract. | false | null | null |
fh-arxiv-title-2609.18844 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Multimodal coding agents are expected to turn visual inputs into usable artifacts, and they act through a harness, the layer of tools, context management, and execution environment around the model. Existing evaluations often isolate short tool calls, API traces, or screenshot resemblance, and a low score under these p... | [
"Mid-Harness: Scaling Actions Between Model and Harness for Terminal Agents",
"Harness or Model? Isolating the Harness Effect in Agentic Coding with a Contamination-Controlled Private Suite",
"ReFigBench: Benchmarking Scientific Figure Reconstruction as Editable PowerPoint Artifacts",
"An Empirical Study of H... | {"Mid-Harness: Scaling Actions Between Model and Harness for Terminal Agents": 0.0, "Harness or Model? Isolating the Harness Effect in Agentic Coding with a Contamination-Controlled Private Suite": 0.0, "ReFigBench: Benchmarking Scientific Figure Reconstruction as Editable PowerPoint Artifacts": 1.0, "An Empirical Stud... | Which title belongs to this abstract? | false | null | null |
fh-arxiv-cat-2609.24016 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Context-Aware Pre-Deployment Evaluation of AI Systems: A Regulatory Framework for Nigerian Fintech
Commercial large language models are increasingly deployed across African fintech infrastructure for fraud detection and customer communication, yet no Nigerian or African continental regulatory instrument specifies what... | [
"Audio and Speech Processing",
"Robotics",
"Sound",
"Computation and Language",
"Artificial Intelligence",
"Information Theory"
] | {"Audio and Speech Processing": 0.0, "Robotics": 0.0, "Sound": 0.0, "Computation and Language": 0.0, "Artificial Intelligence": 1.0, "Information Theory": 0.0} | Classify this paper into its main arXiv category. | false | null | null |
fh-arxiv-cat-2609.29044 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Multi-Agent Orchestration of 3GPP Channel Estimators
Pilot-aided channel estimation is a decisive block in orthogonal frequency-division multiplexing (OFDM) receivers for both 5G New Radio (5G-NR) and Long-Term Evolution (LTE). A large body of estimators exists, from simple least-squares (LS) interpolation to statisti... | [
"cs.NI",
"cs.IT",
"cs.HC",
"cs.LG",
"cs.CR",
"cs.CL"
] | {"cs.NI": 0.0, "cs.IT": 1.0, "cs.HC": 0.0, "cs.LG": 0.0, "cs.CR": 0.0, "cs.CL": 0.0} | Which arXiv category is this paper's primary category? | false | null | null |
fh-arxiv-title-2609.23533 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Multimodal classifiers can converge to modality-dominant solutions in which one modality dominates the joint prediction, suppressing the learning of others. Existing balancing methods mainly adjust losses, gradients, or modality contributions, largely treating modality imbalance as an optimization problem while implici... | [
"PhyMo: A Physical-Field Modality for Multimodal AI4Physics",
"GeoBalance: Geometry-Aware Monitoring and Reconstruction with Asymmetric Optimization for Balanced Multimodal Learning",
"Chaos in the Text: Revealing the Modality Preference in Mixed-Modality Retrievers",
"BackTrend: Evaluating Scientific Weak-Si... | {"PhyMo: A Physical-Field Modality for Multimodal AI4Physics": 0.0, "GeoBalance: Geometry-Aware Monitoring and Reconstruction with Asymmetric Optimization for Balanced Multimodal Learning": 1.0, "Chaos in the Text: Revealing the Modality Preference in Mixed-Modality Retrievers": 0.0, "BackTrend: Evaluating Scientific W... | Which title belongs to this abstract? | false | null | null |
fh-arxiv-title-2609.19090 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Can an attacker turn influence over an artificial intelligence (AI) adviser into a harmful quantum error-correction update? We identify an ambiguity in passive syndrome records that obstructs recovery selection, then show how additional calibration measurements support certified recovery updates under uncertainty and d... | [
"Reinforcement Learning for Syndrome Extraction",
"QEMScore: How Much Does the Measurement Add to Learned Quantum Error Mitigation?",
"Symmetric Models for Syndrome Decoding",
"AutoRecLab: Describe the Experiment, Get the Code!",
"Securing quantum error correction against misleading advice from AI agents"
] | {"Reinforcement Learning for Syndrome Extraction": 0.0, "QEMScore: How Much Does the Measurement Add to Learned Quantum Error Mitigation?": 0.0, "Symmetric Models for Syndrome Decoding": 0.0, "AutoRecLab: Describe the Experiment, Get the Code!": 0.0, "Securing quantum error correction against misleading advice from AI ... | Which title belongs to this abstract? | false | null | null |
fh-arxiv-title-2609.12580 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Arm TrustZone for Armv8-M isolates Secure and Non-secure software, but developers must still coordinate peripheral attribution, interrupt routing, initialization, and gateway interfaces across separately built firmware images. Inconsistent assumptions between these images can compile successfully and emerge only as fau... | [
"Smoothed Analysis of Inconsistent A*",
"Secure RIS-Aided Multicasting: Globally Optimal Beam Management and Discrete-Phase RIS Configuration",
"Last Secure Code Benchmark: Is the Code You Vibe Secure?",
"MicroHasTEE: Bare-Metal Haskell for Type-Level Peripheral Ownership on Armv8-M",
"Custom PX4 firmware f... | {"Smoothed Analysis of Inconsistent A*": 0.0, "Secure RIS-Aided Multicasting: Globally Optimal Beam Management and Discrete-Phase RIS Configuration": 0.0, "Last Secure Code Benchmark: Is the Code You Vibe Secure?": 0.0, "MicroHasTEE: Bare-Metal Haskell for Type-Level Peripheral Ownership on Armv8-M": 1.0, "Custom PX4 f... | Pick the paper title that matches this abstract. | false | null | null |
fh-arxiv-title-2609.16061 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Span-level masked language modeling (MLM) has shown to be advantageous to pre-trained language models over the original single-token MLM, as entities/phrases and their dependencies are critical to language understanding. Previous works only consider span length with some discrete distributions, while the dependencies a... | [
"What masking geometry works best for EEG foundation models?",
"Informed Masking: Structure-Aware Perturbation for Reinforcement Learning in Diffusion Large Language Models",
"Category-Based MLM: Unifying Powertypes with Superclasses",
"Video-STLayout Pre-training",
"POSPAN: Position-Constrained Span Maskin... | {"What masking geometry works best for EEG foundation models?": 0.0, "Informed Masking: Structure-Aware Perturbation for Reinforcement Learning in Diffusion Large Language Models": 0.0, "Category-Based MLM: Unifying Powertypes with Superclasses": 0.0, "Video-STLayout Pre-training": 0.0, "POSPAN: Position-Constrained Sp... | Which title belongs to this abstract? | false | null | null |
fh-arxiv-title-2609.30186 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Vision-language models (VLMs) have become a common foundation for autonomous mobile GUI agents, but most existing systems rely on the VLM for both planning and action grounding at nearly every interaction step, leading to substantial latency and model-serving cost. We introduce Jev-Mobile, which shifts this paradigm to... | [
"SanSi: A Looped Typed Decision Model for System 1.5 Thinking",
"PriMobiBench: Characterizing Visual Privacy Leakage in VLM-Driven Mobile GUI Agents",
"Jev-Mobile: Jev as an Executor for Mobile GUI Agents",
"Latent World Action Model for GUI Agents",
"EchoPath: Execution-Level Replayable Memory for GUI Agen... | {"SanSi: A Looped Typed Decision Model for System 1.5 Thinking": 0.0, "PriMobiBench: Characterizing Visual Privacy Leakage in VLM-Driven Mobile GUI Agents": 0.0, "Jev-Mobile: Jev as an Executor for Mobile GUI Agents": 1.0, "Latent World Action Model for GUI Agents": 0.0, "EchoPath: Execution-Level Replayable Memory for... | Which title belongs to this abstract? | false | null | null |
fh-arxiv-title-2609.38334 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Large language models (LLMs) are increasingly deployed as agents for multi-step decision-making, yet transfer poorly to unseen environments. World-model methods address this by training agents to predict future observations, at the cost of additional training and errors that compound when predictions are used for plann... | [
"DreamTrue: Action-Faithful Robot World Model with Counterfactual Post-Training",
"Making Agents More Consistent: Skills Should Form Habits for Repeat Tasks",
"Agent-Editing World Model: Rethinking World Modeling for LLM Agents",
"MintAct: A Unified Visual Agent for Digital Environments",
"EVOKE: Eliciting ... | {"DreamTrue: Action-Faithful Robot World Model with Counterfactual Post-Training": 0.0, "Making Agents More Consistent: Skills Should Form Habits for Repeat Tasks": 0.0, "Agent-Editing World Model: Rethinking World Modeling for LLM Agents": 0.0, "MintAct: A Unified Visual Agent for Digital Environments": 0.0, "EVOKE: E... | Which title belongs to this abstract? | false | null | null |
fh-arxiv-cat-2609.27681 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | CasCVS-Net: A Staged Multi-Task Cascade for Critical View of Safety Assessment
Automated assessment of the Critical View of Safety (CVS) in laparoscopic cholecystectomy requires both recognition of the three CVS criteria and anatomical grounding in small, rare, and often occluded hepatocystic structures. Learning-base... | [
"cs.CV",
"cs.SE",
"cs.DS",
"cs.LG",
"cs.RO",
"cs.AI"
] | {"cs.CV": 1.0, "cs.SE": 0.0, "cs.DS": 0.0, "cs.LG": 0.0, "cs.RO": 0.0, "cs.AI": 0.0} | Classify this paper into its main arXiv category. | false | null | null |
fh-arxiv-cat-2609.29217 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | AdaptDuplex: from static to adaptive full-duplex spoken dialogue
Full-duplex spoken dialogue requires simultaneous listening and speaking at sub-second latency, under conversational timing and cognitive demands that change moment to moment. Yet current models mostly impose static operating points, lacking a systematic... | [
"Human-Computer Interaction (cs.HC)",
"Artificial Intelligence (cs.AI)",
"Audio and Speech Processing (eess.AS)",
"Computation and Language (cs.CL)",
"Systems and Control (EESS) (eess.SY)",
"Sound (cs.SD)"
] | {"Human-Computer Interaction (cs.HC)": 0.0, "Artificial Intelligence (cs.AI)": 0.0, "Audio and Speech Processing (eess.AS)": 0.0, "Computation and Language (cs.CL)": 0.0, "Systems and Control (EESS) (eess.SY)": 0.0, "Sound (cs.SD)": 1.0} | Classify this paper into its main arXiv category. | false | null | null |
fh-arxiv-title-2609.25743 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Interactive segmentation of 3D medical images supports quantitative analysis of anatomical structures and disease while allowing users to specify and refine their targets. Despite substantial progress by nnInteractive and VISTA3D, reliable segmentation across diverse clinical targets remains challenging, particularly f... | [
"PePESeg3D: Perception Prior Enhances Multi-Scale Segmentation for 3D Gaussian Splatting",
"SAMI3D-DW: Interactive Segmentation of Any 3D Medical Images",
"De-GAN - Dynamic Parameter Tuned GAN for 3D Medical Image Segmentation: A Step Towards Generalisation",
"Multimodal Routing and Region Refinement for Lang... | {"PePESeg3D: Perception Prior Enhances Multi-Scale Segmentation for 3D Gaussian Splatting": 0.0, "SAMI3D-DW: Interactive Segmentation of Any 3D Medical Images": 1.0, "De-GAN - Dynamic Parameter Tuned GAN for 3D Medical Image Segmentation: A Step Towards Generalisation": 0.0, "Multimodal Routing and Region Refinement fo... | Pick the paper title that matches this abstract. | false | null | null |
fh-arxiv-title-2609.17688 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Wearable assistants require episodic memory over egocentric video, yet current vision-language models face bounded frame budgets, growing visual-token costs, and long-context retrieval failures. Under these practical constraints, we study whether textual captions can serve as reusable episodic memory. We define the Epi... | [
"CapMem: A Benchmark for Caption-Based Episodic Memory in Egocentric Video",
"Persistent Planar Memory for Video World Models",
"MemLife: Curating and Reasoning over Long-Term Egocentric Video Memories",
"Ego2Act: Evaluating Goal-Directed Manipulation in Egocentric Video Generation",
"Learning What to Recal... | {"CapMem: A Benchmark for Caption-Based Episodic Memory in Egocentric Video": 1.0, "Persistent Planar Memory for Video World Models": 0.0, "MemLife: Curating and Reasoning over Long-Term Egocentric Video Memories": 0.0, "Ego2Act: Evaluating Goal-Directed Manipulation in Egocentric Video Generation": 0.0, "Learning What... | Which of these is the title of the paper described? | false | null | null |
fh-arxiv-title-2609.15801 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | We introduce loss-conditioned state execution, a model-agnostic method that decides whether to execute a world model's fixed feasible proposal or retain the current state. Predictive informativeness alone, however, does not establish whether an update will reduce downstream loss. Occurrence ranking can approach perfect... | [
"A Ranking Approach for Measuring Calibration",
"When Should a World Model Move? Loss-Conditioned State Execution",
"BranchDrive: A Branch-Structured Dataset for Action-Conditioned Driving Prediction",
"The Lattice of Transition Laws",
"Intelligent Interaction Techniques (IIxT) - Proposal"
] | {"A Ranking Approach for Measuring Calibration": 0.0, "When Should a World Model Move? Loss-Conditioned State Execution": 1.0, "BranchDrive: A Branch-Structured Dataset for Action-Conditioned Driving Prediction": 0.0, "The Lattice of Transition Laws": 0.0, "Intelligent Interaction Techniques (IIxT) - Proposal": 0.0} | Pick the paper title that matches this abstract. | false | null | null |
fh-arxiv-cat-2609.14675 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | PU classification under Non-SCAR: clustering-assisted logistic model with oversampling enhancement
This study addresses the PU classification problem under violations of the SCAR assumption. We investigate logistic regression-based approaches, namely the cluster method and its extensions with strict and non-strict Las... | [
"cs.CR",
"math.OC",
"cs.CV",
"cs.CL",
"cs.SD",
"stat.ML"
] | {"cs.CR": 0.0, "math.OC": 0.0, "cs.CV": 0.0, "cs.CL": 0.0, "cs.SD": 0.0, "stat.ML": 1.0} | Route this paper to the arXiv listing it belongs to. | false | null | null |
fh-arxiv-title-2609.27262 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Production customer-support systems often require LLMs to support multiple skills, such as intent classification, question answering, summarization, or tool-use decisions. A central deployment question is whether these skills should be handled by separate task-specialist models or by a single model trained through mult... | [
"Beyond Uniform Subspaces: Spectrum-Aware and Depth-Adaptive Fusion for Multi-Task Model Merging",
"Training Specialist Models without Reasoning Trajectories for Domain Expert Distillation",
"ReMova: Fine-tuning LLMs for English to Belarusian translation",
"Domain-Specific Jargon in Large Language Models: A C... | {"Beyond Uniform Subspaces: Spectrum-Aware and Depth-Adaptive Fusion for Multi-Task Model Merging": 0.0, "Training Specialist Models without Reasoning Trajectories for Domain Expert Distillation": 0.0, "ReMova: Fine-tuning LLMs for English to Belarusian translation": 0.0, "Domain-Specific Jargon in Large Language Model... | Pick the paper title that matches this abstract. | false | null | null |
fh-arxiv-cat-2609.14088 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | Consistency-Robustness Tradeoffs for Strategyproof Scheduling with Predictions
We study strategyproof scheduling on \(n\) unrelated machines with predictions. Each machine is controlled by an agent with privately known processing times, while the mechanism receives a public prediction of the processing-time matrix bef... | [
"Computer Vision and Pattern Recognition",
"Machine Learning",
"Computer Science and Game Theory",
"Multiagent Systems",
"Robotics",
"Computation and Language"
] | {"Computer Vision and Pattern Recognition": 0.0, "Machine Learning": 0.0, "Computer Science and Game Theory": 1.0, "Multiagent Systems": 0.0, "Robotics": 0.0, "Computation and Language": 0.0} | What is the primary subject area of this paper? | false | null | null |
fh-arxiv-cat-2609.19780 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | A Separation Between Distribution-Free SQ Learning and Dimension Complexity
We show that there exists a class of boolean functions C such that $(i)$ there is a distribution-independent statistical query algorithm for learning C that makes a polynomial number of queries of inverse polynomial tolerance and $(ii)$ for an... | [
"cs.DS",
"cs.LO",
"eess.IV",
"cs.AI",
"cs.CC",
"cs.NI"
] | {"cs.DS": 0.0, "cs.LO": 0.0, "eess.IV": 0.0, "cs.AI": 0.0, "cs.CC": 1.0, "cs.NI": 0.0} | Classify this paper into its main arXiv category. | false | null | null |
fh-arxiv-title-2610.06804 | arxiv | fh-arxiv | gliclass | arXiv author categories (source metadata) | permissive | cc0-1.0 | null | null | null | null | A language model can give a correct answer more probability than any single incorrect answer and still usually sample an incorrect one, because the incorrect answers together hold more probability. The power distribution raises each complete answer's probability to a power above one and renormalizes, shifting probabili... | [
"Graph-Based Stochastic Power-UCT: Monte-Carlo Graph Search with Power Mean Estimation",
"Sharpening Tax in Post-Training",
"When and Where to Trust the Teacher: Unifying On-Policy Distillation and GRPO through Entropy-Calibrated Credit Assignment",
"PreSTO: Predictive Subtree Prefetching for Fast LLM Power S... | {"Graph-Based Stochastic Power-UCT: Monte-Carlo Graph Search with Power Mean Estimation": 0.0, "Sharpening Tax in Post-Training": 0.0, "When and Where to Trust the Teacher: Unifying On-Policy Distillation and GRPO through Entropy-Calibrated Credit Assignment": 0.0, "PreSTO: Predictive Subtree Prefetching for Fast LLM P... | Which of these is the title of the paper described? | false | null | null |
End of preview. Expand in Data Studio
Curation decision mix
The training data of davanstrien/ultrabert-380m-curation, a small decision model that answers curation questions about web pages, prompts, maths problems and library records ("what topic is this page?", "how educational is it?", "how hard is this problem?"). Each row is a text, a question, a set of options and a probability for each option. Every row records where it came from, who made its label and under which licence.
This card is also the data recipe: it says which sources were used, why, and which choices moved the scores (see How the mix was designed). The scripts are in github.com/davanstrien/curation-decision-recipe. If you train decision models, please share how you made your data too so others can learn from your work and open models can keep improving!
| Rows | 516,964 released, from a 535,730-row training mix; 18,766 base rows without a verified open licence are left out (see Licences) |
| Sources | 9: public task collections, DecisionBank, curation labels from web/prompt datasets, library catalogues, program-computed tasks |
| Row shapes | decision rows {state, question, target} and GLiClass rows {text, all_labels, true_labels, prompt} |
| Labels made by | dataset gold, library cataloguers, programs, and LLMs (named per source and per row) |
| Licence | mixed, per row; web page text is not included (labels plus a pointer to the source row; rebuild snippet below) |
| Trains | davanstrien/ultrabert-380m-curation (UltraBERT-380M + GLiClass head; 12,000 steps over this mix) |
| Build | mix builder on Hugging Face Jobs (CPU), then scripts/release_build.py for the release (a record of how this repo was made: it reads the mix from a private bucket); test sets fixed before the mix was built, and every row checked against them |
import json
from datasets import load_dataset
ds = load_dataset("davanstrien/curation-decision-mix", "all", split="train") # or one config per source, e.g. "decisionbank"
ok = ds.filter(lambda r: r["licence_class"] == "permissive") # filter by licence before reuse
decisions = ok.filter(lambda r: r["row_type"] == "decision") # decision rows; the rest are GLiClass rows
q = json.loads(decisions[0]["question"]) # nested fields are JSON strings
Layout
One config per source, plus all. web rows have an empty state by design (see Web text). Splitting by source keeps licence terms together, so a user can take only the parts
whose terms suit them.
| Config | Rows (training mix) | Shape |
|---|---|---|
base |
173,859 (192,625 in training; licence filter) | GLiClass |
decisionbank |
165,277 | decision |
prompt |
52,491 | decision |
web |
41,890 (labels only; state empty until rebuilt) |
decision |
nls |
23,846 | GLiClass |
tool |
21,089 | GLiClass |
harvard |
13,918 | decision |
arxiv |
13,597 | GLiClass |
procpile |
10,997 | GLiClass |
manifest/training_mix_manifest.parquet lists all 535,730 rows of the training mix (including the 18,766 rows not
released here): row id, source, licence class, labeller and a SHA-256 hash of each row's text.
Columns (union of both shapes; unused fields are null):
| Column | Meaning |
|---|---|
id |
stable row id; prefix names the source (db/…, procpile-…, knowledgator/gliclass-v2.0:…) |
config, src |
source group and the finer origin (e.g. db_bitext_support, a tasksource task) |
row_type |
decision or gliclass |
teacher |
who made the label (e.g. Llama-3.1-405B, cataloguer, program, 50% source label + 50% Qwen3.8-27B) |
licence_class, licence |
permissive / sa (share-alike, e.g. CC BY-SA, CDLA-Sharing) / other_open (MPL-2.0, AFL-3.0) / publisher (text not redistributed), and the licence id |
state |
the input (decision rows): a string, or a record as a JSON string |
state_is_record |
decision rows: true when state was a record (dict or list) in training; false when it was text (some text states look like JSON) |
question |
JSON string: {type, instructions, criteria}, type is noul or choice; noul = yes/no; choice options are {key: description}; graded scales use level_<k> keys, lowest first |
target |
JSON string: probability per option (soft targets allowed; sums to 1 for single-answer rows) |
text, all_labels, true_labels, prompt |
GLiClass rows: text, label list, {label: score} or a list (JSON string), optional prompt |
multi_label |
true when several options can be correct (per-option yes/no loss in training) |
source_dataset, source_row |
web rows only: source dataset and row index in its train split |
Example decision row (DecisionBank, shortened):
{"id": "db/decider/bitext_support/286978::q1", "src": "db_bitext_support", "row_type": "decision",
"teacher": "50% decider label + 50% Qwen3.8-27B", "licence_class": "sa", "licence": "cdla-sharing-1.0",
"state": "how do i remove several articles from order {{Order Number}}",
"question": {"type": "choice", "instructions": "What is the specific intent of this request?",
"criteria": {"change order": "change order", "get refund": "get refund", "...": "..."}},
"target": {"change order": 1.0, "get refund": 0.0, "...": 0.0}}
Example GLiClass row (procedural pile, shortened):
{"id": "procpile-game_forced_win-89307", "row_type": "gliclass", "teacher": "program",
"text": "In this graph game, decide whether player can force a win. … The answer is Yes or No.",
"all_labels": ["No", "Yes"], "true_labels": {"No": 0.0, "Yes": 1.0}, "prompt": ""}
Web text. The web config carries labels only. source_dataset and source_row point at the row in the
source dataset's train split; the training text was the first 3,000 characters. To rebuild it for your own use
(do not redistribute it):
from datasets import load_dataset
src = {}
def rebuild(r):
if r["source_dataset"] not in src:
src[r["source_dataset"]] = load_dataset(r["source_dataset"], split="train")
return {"state": src[r["source_dataset"]][r["source_row"]]["text"][:3000]}
web = load_dataset("davanstrien/curation-decision-mix", "web", split="train").map(rebuild)
Row indices assume the source datasets are unchanged since October 2026. The snippet loads each source dataset in full; it has not been tested end to end.
Train on it. data/release_to_jsonl.py
in the recipe repo converts this dataset to the trainer's JSONL input (uv run data/release_to_jsonl.py --out mix.jsonl; filter with --configs and --licence-class). We checked it against the private training mix: all
475,074 released rows with text are identical on every field the trainer reads. The recipe README has a tested
command that converts and trains on one GPU Job.
Sources
| Source | Rows | What the rows ask | Labels made by | Licence |
|---|---|---|---|---|
| base (tasksource + Knowledgator GLiClass data: v2.0, v3-logic) | 173,859 (192,625 trained on) | many public classification tasks | dataset gold; knowledgator gliclass-v2.0 / v3-logic (LLM-generated) | per source repo, in licence; only rows with a verified permissive, share-alike or other open licence are released (see Licences) |
| DecisionBank-260k | 165,277 | public tasks, written and programmatic decision questions | 50 % source label + 50 % Qwen3.8-27B; Qwen3.8-27B with thinking; 50 % Jeff 1.0 teachers (Qwen3.6-27B, Qwen3.5-122B-A10B) + 50 % Qwen3.8-27B (4,188 rows); code | per row: CC BY / CC BY-SA, Apache, MIT, CC0 (public domain), CDLA-Sharing-1.0; non-commercial, research-only, unknown and other terms dropped |
| prompt slice | 52,491 | NuminaMath-1.5 question type / problem type / validity; DeepMath-103K difficulty / topic; WebInstruct-verified answer type / category / difficulty; LMArena criteria and tags | dataset columns (NuminaMath, WebInstruct: labeller not documented); GPT-4o (DeepMath); LMArena classifiers | Apache-2.0; MIT; Apache-2.0; CC-BY-4.0 |
| web slice (ids and labels only) | 41,890 | WebOrganizer topic and format; FineWeb-Edu educational value | Llama-3.1-405B (WebOrganizer; the annotation datasets state no licence); Llama-3-70B (FineWeb-Edu) | labels: none stated; ODC-By. Page text: publisher copyright, not included (rebuild snippet above) |
| National Library of Scotland catalogue | 23,846 | subject areas of library records | National Library of Scotland cataloguers | CC0 |
| tool lane (code decisions) | 21,089 | linter findings (classification and yes/no) and complexity comparisons on functions from Python packages | program-computed (linter and complexity tools) | permissive package licences (MIT / BSD / Apache-2.0 / PSF); the package is not recorded per row |
| Harvard Library modern catalogue | 13,918 | Library of Congress subject class, literary form, government publication | Harvard Library cataloguers (from their catalogue records) | CC0 |
| arXiv | 13,597 | arXiv subject category | author-chosen categories | arXiv metadata (CC0) |
| procedural pile | 10,997 | program-checkable reasoning | program-computed | CC BY 4.0 (attribute) |
How the mix was designed
The mix results from a series of screened experiments (2026-10-23 → 10-27). Each choice below states what it rests on: earlier decision-model reports, published curation work, or our own measurements. "Curation score" is the benchmark used for selection: the mean of 16 curation test units (see the model card for the grouping and all 16 scores).
1. Start from a broad public base, cleaned by task identity. The base follows the GLiClass v4 recipe (knowledgator): tasksource classification tasks plus knowledgator's GLiClass training sets, so the model keeps general zero-shot skill. Contamination was screened by task, not only by canary strings: tasksource re-hosts BIG-bench tasks without the canary, and whole sources (math_qa, cosmos_qa, vitaminc, qasc, social_i_qa, …) were removed (2.7 % of the pool).
2. Prefer clean, audited, family-covering data to volume. Earlier decision-model reports found bulk open decision data neutral or harmful, gains staying inside the task families covered, and source audits worth more than extra rows. Our own runs agreed: adding bulk without audits did not help; diverse, audited sources did (+4 points over the base alone on a sample of a public decision benchmark suite). So every source was audited before training:
- Text-blind baselines on every source (majority label, longest option, options-only ranker), in the spirit of hypothesis-only baselines for NLI (Poliak et al., 2018; Gururangan et al., 2018). A generated-schema source that looked like +4.5 turned out to be mostly a "longest option is right" shortcut (real gain ≈ +2.3), so sources with option-only leaks were fixed or dropped.
- Distractors weighted by gold frequency. With uniformly sampled wrong options, a real-label source (arXiv) leaked a label prior (+18 points over chance from the options alone). The Harvard rows needed the same fix (options-only 52 % before it).
- Same writing style for every class. A source where one class was human-written and the other LLM-written taught "LLM style means unsupported" and was dropped.
3. Train on the labels real curation pipelines use. The curation slices reuse public labels from curation work: WebOrganizer topic and format (Wettig et al., 2025, Organize the Web), FineWeb-Edu educational value (Penedo et al., 2024), and prompt tags from NuminaMath-1.5, DeepMath-103K, WebInstruct-verified and LMArena. Fixed-head classifiers trained on the same labels were used as a ceiling check before any claim (they still win their own labels: 0.87 / 0.83 vs 0.84 / 0.74 for WebOrganizer topic/format).
4. Keep the third-party labels; document them. We tested replacing every LLM-made label with labels from a panel
of open models (2 seeds). The panel matched the 405B topic labels well (0.815) but lost 5.1 curation-score points,
almost all on graded axes (maths difficulty, educational value). Rubric-anchored prompts helped the panel (+0.07
rank correlation against Omni-MATH human ratings) but did not close the gap. So the release keeps the original labels and
states each labeller, model and licence (Sources table and the teacher column).
5. Add human-made labels where they exist. Library catalogue records (National Library of Scotland, Harvard) carry labels made by cataloguers. Harvard rows lifted an unseen library scheme by +4.6 accuracy (3 seeds, all 9 seed pairs positive) without hurting the curation score; the cost to watch is the Aegis safety test set (−4.7).
6. Add DecisionBank, licence-filtered. DecisionBank rows added +0.70 curation-score points at equal steps and +0.77 at equal passes over the other rows (3 seeds each, all 9 pairs positive), mostly on evals whose gold is independent of any training labeller. We kept only rows with allowed licences; we dropped the HelpSteer families because they are held-out test sets.
7. Program-computed lanes for reasoning and tools. Procedural reasoning tasks, tool selection and arXiv categories have labels that are computed or author-chosen, so no LLM labeller is involved (the task design still shapes them). They help general decision tasks more than the curation score.
How train and eval were split
- Evals first. The eval sets (v3: 40 sets, 72,533 rows) were frozen before any mix was built, and every mix was decontaminated against all of them (exact, normalised, prefix, number-stripped skeleton, containment, 13-gram).
- Hold out whole sources and schemes, not random rows. Dev sets drawn from the training pool overestimate held-out scores (earlier reports saw dev scores rise while scores on private test sets fell, and data shaped like a benchmark inflate that benchmark's public score). So held-out test sets are whole sources (HelpSteer2/3, Polaris, Aegis, WildGuard) and label schemes not seen in training (the item types of US company filings, SEC form 8-K; a library item-form scheme).
- Separate same-labeller from independent gold. Where an eval's gold comes from the same LLM that labelled training rows (WebOrganizer, FineWeb-Edu, DeepMath), it partly measures agreement with that labeller. The fair comparison uses independent gold (Omni-MATH human ratings, Essential-Web), reported apart in the model card.
- Seeds and review. Each data choice was screened with ≥ 2 training seeds and claimed with ≥ 3; every claimed result was checked by two independent automated reviews (separate AI models) before adoption.
Decontamination
We checked every row against every text in the test sets (version v3). Removed: 1,562 base rows (478 overlaps + 1,084 rows from HelpSteer3 sources, a held-out test source), 1,035 DecisionBank rows (mostly HelpSteer3 / Aegis prompts), 1 Harvard row; the HelpSteer families of DecisionBank were dropped before the build. A 12-gram near-duplicate check found 15 of 1,038 HelpSteer2 and 15 of 1,832 HelpSteer3 eval items sharing text with a DecisionBank row. The catalogue test sets were kept out by id and title when the Harvard and National Library of Scotland sources were built.
Licences
Licence classes in this release: permissive 375,637 rows, share-alike 98,941, other open 496, publisher terms 41,890
(web labels only; page text trained on, never redistributed). DecisionBank's own parts are CC BY-SA 4.0;
share-alike rows keep their licence in this dataset. The model is released under Apache-2.0. These are the licences
stated by each source; this is not legal advice, so check the terms of the sources you reuse.
Base rows, by the licence on the source repo's card (or, for tasksource tasks, the origin dataset's card; not legal
advice): of 200,000 rows before decontamination, 178,084 verified permissive, 1,127 share-alike (CC BY-SA), 497 under
other open licences (Mozilla Public License, Academic Free License), 13,263 not traced to an origin, 5,971 with no
licence stated and 1,058 under the GPL. The model was trained on all of them. This dataset releases only the
permissive, share-alike and other-open rows (173,859 after decontamination); the per-row licence is in licence.
Compute
Every step ran on Hugging Face Jobs, with sources, mixes and test sets in Hugging Face storage buckets mounted into each job.
- Building the data (source conversion, licence filtering, decontamination, the mix itself) ran on CPU jobs; the final mix build takes about 16 minutes.
- Labelling experiments (open-model panels, rubric and pairwise checks; see design note 4) called LLMs through Hugging Face Inference Providers from Jobs. Those labels are not in the final mix.
- Training and evaluation of each data choice ran on single RTX PRO 6000 GPU jobs, about an hour per training run.
- The release build (
scripts/release_build.py) is a CPU job that reads the mix from a bucket and writes this repo.
Known issues
- Several label sources are LLM outputs (named above and per row in
teacher). The WebOrganizer annotation datasets carry no licence statement. The Llama 3 and Llama 3.1 licences cover the Llama-made labels (WebOrganizer, FineWeb-Edu); read them before you reuse those rows. - Use terms of the original sources apply to their rows; check
licence/srcper row before reuse. - Multilingual rows (DecisionBank) are present; the model was evaluated in English only.
- Difficulty labels are subjective and depend on the labeller (LLM-made here; human ratings agree only moderately).
Notes for agents
- Filter on
licence_class/licencebefore redistributing any subset; never redistribute rebuiltwebtext. targetis a distribution, not a class index; takeargmaxfor hard labels. Graded questions (level_<k>keys) are ordered: use the expected level or a rank correlation (Spearman), not top-1 accuracy alone.- This release is not the exact training mix: 18,766 base rows are left out and web text is removed.
- Evals are not in this dataset; they are listed by name in the model card. Do not train on Omni-MATH, Essential-Web, HelpSteer, Polaris, Aegis or WildGuard if you want to compare with the model card.
- Downloads last month
- 31
Models trained or fine-tuned on davanstrien/curation-decision-mix
Space using davanstrien/curation-decision-mix 1
Collection including davanstrien/curation-decision-mix
Papers for davanstrien/curation-decision-mix
Organize the Web: Constructing Domains Enhances Pre-Training Data Curation
Paper • 2502.10341 • Published • 3
The FineWeb Datasets: Decanting the Web for the Finest Text Data at Scale
Paper • 2406.17557 • Published • 108
Hypothesis Only Baselines in Natural Language Inference
Paper • 1805.01042 • Published
Annotation Artifacts in Natural Language Inference Data
Paper • 1803.02324 • Published