Search is not available for this dataset
id unknown | title unknown | abstract unknown | categories unknown | update_date timestamp[us]date 2007-05-23 00:00:00 2026-01-30 00:00:00 | stratum unknown | minimax_label unknown | minimax_is_new bool 2
classes | minimax_confidence float64 0.6 1 ⌀ | minimax_reasoning unknown | qwen_is_new bool 2
classes | qwen_confidence unknown | qwen_reason unknown | qwen_raw_response unknown | consensus bool 2
classes | consensus_label unknown |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
"2406.11192" | "Beyond Boundaries: Learning a Universal Entity Taxonomy across Datasets\n and Languages for Open Named Entity Recognition" | " Open Named Entity Recognition (NER), which involves identifying arbitrary\ntypes of entities from arbitrary domains, remains challenging for Large\nLanguage Models (LLMs). Recent studies suggest that fine-tuning LLMs on\nextensive NER data can boost their performance. However, training directly on\nexisting datasets... | "cs.CL" | 2025-04-22T00:00:00 | "dataset_in_title" | "new_dataset" | true | 0.95 | "This paper introduces B2NERD as a new dataset for Open NER. Key indicators: (1) They explicitly 'present' B2NERD as a compact dataset designed to guide LLMs' generalization; (2) It's refined from 54 existing datasets through a two-step process involving entity taxonomy construction (400+ types) and data pruning for di... | true | "HIGH" | "The paper creates and releases B2NERD, a new curated dataset derived from 54 existing datasets through a two-step refinement process involving a universal taxonomy and data pruning, resulting in a novel annotated resource for Open NER." | "INTRODUCES_NEW_DATASET: YES \nCONFIDENCE: HIGH \nREASON: The paper creates and releases B2NERD, a new curated dataset derived from 54 existing datasets through a two-step refinement process involving a universal taxonomy and data pruning, resulting in a novel annotated resource for Open NER." | true | "new_dataset" |
"2101.12457" | "RetaGNN: Relational Temporal Attentive Graph Neural Networks for\n Holistic Sequential Recommendation" | " Sequential recommendation (SR) is to accurately recommend a list of items for\na user based on her current accessed ones. While new-coming users continuously\narrive in the real world, one crucial task is to have inductive SR that can\nproduce embeddings of users and items without re-training. Given user-item\ninter... | "cs.IR cs.LG cs.SI" | 2021-02-01T00:00:00 | "random" | "not_new_dataset" | false | 0.95 | "This paper proposes a new model (RetaGNN) for sequential recommendation and evaluates it on existing datasets (MovieLens, Instagram, Book-Crossing). The paper does not introduce, create, collect, curate, or release any new dataset. The primary contribution is the model architecture and its components (relational atten... | false | "HIGH" | "The paper uses existing datasets (MovieLens, Instagram, and Book-Crossing) for experiments without creating or releasing a new data collection." | "INTRODUCES_NEW_DATASET: NO \nCONFIDENCE: HIGH \nREASON: The paper uses existing datasets (MovieLens, Instagram, and Book-Crossing) for experiments without creating or releasing a new data collection." | true | "not_new_dataset" |
"2306.09424" | "SSL4EO-L: Datasets and Foundation Models for Landsat Imagery" | " The Landsat program is the longest-running Earth observation program in\nhistory, with 50+ years of data acquisition by 8 satellites. The multispectral\nimagery captured by sensors onboard these satellites is critical for a wide\nrange of scientific fields. Despite the increasing popularity of deep learning\nand rem... | "cs.LG cs.CV eess.IV" | 2023-10-24T00:00:00 | "dataset_in_title" | "new_dataset" | true | 0.95 | "The paper explicitly introduces multiple new datasets as primary contributions: (1) SSL4EO-L, a new 5M image patch dataset for self-supervised learning on Landsat imagery, described as 'the first ever dataset' and 'largest Landsat dataset in history'; (2) modernized versions of existing cloud detection datasets (L7 Ir... | true | "HIGH" | "The paper introduces SSL4EO-L, a new dataset comprising 5M image patches specifically designed for self-supervised learning on Landsat imagery, and also releases new benchmark datasets for Landsat 4-5 TM and Landsat 7 ETM+ SR, which were previously unavailable." | "INTRODUCES_NEW_DATASET: YES \nCONFIDENCE: HIGH \nREASON: The paper introduces SSL4EO-L, a new dataset comprising 5M image patches specifically designed for self-supervised learning on Landsat imagery, and also releases new benchmark datasets for Landsat 4-5 TM and Landsat 7 ETM+ SR, which were previously unavailable... | true | "new_dataset" |
"2508.06763" | "SafePLUG: Empowering Multimodal LLMs with Pixel-Level Insight and Temporal Grounding for Traffic Accident Understanding" | "Multimodal large language models (MLLMs) have achieved remarkable progress across a range of vision-language tasks and demonstrate strong potential for traffic accident understanding. However, existing MLLMs in this domain primarily focus on coarse-grained image-level or video-level comprehension and often struggle to... | "cs.CV cs.AI" | 2025-11-03T00:00:00 | "corpus_annotation" | "new_dataset" | true | 0.95 | "The abstract explicitly states: 'To advance the development of MLLMs for traffic accident understanding, we curate a new dataset containing multimodal question-answer pairs centered on diverse accident scenarios, with detailed pixel-level annotations and temporal event boundaries.' This indicates the dataset is a prim... | true | "HIGH" | "The paper explicitly states that it \"curates a new dataset containing multimodal question-answer pairs centered on diverse accident scenarios, with detailed pixel-level annotations and temporal event boundaries,\" and provides a public link for access." | "INTRODUCES_NEW_DATASET: YES \nCONFIDENCE: HIGH \nREASON: The paper explicitly states that it \"curates a new dataset containing multimodal question-answer pairs centered on diverse accident scenarios, with detailed pixel-level annotations and temporal event boundaries,\" and provides a public link for access." | true | "new_dataset" |
"2211.01551" | "Crime Prediction using Machine Learning with a Novel Crime Dataset" | " Crime is an unlawful act that carries legal repercussions. Bangladesh has a\nhigh crime rate due to poverty, population growth, and many other\nsocio-economic issues. For law enforcement agencies, understanding crime\npatterns is essential for preventing future criminal activity. For this\npurpose, these agencies ne... | "cs.LG cs.AI cs.CY" | 2022-11-04T00:00:00 | "corpus_annotation" | "new_dataset" | true | 0.98 | "This paper explicitly introduces a novel crime dataset as its primary contribution. The authors manually collected 6574 crime incident records from Bangladeshi newspaper archives over a seven-year period, enriched with geographic, weather, and demographic data from external sources. The abstract states 'This paper int... | true | "HIGH" | "The paper explicitly states it creates and releases a novel crime dataset by manually gathering and integrating crime news, geo-location, weather, and demographic data into a structured dataset with 36 engineered features." | "INTRODUCES_NEW_DATASET: YES \nCONFIDENCE: HIGH \nREASON: The paper explicitly states it creates and releases a novel crime dataset by manually gathering and integrating crime news, geo-location, weather, and demographic data into a structured dataset with 36 engineered features." | true | "new_dataset" |
"2403.06668" | "PeerAiD: Improving Adversarial Distillation from a Specialized Peer\n Tutor" | " Adversarial robustness of the neural network is a significant concern when it\nis applied to security-critical domains. In this situation, adversarial\ndistillation is a promising option which aims to distill the robustness of the\nteacher network to improve the robustness of a small student network. Previous\nworks... | "cs.LG cs.CV" | 2024-05-20T00:00:00 | "random" | "not_new_dataset" | false | 0.95 | "This paper introduces PeerAiD, a novel adversarial distillation framework that trains peer and student networks simultaneously to improve adversarial robustness. The primary contribution is the PeerAiD methodology, not a dataset. The paper evaluates the proposed method on the existing TinyImageNet benchmark dataset us... | false | "HIGH" | "The paper uses the existing TinyImageNet dataset and does not create, curate, or release any new data collection, annotated dataset, or benchmark." | "INTRODUCES_NEW_DATASET: NO \nCONFIDENCE: HIGH \nREASON: The paper uses the existing TinyImageNet dataset and does not create, curate, or release any new data collection, annotated dataset, or benchmark." | true | "not_new_dataset" |
"2503.10522" | "AudioX: Diffusion Transformer for Anything-to-Audio Generation" | " Audio and music generation have emerged as crucial tasks in many\napplications, yet existing approaches face significant limitations: they\noperate in isolation without unified capabilities across modalities, suffer\nfrom scarce high-quality, multi-modal training data, and struggle to\neffectively integrate diverse ... | "cs.MM cs.CV cs.LG cs.SD eess.AS" | 2025-04-24T00:00:00 | "corpus_annotation" | "new_dataset" | true | 0.95 | "The abstract explicitly states: 'To address data scarcity, we curate two comprehensive datasets: vggsound-caps with 190K audio captions based on the VGGSound dataset, and V2M-caps with 6 million music captions derived from the V2M dataset.' These datasets are presented as key contributions to address data scarcity lim... | true | "HIGH" | "The paper explicitly curates and releases two new datasets, vggsound-caps and V2M-caps, which are novel annotated collections of audio and music captions derived from existing datasets but significantly expanded and restructured for training." | "INTRODUCES_NEW_DATASET: YES \nCONFIDENCE: HIGH \nREASON: The paper explicitly curates and releases two new datasets, vggsound-caps and V2M-caps, which are novel annotated collections of audio and music captions derived from existing datasets but significantly expanded and restructured for training." | true | "new_dataset" |
"2309.03907" | "DrugChat: Towards Enabling ChatGPT-Like Capabilities on Drug Molecule\n Graphs" | " A ChatGPT-like system for drug compounds could be a game-changer in\npharmaceutical research, accelerating drug discovery, enhancing our\nunderstanding of structure-activity relationships, guiding lead optimization,\naiding drug repurposing, reducing the failure rate, and streamlining clinical\ntrials. In this work,... | "q-bio.BM cs.LG" | 2023-09-11T00:00:00 | "corpus_annotation" | "new_dataset" | true | 0.95 | "The paper explicitly states they 'collected instruction tuning datasets which contain 10,834 drug compounds and 143,517 question-answer pairs' and makes the code and data available. This is a substantial newly collected dataset of drug compounds with QA pairs, released as a resource for the community. While used to tr... | true | "HIGH" | "The paper explicitly states that it collected and released a new instruction tuning dataset containing 10,834 drug compounds and 143,517 question-answer pairs, which constitutes a novel annotated dataset." | "INTRODUCES_NEW_DATASET: YES \nCONFIDENCE: HIGH \nREASON: The paper explicitly states that it collected and released a new instruction tuning dataset containing 10,834 drug compounds and 143,517 question-answer pairs, which constitutes a novel annotated dataset." | true | "new_dataset" |
"2512.11089" | "TPV: Parameter Perturbations Through the Lens of Test Prediction Variance" | "We identify test prediction variance (TPV)-- the first-order sensitivity of model outputs to parameter perturbations around a trained solution-- as a unifying quantity that links several classical observations about generalization in deep networks. TPV is a fully label-free object whose trace form separates the geomet... | "stat.ML cs.LG" | 2026-01-23T00:00:00 | "random" | "not_new_dataset" | false | 0.95 | "This paper introduces TPV (Test Prediction Variance) - a theoretical framework/metric for analyzing model generalization through parameter perturbations. The paper does not create, collect, or release any new dataset. It evaluates the proposed TPV method on existing benchmark datasets (implied by 'across datasets and ... | false | "HIGH" | "The paper proposes a new theoretical and empirical analysis of test prediction variance using existing datasets and does not create or release a new dataset, annotated or otherwise." | "INTRODUCES_NEW_DATASET: NO \nCONFIDENCE: HIGH \nREASON: The paper proposes a new theoretical and empirical analysis of test prediction variance using existing datasets and does not create or release a new dataset, annotated or otherwise." | true | "not_new_dataset" |
"2405.14121" | "One-shot Active Learning Based on Lewis Weight Sampling for Multiple\n Deep Models" | " Active learning (AL) for multiple target models aims to reduce labeled data\nquerying while effectively training multiple models concurrently. Existing AL\nalgorithms often rely on iterative model training, which can be computationally\nexpensive, particularly for deep models. In this paper, we propose a one-shot\nA... | "cs.LG" | 2024-10-03T00:00:00 | "benchmark" | "not_new_dataset" | false | 0.95 | "This paper proposes a new one-shot active learning method using Lewis weight sampling for training multiple deep models. The primary contribution is algorithmic - a novel approach to reduce computational expense in active learning by performing label queries without repeated model training. The paper evaluates its met... | false | "HIGH" | "The paper uses 11 existing benchmarks for evaluation and does not create, curate, or release any new dataset, annotated or otherwise." | "INTRODUCES_NEW_DATASET: NO \nCONFIDENCE: HIGH \nREASON: The paper uses 11 existing benchmarks for evaluation and does not create, curate, or release any new dataset, annotated or otherwise." | true | "not_new_dataset" |
"2505.14438" | "S2SBench: A Benchmark for Quantifying Intelligence Degradation in Speech-to-Speech Large Language Models" | "End-to-end speech large language models ((LLMs)) extend the capabilities of text-based models to directly process and generate audio tokens. However, this often leads to a decline in reasoning and generation performance compared to text input, a phenomenon referred to as intelligence degradation. To systematically eva... | "cs.SD cs.CL eess.AS" | 2025-05-21T00:00:00 | "benchmark" | "new_dataset" | true | 0.9 | "The paper introduces S2SBench, a benchmark that includes newly created diagnostic datasets for sentence continuation and commonsense reasoning under audio input. The abstract explicitly states 'It includes diagnostic datasets targeting...' and mentions 'All datasets and evaluation code are available,' indicating the d... | true | "HIGH" | "The paper presents and releases new diagnostic datasets specifically designed for evaluating intelligence degradation in speech-to-speech LLMs, including sentence continuation and commonsense reasoning tasks under audio input." | "INTRODUCES_NEW_DATASET: YES \nCONFIDENCE: HIGH \nREASON: The paper presents and releases new diagnostic datasets specifically designed for evaluating intelligence degradation in speech-to-speech LLMs, including sentence continuation and commonsense reasoning tasks under audio input." | true | "new_dataset" |
"2207.11382" | "Density-Aware Personalized Training for Risk Prediction in Imbalanced\n Medical Data" | " Medical events of interest, such as mortality, often happen at a low rate in\nelectronic medical records, as most admitted patients survive. Training models\nwith this imbalance rate (class density discrepancy) may lead to suboptimal\nprediction. Traditionally this problem is addressed through ad-hoc methods such\na... | "cs.LG" | 2022-08-02T00:00:00 | "random" | "not_new_dataset" | false | 0.95 | "The paper uses existing medical datasets (TOPCAT and MIMIC-III) for evaluation purposes but does not introduce a new dataset. The primary contribution is a methodology/framework for handling class imbalance in risk prediction. The datasets mentioned are pre-existing benchmarks used to demonstrate the method's performa... | false | "HIGH" | "The paper uses existing medical datasets (TOPCAT and MIMIC-III) without creating or releasing a new collection of data." | "INTRODUCES_NEW_DATASET: NO \nCONFIDENCE: HIGH \nREASON: The paper uses existing medical datasets (TOPCAT and MIMIC-III) without creating or releasing a new collection of data." | true | "not_new_dataset" |
"2410.01611" | "DRUPI: Dataset Reduction Using Privileged Information" | " Dataset reduction (DR) seeks to select or distill samples from large datasets\ninto smaller subsets while preserving performance on target tasks. Existing\nmethods primarily focus on pruning or synthesizing data in the same format as\nthe original dataset, typically the input data and corresponding labels.\nHowever,... | "cs.CV cs.AI cs.LG" | 2024-10-10T00:00:00 | "dataset_in_title" | "not_new_dataset" | false | 0.95 | "This paper proposes a method (DRUPI) for dataset reduction that synthesizes privileged information (feature labels, attention labels) to improve training on reduced datasets. However, it does NOT introduce a new machine learning dataset as a primary contribution. The paper: (1) evaluates exclusively on existing benchm... | false | "HIGH" | "The paper proposes a method (DRUPI) for dataset reduction using privileged information but does not create or release a new dataset; it uses existing datasets like ImageNet, CIFAR-10/100, and Tiny ImageNet." | "INTRODUCES_NEW_DATASET: NO \nCONFIDENCE: HIGH \nREASON: The paper proposes a method (DRUPI) for dataset reduction using privileged information but does not create or release a new dataset; it uses existing datasets like ImageNet, CIFAR-10/100, and Tiny ImageNet." | true | "not_new_dataset" |
"2105.02001" | "Contrastive Learning and Self-Training for Unsupervised Domain\n Adaptation in Semantic Segmentation" | " Deep convolutional neural networks have considerably improved\nstate-of-the-art results for semantic segmentation. Nevertheless, even modern\narchitectures lack the ability to generalize well to a test dataset that\noriginates from a different domain. To avoid the costly annotation of training\ndata for unseen domai... | "cs.CV cs.AI" | 2021-05-06T00:00:00 | "benchmark" | "not_new_dataset" | false | 0.95 | "This paper proposes a method (CLST) combining contrastive learning and self-training for unsupervised domain adaptation in semantic segmentation. It evaluates the approach on existing benchmark datasets (GTA5 → Cityscapes and SYNTHIA → Cityscapes), which are well-established datasets in the field. The paper does not c... | false | "HIGH" | "The paper uses existing domain adaptation benchmarks (GTA5 → Cityscapes and SYNTHIA → Cityscapes) without creating or releasing a new dataset." | "INTRODUCES_NEW_DATASET: NO \nCONFIDENCE: HIGH \nREASON: The paper uses existing domain adaptation benchmarks (GTA5 → Cityscapes and SYNTHIA → Cityscapes) without creating or releasing a new dataset." | true | "not_new_dataset" |
"2401.04138" | "Expanding Horizons in HCI Research Through LLM-Driven Qualitative\n Analysis" | " How would research be like if we still needed to \"send\" papers typed with a\ntypewriter? Our life and research environment have continually evolved, often\naccompanied by controversial opinions about new methodologies. In this paper,\nwe embrace this change by introducing a new approach to qualitative analysis in\... | "cs.HC cs.AI" | 2024-01-10T00:00:00 | "benchmark" | "new_dataset" | true | 0.95 | "The abstract explicitly states \"Through a novel dataset and benchmark, we explore LLMs' characteristics in HCI research.\" This indicates the paper creates and releases a new dataset as a primary contribution, not merely using existing data." | true | "HIGH" | "The paper explicitly states it presents a \"novel dataset and benchmark\" for evaluating LLMs in HCI research, indicating the creation and release of a new annotated data resource." | "INTRODUCES_NEW_DATASET: YES \nCONFIDENCE: HIGH \nREASON: The paper explicitly states it presents a \"novel dataset and benchmark\" for evaluating LLMs in HCI research, indicating the creation and release of a new annotated data resource." | true | "new_dataset" |
"2407.18691" | "Graph Neural Networks for Virtual Sensing in Complex Systems: Addressing\n Heterogeneous Temporal Dynamics" | " Real-time condition monitoring is crucial for the reliable and efficient\noperation of complex systems. However, relying solely on physical sensors can\nbe limited due to their cost, placement constraints, or inability to directly\nmeasure certain critical parameters. Virtual sensing addresses these\nlimitations by ... | "cs.LG cs.AI cs.CE" | 2025-03-07T00:00:00 | "recent" | "new_dataset" | true | 0.85 | "The paper explicitly mentions evaluating on 'two newly released datasets': (1) a bearing dataset with diverse load conditions for bearing load prediction, and (2) a year-long simulated dataset for predicting bridge live loads. The abstract states 'Our code and data are available,' indicating these datasets are being r... | true | "HIGH" | "The paper explicitly states that it evaluates its method using two newly released datasets—one for bearing load prediction and another for bridge live load prediction—indicating the creation and release of new data resources." | "INTRODUCES_NEW_DATASET: YES \nCONFIDENCE: HIGH \nREASON: The paper explicitly states that it evaluates its method using two newly released datasets—one for bearing load prediction and another for bridge live load prediction—indicating the creation and release of new data resources." | true | "new_dataset" |
"2403.06412" | "CLIcK: A Benchmark Dataset of Cultural and Linguistic Intelligence in\n Korean" | " Despite the rapid development of large language models (LLMs) for the Korean\nlanguage, there remains an obvious lack of benchmark datasets that test the\nrequisite Korean cultural and linguistic knowledge. Because many existing\nKorean benchmark datasets are derived from the English counterparts through\ntranslatio... | "cs.CL" | 2024-07-08T00:00:00 | "benchmark" | "new_dataset" | true | 0.99 | "This paper explicitly introduces CLIcK, a new benchmark dataset comprising 1,995 QA pairs sourced from official Korean exams and textbooks. The dataset is presented as the primary contribution - it's in the title and abstract as the main focus. The authors create new data, partition it into categories, and provide fin... | true | "HIGH" | "The paper presents CLIcK, a new benchmark dataset with 1,995 QA pairs sourced from Korean exams and textbooks, featuring fine-grained annotations for cultural and linguistic knowledge, and explicitly released as a novel resource for evaluating Korean LLMs." | "INTRODUCES_NEW_DATASET: YES \nCONFIDENCE: HIGH \nREASON: The paper presents CLIcK, a new benchmark dataset with 1,995 QA pairs sourced from Korean exams and textbooks, featuring fine-grained annotations for cultural and linguistic knowledge, and explicitly released as a novel resource for evaluating Korean LLMs." | true | "new_dataset" |
"2404.18508" | "Scalable Event-by-event Processing of Neuromorphic Sensory Signals With\n Deep State-Space Models" | " Event-based sensors are well suited for real-time processing due to their\nfast response times and encoding of the sensory data as successive temporal\ndifferences. These and other valuable properties, such as a high dynamic range,\nare suppressed when the data is converted to a frame-based format. However,\nmost cu... | "cs.LG cs.AI cs.NE" | 2024-10-10T00:00:00 | "benchmark" | "not_new_dataset" | false | 0.95 | "This paper introduces a deep state-space model architecture for processing neuromorphic event streams. While it mentions evaluating on 'several event stream benchmarks' including the 'Spiking Speech Commands task' and 'DVS128-Gestures dataset', these are existing benchmarks, not newly created datasets. The paper's pri... | false | "HIGH" | "The paper uses existing event-based benchmarks (Spiking Speech Commands and DVS128-Gestures) without creating or releasing a new dataset." | "INTRODUCES_NEW_DATASET: NO \nCONFIDENCE: HIGH \nREASON: The paper uses existing event-based benchmarks (Spiking Speech Commands and DVS128-Gestures) without creating or releasing a new dataset." | true | "not_new_dataset" |
"2310.03295" | "Can pre-trained models assist in dataset distillation?" | " Dataset Distillation (DD) is a prominent technique that encapsulates\nknowledge from a large-scale original dataset into a small synthetic dataset\nfor efficient training. Meanwhile, Pre-trained Models (PTMs) function as\nknowledge repositories, containing extensive information from the original\ndataset. This natur... | "cs.CV" | 2023-10-06T00:00:00 | "dataset_in_title" | "not_new_dataset" | false | 0.95 | "This paper studies how pre-trained models can improve dataset distillation techniques, but does not introduce a new dataset as a primary contribution. The paper's main contributions are: (1) investigating whether PTMs can assist in dataset distillation, (2) systematically studying different PTM options (initialization... | false | "HIGH" | "The paper uses existing datasets and pre-trained models to study dataset distillation but does not create or release a new dataset, annotated or otherwise." | "INTRODUCES_NEW_DATASET: NO \nCONFIDENCE: HIGH \nREASON: The paper uses existing datasets and pre-trained models to study dataset distillation but does not create or release a new dataset, annotated or otherwise." | true | "not_new_dataset" |
"2410.04456" | "SWEb: A Large Web Dataset for the Scandinavian Languages" | " This paper presents the hitherto largest pretraining dataset for the\nScandinavian languages: the Scandinavian WEb (SWEb), comprising over one\ntrillion tokens. The paper details the collection and processing pipeline, and\nintroduces a novel model-based text extractor that significantly reduces\ncomplexity in compa... | "cs.CL" | 2024-10-08T00:00:00 | "dataset_in_title" | "new_dataset" | true | 1 | "This paper clearly introduces a new dataset (SWEb) as its primary contribution. The paper creates and releases a large-scale pretraining dataset for Scandinavian languages comprising over one trillion tokens. The dataset is explicitly named in the title, detailed in the abstract as the main contribution, and the autho... | true | "HIGH" | "The paper presents SWEb, a newly created and publicly released dataset comprising over one trillion tokens for Scandinavian languages, along with a novel text extraction method and a new benchmark, fulfilling all criteria for a new dataset introduction." | "INTRODUCES_NEW_DATASET: YES \nCONFIDENCE: HIGH \nREASON: The paper presents SWEb, a newly created and publicly released dataset comprising over one trillion tokens for Scandinavian languages, along with a novel text extraction method and a new benchmark, fulfilling all criteria for a new dataset introduction." | true | "new_dataset" |
"2404.02415" | "What Are We Measuring When We Evaluate Large Vision-Language Models? An\n Analysis of Latent Factors and Biases" | " Vision-language (VL) models, pretrained on colossal image-text datasets, have\nattained broad VL competence that is difficult to evaluate. A common belief is\nthat a small number of VL skills underlie the variety of VL tests. In this\npaper, we perform a large-scale transfer learning experiment aimed at\ndiscovering... | "cs.CV" | 2024-04-04T00:00:00 | "benchmark" | "new_dataset" | true | 0.85 | "The paper explicitly presents a new dataset called OLIVE, which simulates user instructions in the wild and presents challenges dissimilar to all tested datasets. The abstract lists this as one of three main contributions alongside length bias analysis and factor analysis. The dataset is released with a GitHub link (h... | true | "HIGH" | "The paper explicitly states it presents a new dataset, OLIVE, which is released publicly and designed to simulate real-world user instructions with unique challenges not found in existing datasets." | "INTRODUCES_NEW_DATASET: YES \nCONFIDENCE: HIGH \nREASON: The paper explicitly states it presents a new dataset, OLIVE, which is released publicly and designed to simulate real-world user instructions with unique challenges not found in existing datasets." | true | "new_dataset" |
"2506.22718" | "Part Segmentation and Motion Estimation for Articulated Objects with Dynamic 3D Gaussians" | "Part segmentation and motion estimation are two fundamental problems for articulated object motion analysis. In this paper, we present a method to solve these two problems jointly from a sequence of observed point clouds of a single articulated object. The main challenge in our problem setting is that the point clouds... | "cs.CV" | 2025-08-08T00:00:00 | "recent" | "not_new_dataset" | false | 0.95 | "The paper's primary contribution is a method for joint part segmentation and motion estimation using dynamic 3D Gaussians. While it mentions 'extending existing datasets to emulate real-world scenarios,' this refers to modifying existing benchmarks for evaluation purposes, not creating a new dataset as a primary contr... | false | "HIGH" | "The paper extends existing datasets to simulate real-world scenarios with occlusions but does not create or release a new dataset." | "INTRODUCES_NEW_DATASET: NO \nCONFIDENCE: HIGH \nREASON: The paper extends existing datasets to simulate real-world scenarios with occlusions but does not create or release a new dataset." | true | "not_new_dataset" |
"2504.19149" | "A Kinematic and Kinetic Dataset of Lower Limb Joints During Obstacle Crossing in Healthy Young Adults" | "Obstacle crossing is an essential component of human locomotion, particularly for individuals with lower limb amputations who face elevated risks of imbalance and falls. While prior studies have explored this task, they often lack a comprehensive examination of kinematic and kinetic changes throughout the entire gait ... | "physics.med-ph cs.SY eess.SY" | 2025-09-24T00:00:00 | "random" | "new_dataset" | true | 0.95 | "The paper explicitly states it 'creates a novel dataset' collected from ten healthy adults performing obstacle crossing at four different heights. The dataset contains original kinematic and kinetic measurements (joint angles and torques) collected through experimental procedures. The dataset is presented as a primary... | true | "HIGH" | "The paper explicitly states that it creates and releases a novel dataset collected from ten healthy adults performing obstacle crossing at four different heights, including detailed kinematic and kinetic measurements of lower limb joints." | "INTRODUCES_NEW_DATASET: YES \nCONFIDENCE: HIGH \nREASON: The paper explicitly states that it creates and releases a novel dataset collected from ten healthy adults performing obstacle crossing at four different heights, including detailed kinematic and kinetic measurements of lower limb joints." | true | "new_dataset" |
"2501.10811" | "MusicEval: A Generative Music Dataset with Expert Ratings for Automatic\n Text-to-Music Evaluation" | " The technology for generating music from textual descriptions has seen rapid\nadvancements. However, evaluating text-to-music (TTM) systems remains a\nsignificant challenge, primarily due to the difficulty of balancing performance\nand cost with existing objective and subjective evaluation methods. In this\npaper, w... | "cs.SD eess.AS" | 2025-03-25T00:00:00 | "corpus_annotation" | "new_dataset" | true | 0.98 | "This paper clearly introduces a new dataset as its primary contribution. The authors explicitly state they 'collect MusicEval, the first generative music assessment dataset' containing 2,748 music clips generated by 31 models responding to 384 text prompts, along with 13,740 expert ratings. The dataset is named in the... | true | "HIGH" | "The paper explicitly creates and releases MusicEval, a new generative music dataset with 2,748 music clips and 13,740 expert ratings, which is novel and not derived from existing datasets." | "INTRODUCES_NEW_DATASET: YES \nCONFIDENCE: HIGH \nREASON: The paper explicitly creates and releases MusicEval, a new generative music dataset with 2,748 music clips and 13,740 expert ratings, which is novel and not derived from existing datasets." | true | "new_dataset" |
"2206.04365" | "CARLA-GeAR: a Dataset Generator for a Systematic Evaluation of Adversarial Robustness of Vision Models" | "Adversarial examples represent a serious threat for deep neural networks in several application domains and a huge amount of work has been produced to investigate them and mitigate their effects. Nevertheless, no much work has been devoted to the generation of datasets specifically designed to evaluate the adversarial... | "cs.CV" | 2025-09-16T00:00:00 | "dataset_in_title" | "new_dataset" | true | 0.95 | "This paper introduces CARLA-GeAR, a tool that generates NEW synthetic datasets specifically designed for evaluating adversarial robustness of vision models. Key indicators: (1) It creates photo-realistic synthetic datasets using the CARLA simulator, (2) these datasets are designed for a specific evaluation purpose (ad... | true | "HIGH" | "The paper presents CARLA-GeAR, a tool that generates photo-realistic synthetic datasets with adversarial patches for evaluating adversarial robustness, and explicitly releases these datasets for use as a benchmark." | "INTRODUCES_NEW_DATASET: YES \nCONFIDENCE: HIGH \nREASON: The paper presents CARLA-GeAR, a tool that generates photo-realistic synthetic datasets with adversarial patches for evaluating adversarial robustness, and explicitly releases these datasets for use as a benchmark." | true | "new_dataset" |
"2406.10304" | "Enhancing Voice Wake-Up for Dysarthria: Mandarin Dysarthria Speech\n Corpus Release and Customized System Design" | " Smart home technology has gained widespread adoption, facilitating effortless\ncontrol of devices through voice commands. However, individuals with\ndysarthria, a motor speech disorder, face challenges due to the variability of\ntheir speech. This paper addresses the wake-up word spotting (WWS) task for\ndysarthric ... | "cs.CL" | 2024-06-18T00:00:00 | "corpus_annotation" | "new_dataset" | true | 0.95 | "The paper explicitly releases a new dataset called the Mandarin Dysarthria Speech Corpus (MDSC). The abstract states 'we release the open-source Mandarin Dysarthria Speech Corpus (MDSC), a dataset designed for dysarthric individuals in home environments.' The dataset includes specific annotations (age, gender, disease... | true | "HIGH" | "The paper explicitly releases the Mandarin Dysarthria Speech Corpus (MDSC), a new, annotated dataset for dysarthric speech in Mandarin, with detailed metadata and evaluations, fulfilling all criteria for a new dataset introduction." | "INTRODUCES_NEW_DATASET: YES \nCONFIDENCE: HIGH \nREASON: The paper explicitly releases the Mandarin Dysarthria Speech Corpus (MDSC), a new, annotated dataset for dysarthric speech in Mandarin, with detailed metadata and evaluations, fulfilling all criteria for a new dataset introduction." | true | "new_dataset" |
"2305.15776" | "AUC Optimization from Multiple Unlabeled Datasets" | " Weakly supervised learning aims to empower machine learning when the perfect\nsupervision is unavailable, which has drawn great attention from researchers.\nAmong various types of weak supervision, one of the most challenging cases is\nto learn from multiple unlabeled (U) datasets with only a little knowledge of\nth... | "cs.LG" | 2023-09-18T00:00:00 | "dataset_in_title" | "not_new_dataset" | false | 0.95 | "This paper proposes a METHOD called U^m-AUC for AUC optimization from multiple unlabeled datasets. The contribution is algorithmic - converting weakly supervised data into a multi-label AUC optimization problem. The paper does not create, collect, curate, or release any new dataset. The 'U^m' refers to a learning para... | false | "HIGH" | "The paper proposes a method (U^m-AUC) for AUC optimization using multiple unlabeled datasets but does not create or release any new dataset, instead relying on existing unlabeled data." | "INTRODUCES_NEW_DATASET: NO \nCONFIDENCE: HIGH \nREASON: The paper proposes a method (U^m-AUC) for AUC optimization using multiple unlabeled datasets but does not create or release any new dataset, instead relying on existing unlabeled data." | true | "not_new_dataset" |
"2404.00401" | "How Robust are the Tabular QA Models for Scientific Tables? A Study\n using Customized Dataset" | " Question-answering (QA) on hybrid scientific tabular and textual data deals\nwith scientific information, and relies on complex numerical reasoning. In\nrecent years, while tabular QA has seen rapid progress, understanding their\nrobustness on scientific information is lacking due to absence of any benchmark\ndatase... | "cs.CL" | 2024-04-02T00:00:00 | "recent" | "new_dataset" | true | 0.95 | "The abstract explicitly states 'we propose a new dataset, \"SciTabQA\", consisting of 822 question-answer pairs from scientific tables and their descriptions.' The paper's primary contribution is creating this new benchmark dataset to address the absence of any benchmark dataset for scientific tabular QA robustness te... | true | "HIGH" | "The paper explicitly states the creation and release of a new dataset called \"SciTabQA\" with 822 question-answer pairs from scientific tables and their descriptions, fulfilling the criteria of introducing a novel annotated dataset." | "INTRODUCES_NEW_DATASET: YES \nCONFIDENCE: HIGH \nREASON: The paper explicitly states the creation and release of a new dataset called \"SciTabQA\" with 822 question-answer pairs from scientific tables and their descriptions, fulfilling the criteria of introducing a novel annotated dataset." | true | "new_dataset" |
"2509.24177" | "High-Order Progressive Trajectory Matching for Medical Image Dataset Distillation" | "Medical image analysis faces significant challenges in data sharing due to privacy regulations and complex institutional protocols. Dataset distillation offers a solution to address these challenges by synthesizing compact datasets that capture essential information from real, large medical datasets. Trajectory matchi... | "cs.CV" | 2025-09-30T00:00:00 | "dataset_in_title" | "not_new_dataset" | false | 0.95 | "This paper proposes a method (High-Order Progressive Trajectory Matching) for dataset distillation in medical imaging. While it involves synthesizing compact datasets from real medical data, the primary contribution is the trajectory matching methodology, not the creation of a new dataset. The paper evaluates on exist... | false | "HIGH" | "The paper proposes a method for dataset distillation using existing medical image datasets but does not create or release a new dataset, corpus, or annotated collection." | "INTRODUCES_NEW_DATASET: NO \nCONFIDENCE: HIGH \nREASON: The paper proposes a method for dataset distillation using existing medical image datasets but does not create or release a new dataset, corpus, or annotated collection." | true | "not_new_dataset" |
"2203.08630" | "Extended vehicle energy dataset (eVED): an enhanced large-scale dataset\n for deep learning on vehicle trip energy consumption" | " This work presents an extended version of the Vehicle Energy Dataset (VED),\nwhich is a openly released large-scale dataset for vehicle energy consumption\nanalysis. Compared with its original version, the extended VED (eVED) dataset\nis enhanced with accurate vehicle trip GPS coordinates, serving as a basis to\nass... | "cs.LG" | 2022-03-17T00:00:00 | "dataset_in_title" | "new_dataset" | true | 0.95 | "This paper introduces eVED (extended vehicle energy dataset) as a new, enhanced dataset. While it extends the original VED, the paper presents it as a substantially new resource with: (1) calibrated GPS coordinates for all records, (2) enriched features from multiple external APIs (road elevation: 12.6M records, speed... | true | "HIGH" | "The paper presents an extended and enhanced version of the original VED dataset with newly calibrated GPS coordinates and enriched attributes derived from multiple external sources, creating a novel, larger, and more detailed dataset for vehicle energy consumption analysis." | "INTRODUCES_NEW_DATASET: YES \nCONFIDENCE: HIGH \nREASON: The paper presents an extended and enhanced version of the original VED dataset with newly calibrated GPS coordinates and enriched attributes derived from multiple external sources, creating a novel, larger, and more detailed dataset for vehicle energy consump... | true | "new_dataset" |
"1908.09140" | "LANTERN: learn analysis transform network for dynamic magnetic resonance\n imaging with small dataset" | " This paper proposes to learn analysis transform network for dynamic magnetic\nresonance imaging (LANTERN) with small dataset. Integrating the strength of\nCS-MRI and deep learning, the proposed framework is highlighted in three\ncomponents: (i) The spatial and temporal domains are sparsely constrained by\nusing adap... | "eess.IV cs.CV" | 2019-08-27T00:00:00 | "dataset_in_title" | "not_new_dataset" | false | 0.95 | "This paper proposes a deep learning framework (LANTERN) for dynamic MRI reconstruction. While it mentions working with 'small dataset' and analyzing 'cardiac datasets,' the primary contribution is the proposed neural network architecture and end-to-end learning framework, not dataset creation. The experiments appear t... | false | "HIGH" | "The paper uses existing cardiac MRI datasets for evaluation and does not create or release a new dataset, corpus, or annotated collection." | "INTRODUCES_NEW_DATASET: NO \nCONFIDENCE: HIGH \nREASON: The paper uses existing cardiac MRI datasets for evaluation and does not create or release a new dataset, corpus, or annotated collection." | true | "not_new_dataset" |
"2307.10189" | "Subjective Crowd Disagreements for Subjective Data: Uncovering\n Meaningful CrowdOpinion with Population-level Learning" | " Human-annotated data plays a critical role in the fairness of AI systems,\nincluding those that deal with life-altering decisions or moderating\nhuman-created web/social media content. Conventionally, annotator disagreements\nare resolved before any learning takes place. However, researchers are\nincreasingly identi... | "cs.IR cs.CL cs.SI" | 2023-07-21T00:00:00 | "benchmark" | "not_new_dataset" | false | 0.95 | "The paper introduces CrowdOpinion, an unsupervised learning approach for handling annotator disagreements. The primary contribution is the methodology, not a dataset. The authors explicitly state they use 'five publicly available benchmark datasets' from Twitter, Gab, and Reddit, plus 'a dataset from Facebook' using e... | false | "HIGH" | "The paper uses five publicly available benchmark datasets and one dataset from Facebook, but does not create or release a new annotated dataset, corpus, or data resource." | "INTRODUCES_NEW_DATASET: NO \nCONFIDENCE: HIGH \nREASON: The paper uses five publicly available benchmark datasets and one dataset from Facebook, but does not create or release a new annotated dataset, corpus, or data resource." | true | "not_new_dataset" |
"2410.14536" | "A Hybrid Feature Fusion Deep Learning Framework for Leukemia Cancer\n Detection in Microscopic Blood Sample Using Gated Recurrent Unit and\n Uncertainty Quantification" | " Acute lymphoblastic leukemia (ALL) is the most malignant form of leukemia and\nthe most common cancer in adults and children. Traditionally, leukemia is\ndiagnosed by analyzing blood and bone marrow smears under a microscope, with\nadditional cytochemical tests for confirmation. However, these methods are\nexpensive... | "eess.IV cs.CV" | 2024-10-21T00:00:00 | "recent" | "not_new_dataset" | false | 0.95 | "The paper uses existing publicly available datasets (ALL-IDB1 and ALL-IDB2) for training and evaluation. The abstract explicitly states: 'The proposed models were trained on the publicly available datasets ALL-IDB1 and ALL-IDB2.' The primary contributions are methodological: (1) hybrid deep learning architectures comb... | false | "HIGH" | "The paper explicitly states that the models were trained on the publicly available datasets ALL-IDB1 and ALL-IDB2, and no new data collection, curation, or release is described." | "INTRODUCES_NEW_DATASET: NO \nCONFIDENCE: HIGH \nREASON: The paper explicitly states that the models were trained on the publicly available datasets ALL-IDB1 and ALL-IDB2, and no new data collection, curation, or release is described." | true | "not_new_dataset" |
"2005.05501" | "3DV: 3D Dynamic Voxel for Action Recognition in Depth Video" | " To facilitate depth-based 3D action recognition, 3D dynamic voxel (3DV) is\nproposed as a novel 3D motion representation. With 3D space voxelization, the\nkey idea of 3DV is to encode 3D motion information within depth video into a\nregular voxel set (i.e., 3DV) compactly, via temporal rank pooling. Each\navailable ... | "cs.CV" | 2020-05-13T00:00:00 | "benchmark" | "not_new_dataset" | false | 0.95 | "This paper proposes a novel 3D motion representation called '3D Dynamic Voxel (3DV)' for depth-based action recognition. The primary contribution is the 3DV method itself - a voxelization approach combined with temporal rank pooling and PointNet++ processing. The paper explicitly states it conducts 'extensive experime... | false | "HIGH" | "The paper uses four well-established benchmark datasets for evaluation and does not create or release any new data collection, corpus, or annotated dataset." | "INTRODUCES_NEW_DATASET: NO \nCONFIDENCE: HIGH \nREASON: The paper uses four well-established benchmark datasets for evaluation and does not create or release any new data collection, corpus, or annotated dataset." | true | "not_new_dataset" |
"2508.20259" | "Latent Variable Modeling for Robust Causal Effect Estimation" | "Latent variable models provide a powerful framework for incorporating and inferring unobserved factors in observational data. In causal inference, they help account for hidden factors influencing treatment or outcome, thereby addressing challenges posed by missing or unmeasured covariates. This paper proposes a new fr... | "cs.LG" | 2025-08-29T00:00:00 | "recent" | "not_new_dataset" | false | 0.95 | "This paper proposes a new methodological framework for causal inference that integrates latent variable modeling with double machine learning. The primary contribution is the framework itself, not any dataset. The mention of 'synthetic and real-world datasets' is in the context of evaluating the proposed method, not r... | false | "HIGH" | "The paper uses both synthetic and real-world datasets without creating or releasing a new annotated dataset, corpus, or data resource." | "INTRODUCES_NEW_DATASET: NO \nCONFIDENCE: HIGH \nREASON: The paper uses both synthetic and real-world datasets without creating or releasing a new annotated dataset, corpus, or data resource." | true | "not_new_dataset" |
"2412.03205" | "U-MATH: A University-Level Benchmark for Evaluating Mathematical Skills\n in LLMs" | " The current evaluation of mathematical skills in LLMs is limited, as existing\nbenchmarks are either relatively small, primarily focus on elementary and\nhigh-school problems, or lack diversity in topics. Additionally, the inclusion\nof visual elements in tasks remains largely under-explored.\n To address these gap... | "cs.CL cs.AI" | 2025-01-16T00:00:00 | "benchmark" | "new_dataset" | true | 0.95 | "This paper introduces TWO new datasets as primary contributions: (1) U-MATH, a novel benchmark of 1,100 unpublished university-level math problems collected from teaching materials, balanced across six subjects with 20% multimodal problems; and (2) $\\mu$-MATH, a dataset specifically released to evaluate LLMs' capabil... | true | "HIGH" | "The paper introduces U-MATH, a new benchmark of 1,100 unpublished university-level problems, and releases μ-MATH, a new dataset for evaluating LLMs' ability to judge solutions, both of which are novel, curated, and publicly released datasets." | "INTRODUCES_NEW_DATASET: YES \nCONFIDENCE: HIGH \nREASON: The paper introduces U-MATH, a new benchmark of 1,100 unpublished university-level problems, and releases μ-MATH, a new dataset for evaluating LLMs' ability to judge solutions, both of which are novel, curated, and publicly released datasets." | true | "new_dataset" |
"2306.05323" | "Advancing Italian Biomedical Information Extraction with\n Transformers-based Models: Methodological Insights and Multicenter Practical\n Application" | " The introduction of computerized medical records in hospitals has reduced\nburdensome activities like manual writing and information fetching. However,\nthe data contained in medical records are still far underutilized, primarily\nbecause extracting data from unstructured textual medical records takes time\nand effo... | "cs.CL cs.AI cs.LG" | 2024-01-17T00:00:00 | "corpus_annotation" | "new_dataset" | true | 0.95 | "The paper explicitly states they 'created the first Italian neuropsychiatric Named Entity Recognition dataset, PsyNIT.' This is presented as a primary contribution alongside their model development. The dataset is named, described as the first of its kind for Italian neuropsychiatric NER, and the abstract highlights t... | true | "HIGH" | "The paper explicitly states the creation and release of PsyNIT, the first Italian neuropsychiatric Named Entity Recognition dataset, which qualifies as a new annotated dataset." | "INTRODUCES_NEW_DATASET: YES \nCONFIDENCE: HIGH \nREASON: The paper explicitly states the creation and release of PsyNIT, the first Italian neuropsychiatric Named Entity Recognition dataset, which qualifies as a new annotated dataset." | true | "new_dataset" |
"1909.06897" | "Brno Urban Dataset -- The New Data for Self-Driving Agents and Mapping\n Tasks" | " Autonomous driving is a dynamically growing field of research, where quality\nand amount of experimental data is critical. Although several rich datasets are\navailable these days, the demands of researchers and technical possibilities\nare evolving. Through this paper, we bring a new dataset recorded in Brno,\nCzec... | "cs.RO cs.CV" | 2021-11-02T00:00:00 | "dataset_in_title" | "new_dataset" | true | 0.98 | "This paper explicitly introduces a new dataset as its primary contribution. The authors collected new data (350+ km of rides) in Brno, Czech Republic, with multiple sensors (cameras, LiDARs, IMU, infrared camera, RTK GNSS). The dataset is publicly released via GitHub and offers unique features (centimetre-accurate RTK... | true | "HIGH" | "The paper presents a newly recorded, publicly released dataset from Brno, Czech Republic, featuring unique sensor configurations including centimeter-accurate differential RTK GNSS and sub-millisecond timestamping, which distinguishes it from existing datasets." | "INTRODUCES_NEW_DATASET: YES \nCONFIDENCE: HIGH \nREASON: The paper presents a newly recorded, publicly released dataset from Brno, Czech Republic, featuring unique sensor configurations including centimeter-accurate differential RTK GNSS and sub-millisecond timestamping, which distinguishes it from existing datasets... | true | "new_dataset" |
"2106.07696" | "Face Age Progression With Attribute Manipulation" | " Face is one of the predominant means of person recognition. In the process of\nageing, human face is prone to many factors such as time, attributes, weather\nand other subject specific variations. The impact of these factors were not\nwell studied in the literature of face aging. In this paper, we propose a novel\nh... | "cs.CV cs.AI" | 2021-06-16T00:00:00 | "random" | "not_new_dataset" | false | 0.95 | "This paper proposes a novel GAN-based model (FAWAM) for face age progression and attribute manipulation. The abstract explicitly states they 'conduct extensive analysis in standard large scale datasets' - indicating they use existing datasets for evaluation rather than creating new ones. The primary contribution is th... | false | "HIGH" | "The paper uses standard large-scale datasets for evaluation but does not create, curate, or release a new annotated dataset, corpus, or data resource." | "INTRODUCES_NEW_DATASET: NO \nCONFIDENCE: HIGH \nREASON: The paper uses standard large-scale datasets for evaluation but does not create, curate, or release a new annotated dataset, corpus, or data resource." | true | "not_new_dataset" |
"2106.14366" | "Rail-5k: a Real-World Dataset for Rail Surface Defects Detection" | " This paper presents the Rail-5k dataset for benchmarking the performance of\nvisual algorithms in a real-world application scenario, namely the rail surface\ndefects detection task. We collected over 5k high-quality images from railways\nacross China, and annotated 1100 images with the help from railway experts to\n... | "cs.CV" | 2021-06-29T00:00:00 | "corpus_annotation" | "new_dataset" | true | 0.99 | "The paper explicitly introduces a new dataset called Rail-5k as its primary contribution. The authors collected over 5k images from railways across China and annotated 1100 images with 13 types of rail defects. The dataset is presented as a new benchmark for rail surface defect detection, with both fully-supervised an... | true | "HIGH" | "The paper explicitly describes the creation and release of a new dataset, Rail-5k, consisting of over 5,000 high-quality rail images with 1,100 annotated for 13 types of rail defects, which is not a reuse or combination of existing datasets." | "INTRODUCES_NEW_DATASET: YES \nCONFIDENCE: HIGH \nREASON: The paper explicitly describes the creation and release of a new dataset, Rail-5k, consisting of over 5,000 high-quality rail images with 1,100 annotated for 13 types of rail defects, which is not a reuse or combination of existing datasets." | true | "new_dataset" |
"2508.11144" | "CTRL Your Shift: Clustered Transfer Residual Learning for Many Small Datasets" | "Machine learning (ML) tasks often utilize large-scale data that is drawn from several distinct sources, such as different locations, treatment arms, or groups. In such settings, practitioners often desire predictions that not only exhibit good overall accuracy, but also remain reliable within each source and preserve ... | "cs.LG" | 2025-08-18T00:00:00 | "random" | "not_new_dataset" | false | 0.95 | "This paper introduces CTRL, a meta-learning method combining cross-domain residual learning with adaptive pooling/clustering. The abstract indicates the paper evaluates this method on 5 existing large-scale datasets (including an operational Swiss asylum program dataset). While the paper uses data from real-world appl... | false | "HIGH" | "The paper mentions using a dataset from the national asylum program in Switzerland but does not claim to have created or released a new dataset; it uses existing data for evaluation." | "INTRODUCES_NEW_DATASET: NO \nCONFIDENCE: HIGH \nREASON: The paper mentions using a dataset from the national asylum program in Switzerland but does not claim to have created or released a new dataset; it uses existing data for evaluation." | true | "not_new_dataset" |
"2104.03158" | "Simple Imputation Rules for Prediction with Missing Data: Contrasting\n Theoretical Guarantees with Empirical Performance" | " Missing data is a common issue in real-world datasets. This paper studies the\nperformance of impute-then-regress pipelines by contrasting theoretical and\nempirical evidence. We establish the asymptotic consistency of such pipelines\nfor a broad family of imputation methods. While common sense suggests that a\n`goo... | "stat.ML cs.LG" | 2025-01-08T00:00:00 | "corpus_annotation" | "not_new_dataset" | false | 0.9 | "This paper does NOT introduce a new machine learning dataset as a primary contribution. The abstract indicates they use a 'large corpus of synthetic, semi-real, and real datasets' for empirical evaluation, but these appear to be existing or modified datasets used to validate their theoretical findings. The paper's pri... | false | "HIGH" | "The paper uses a large corpus of synthetic, semi-real, and real datasets for evaluation but does not create or release a new dataset, annotated or otherwise." | "INTRODUCES_NEW_DATASET: NO \nCONFIDENCE: HIGH \nREASON: The paper uses a large corpus of synthetic, semi-real, and real datasets for evaluation but does not create or release a new dataset, annotated or otherwise." | true | "not_new_dataset" |
"2302.01722" | "Leveraging Contaminated Datasets to Learn Clean-Data Distribution with\n Purified Generative Adversarial Networks" | " Generative adversarial networks (GANs) are known for their strong abilities\non capturing the underlying distribution of training instances. Since the\nseminal work of GAN, many variants of GAN have been proposed. However, existing\nGANs are almost established on the assumption that the training dataset is\nclean. B... | "cs.LG" | 2023-02-06T00:00:00 | "dataset_in_title" | "not_new_dataset" | false | 0.95 | "This paper introduces PuriGAN, a new method for learning clean-data distributions from contaminated datasets. The primary contribution is the proposed framework/architecture, not a new dataset. While the method uses 'an extra dataset solely composed of contamination instances,' this appears to be part of their experim... | false | "HIGH" | "The paper uses existing datasets for experiments but does not create, curate, or release a new dataset." | "INTRODUCES_NEW_DATASET: NO \nCONFIDENCE: HIGH \nREASON: The paper uses existing datasets for experiments but does not create, curate, or release a new dataset." | true | "not_new_dataset" |
"2204.07660" | "It is Okay to Not Be Okay: Overcoming Emotional Bias in Affective Image\n Captioning by Contrastive Data Collection" | " Datasets that capture the connection between vision, language, and affection\nare limited, causing a lack of understanding of the emotional aspect of human\nintelligence. As a step in this direction, the ArtEmis dataset was recently\nintroduced as a large-scale dataset of emotional reactions to images along with\nla... | "cs.CV cs.AI cs.CL cs.LG" | 2022-04-19T00:00:00 | "corpus_annotation" | "new_dataset" | true | 0.9 | "This paper introduces ArtEmis v2.0 as a new dataset, which is a primary contribution. The authors collected 260,533 new instances using their contrastive data collection approach and combined them with the original ArtEmis dataset to create this new version. The abstract explicitly states 'The new combined dataset, du... | true | "HIGH" | "The paper presents a new, large-scale, contrastively collected dataset of 260,533 emotionally contrasting image-caption pairs, which is released alongside the original ArtEmis dataset as ArtEmis v2.0, making it a novel, curated, and annotated dataset." | "INTRODUCES_NEW_DATASET: YES \nCONFIDENCE: HIGH \nREASON: The paper presents a new, large-scale, contrastively collected dataset of 260,533 emotionally contrasting image-caption pairs, which is released alongside the original ArtEmis dataset as ArtEmis v2.0, making it a novel, curated, and annotated dataset." | true | "new_dataset" |
"2004.12752" | "The Gutenberg Dialogue Dataset" | " Large datasets are essential for neural modeling of many NLP tasks. Current\npublicly available open-domain dialogue datasets offer a trade-off between\nquality (e.g., DailyDialog) and size (e.g., Opensubtitles). We narrow this gap\nby building a high-quality dataset of 14.8M utterances in English, and smaller\ndata... | "cs.CL" | 2021-01-25T00:00:00 | "dataset_in_title" | "new_dataset" | true | 0.98 | "This paper explicitly introduces a NEW dataset as its primary contribution. The authors build a high-quality dialogue dataset of 14.8M utterances extracted from Project Gutenberg books, along with datasets in 6 other languages. The abstract states they 'narrow this gap by building a high-quality dataset' and describes... | true | "HIGH" | "The paper creates and releases a new dataset of 14.8M English utterances and smaller versions in six other languages by extracting and processing dialogues from public-domain books via Project Gutenberg, which constitutes a novel, curated, and annotated dataset not previously available." | "INTRODUCES_NEW_DATASET: YES \nCONFIDENCE: HIGH \nREASON: The paper creates and releases a new dataset of 14.8M English utterances and smaller versions in six other languages by extracting and processing dialogues from public-domain books via Project Gutenberg, which constitutes a novel, curated, and annotated datase... | true | "new_dataset" |
"2212.08801" | "Comparison of Model-Free and Model-Based Learning-Informed Planning for\n PointGoal Navigation" | " In recent years several learning approaches to point goal navigation in\npreviously unseen environments have been proposed. They vary in the\nrepresentations of the environments, problem decomposition, and experimental\nevaluation. In this work, we compare the state-of-the-art Deep Reinforcement\nLearning based appr... | "cs.RO cs.CV" | 2022-12-20T00:00:00 | "random" | "not_new_dataset" | false | 0.95 | "The paper does not introduce a new dataset. It explicitly uses an existing benchmark dataset (Matterport3D) with the Habitat Simulator to compare model-free (DD-PPO, ANS) and model-based (POMDP) approaches for point goal navigation. The primary contributions are methodological comparisons and adaptations of existing f... | false | "HIGH" | "The paper uses the existing Matterport3D dataset and Habitat Simulator for evaluation without creating or releasing any new data collection, annotation, or benchmark." | "INTRODUCES_NEW_DATASET: NO \nCONFIDENCE: HIGH \nREASON: The paper uses the existing Matterport3D dataset and Habitat Simulator for evaluation without creating or releasing any new data collection, annotation, or benchmark." | true | "not_new_dataset" |
"2404.12299" | "Simultaneous Interpretation Corpus Construction by Large Language Models\n in Distant Language Pair" | " In Simultaneous Machine Translation (SiMT) systems, training with a\nsimultaneous interpretation (SI) corpus is an effective method for achieving\nhigh-quality yet low-latency systems. However, it is very challenging to curate\nsuch a corpus due to limitations in the abilities of annotators, and hence,\nexisting SI ... | "cs.CL cs.AI cs.LG cs.SD eess.AS" | 2024-04-19T00:00:00 | "corpus_annotation" | "new_dataset" | true | 0.95 | "This paper introduces LLM-SI-Corpus as a new dataset for Simultaneous Machine Translation. The authors propose a method to convert existing speech translation corpora into interpretation-style data, creating a new resource that didn't exist before. The abstract explicitly states 'The LLM-SI-Corpus is available at...' ... | true | "HIGH" | "The paper creates and releases a new annotated corpus (LLM-SI-Corpus) by transforming existing speech translation data into interpretation-style data using large language models, fulfilling the criteria of a novel, curated, and publicly available dataset." | "INTRODUCES_NEW_DATASET: YES \nCONFIDENCE: HIGH \nREASON: The paper creates and releases a new annotated corpus (LLM-SI-Corpus) by transforming existing speech translation data into interpretation-style data using large language models, fulfilling the criteria of a novel, curated, and publicly available dataset." | true | "new_dataset" |
"2109.05539" | "BioLCNet: Reward-modulated Locally Connected Spiking Neural Networks" | " Brain-inspired computation and information processing alongside compatibility\nwith neuromorphic hardware have made spiking neural networks (SNN) a promising\nmethod for solving learning tasks in machine learning (ML). Spiking neurons are\nonly one of the requirements for building a bio-plausible learning model.\nNe... | "cs.NE cs.CV cs.LG q-bio.NC" | 2022-07-08T00:00:00 | "random" | "not_new_dataset" | false | 0.95 | "The paper proposes BioLCNet, a reward-modulated locally connected spiking neural network architecture. For evaluation, it uses existing benchmark datasets (MNIST and XOR MNIST). The paper does not create, collect, curate, or release any new dataset as a primary contribution. The datasets are used solely for evaluating... | false | "HIGH" | "The paper evaluates its model on existing datasets (MNIST and XOR MNIST) without creating or releasing any new data collection, annotation, or benchmark." | "INTRODUCES_NEW_DATASET: NO \nCONFIDENCE: HIGH \nREASON: The paper evaluates its model on existing datasets (MNIST and XOR MNIST) without creating or releasing any new data collection, annotation, or benchmark." | true | "not_new_dataset" |
"2312.15247" | "Prompt-Propose-Verify: A Reliable Hand-Object-Interaction Data\n Generation Framework using Foundational Models" | " Diffusion models when conditioned on text prompts, generate realistic-looking\nimages with intricate details. But most of these pre-trained models fail to\ngenerate accurate images when it comes to human features like hands, teeth,\netc. We hypothesize that this inability of diffusion models can be overcome\nthrough... | "cs.CV cs.AI" | 2023-12-27T00:00:00 | "corpus_annotation" | "new_dataset" | true | 0.95 | "The paper explicitly states they 'collect a well annotated hand-object interaction synthetic dataset' using their Prompt-Propose-Verify framework. The dataset is a primary contribution - they curate new synthetic data specifically for hand-object interaction image generation, annotate it, and use it to finetune a stab... | true | "HIGH" | "The paper explicitly states it collects and curates a new, well-annotated hand-object interaction synthetic dataset using the Prompt-Propose-Verify framework and releases it for finetuning and evaluation." | "INTRODUCES_NEW_DATASET: YES \nCONFIDENCE: HIGH \nREASON: The paper explicitly states it collects and curates a new, well-annotated hand-object interaction synthetic dataset using the Prompt-Propose-Verify framework and releases it for finetuning and evaluation." | true | "new_dataset" |
"2204.08696" | "CTCNet: A CNN-Transformer Cooperation Network for Face Image\n Super-Resolution" | " Recently, deep convolution neural networks (CNNs) steered face\nsuper-resolution methods have achieved great progress in restoring degraded\nfacial details by jointly training with facial priors. However, these methods\nhave some obvious limitations. On the one hand, multi-task joint learning\nrequires additional ma... | "cs.CV" | 2023-03-24T00:00:00 | "random" | "not_new_dataset" | false | 0.95 | "This paper proposes a new network architecture (CTCNet) with novel modules (LGCM, FSAU, FRM, MFFU) for face super-resolution. The abstract explicitly states 'Extensive evaluations on various datasets' - indicating they use existing benchmarks rather than creating new ones. The primary contributions are methodological ... | false | "HIGH" | "The paper evaluates its method on various existing datasets without creating or releasing a new dataset." | "INTRODUCES_NEW_DATASET: NO \nCONFIDENCE: HIGH \nREASON: The paper evaluates its method on various existing datasets without creating or releasing a new dataset." | true | "not_new_dataset" |
"2209.05654" | "ComplETR: Reducing the cost of annotations for object detection in dense\n scenes with vision transformers" | " Annotating bounding boxes for object detection is expensive, time-consuming,\nand error-prone. In this work, we propose a DETR based framework called\nComplETR that is designed to explicitly complete missing annotations in\npartially annotated dense scene datasets. This reduces the need to annotate\nevery object ins... | "cs.CV" | 2022-09-14T00:00:00 | "corpus_annotation" | "not_new_dataset" | false | 0.95 | "This paper introduces ComplETR, a DETR-based framework designed to complete missing annotations in partially annotated datasets. The primary contribution is the method/framework itself, not a new dataset. The abstract indicates they evaluate on 'multiple dense scene datasets' (existing benchmarks) and compare against ... | false | "HIGH" | "The paper proposes a method (ComplETR) to complete missing annotations in existing dense scene datasets but does not create or release a new dataset." | "INTRODUCES_NEW_DATASET: NO \nCONFIDENCE: HIGH \nREASON: The paper proposes a method (ComplETR) to complete missing annotations in existing dense scene datasets but does not create or release a new dataset." | true | "not_new_dataset" |
"2303.01475" | "Over-training with Mixup May Hurt Generalization" | " Mixup, which creates synthetic training instances by linearly interpolating\nrandom sample pairs, is a simple and yet effective regularization technique to\nboost the performance of deep models trained with SGD. In this work, we report\na previously unobserved phenomenon in Mixup training: on a number of standard\nd... | "cs.LG" | 2023-03-03T00:00:00 | "benchmark" | "not_new_dataset" | false | 0.95 | "This paper studies the Mixup regularization technique and its generalization behavior over training epochs. It uses 'standard datasets' and 'benchmark datasets' for experiments but does not create, collect, curate, or release any new dataset. The paper's contributions are theoretical analysis of why Mixup can cause U-... | false | "HIGH" | "The paper uses existing benchmark datasets to study the U-shaped generalization curve in Mixup training and does not create, curate, or release any new dataset." | "INTRODUCES_NEW_DATASET: NO \nCONFIDENCE: HIGH \nREASON: The paper uses existing benchmark datasets to study the U-shaped generalization curve in Mixup training and does not create, curate, or release any new dataset." | true | "not_new_dataset" |
"2412.07155" | "Annotation Techniques for Judo Combat Phase Classification from\n Tournament Footage" | " This paper presents a semi-supervised approach to extracting and analyzing\ncombat phases in judo tournaments using live-streamed footage. The objective is\nto automate the annotation and summarization of live streamed judo matches. We\ntrain models that extract relevant entities and classify combat phases from\nfix... | "cs.CV cs.MM" | 2024-12-11T00:00:00 | "corpus_annotation" | "not_new_dataset" | false | 0.85 | "The paper proposes a semi-supervised approach for classifying judo combat phases using transfer learning. While it mentions evaluating on 'a dataset of 19 thirty-second judo clips,' this appears to be a small evaluation set used to validate their method rather than a new dataset being introduced as a primary contribut... | false | "HIGH" | "The paper uses a dataset of 19 thirty-second judo clips but does not describe creating or releasing a new annotated dataset; it only evaluates on existing footage." | "INTRODUCES_NEW_DATASET: NO \nCONFIDENCE: HIGH \nREASON: The paper uses a dataset of 19 thirty-second judo clips but does not describe creating or releasing a new annotated dataset; it only evaluates on existing footage." | true | "not_new_dataset" |
"2508.19182" | "SoccerNet 2025 Challenges Results" | "The SoccerNet 2025 Challenges mark the fifth annual edition of the SoccerNet open benchmarking effort, dedicated to advancing computer vision research in football video understanding. This year's challenges span four vision-based tasks: (1) Team Ball Action Spotting, focused on detecting ball-related actions in footba... | "cs.CV" | 2025-08-27T00:00:00 | "benchmark" | "not_new_dataset" | false | 0.95 | "This paper is a challenge report describing the fifth annual SoccerNet 2025 Challenges. While it mentions that participants were 'provided with large-scale annotated datasets,' these are existing SoccerNet datasets (part of an ongoing benchmarking effort since at least 2021), not newly created datasets. The paper's pr... | false | "HIGH" | "The paper reports on the results of the SoccerNet 2025 Challenges, which use existing annotated datasets rather than introducing a new one." | "INTRODUCES_NEW_DATASET: NO \nCONFIDENCE: HIGH \nREASON: The paper reports on the results of the SoccerNet 2025 Challenges, which use existing annotated datasets rather than introducing a new one." | true | "not_new_dataset" |
"2107.13054" | "Exceeding the Limits of Visual-Linguistic Multi-Task Learning" | " By leveraging large amounts of product data collected across hundreds of live\ne-commerce websites, we construct 1000 unique classification tasks that share\nsimilarly-structured input data, comprised of both text and images. These\nclassification tasks focus on learning the product hierarchy of different\ne-commerc... | "cs.AI cs.CL cs.CV cs.LG" | 2021-07-29T00:00:00 | "random" | "new_dataset" | true | 0.95 | "The paper introduces a new dataset as a primary contribution. The authors collect original product data from hundreds of e-commerce websites and construct 1000 unique classification tasks from this data. The abstract explicitly mentions an 'initial 100-task dataset' and presents experiments revealing best practices fo... | true | "HIGH" | "The paper explicitly constructs a new dataset comprising 1000 unique classification tasks using product data from hundreds of e-commerce websites, which is curated and released as a novel benchmark for large-scale multi-task learning." | "INTRODUCES_NEW_DATASET: YES \nCONFIDENCE: HIGH \nREASON: The paper explicitly constructs a new dataset comprising 1000 unique classification tasks using product data from hundreds of e-commerce websites, which is curated and released as a novel benchmark for large-scale multi-task learning." | true | "new_dataset" |
"2405.13378" | "FedCache 2.0: Federated Edge Learning with Knowledge Caching and Dataset\n Distillation" | " Federated Edge Learning (FEL) has emerged as a promising approach for\nenabling edge devices to collaboratively train machine learning models while\npreserving data privacy. Despite its advantages, practical FEL deployment faces\nsignificant challenges related to device constraints and device-server\ninteractions, n... | "cs.LG" | 2024-10-15T00:00:00 | "dataset_in_title" | "not_new_dataset" | false | 0.95 | "This paper introduces FedCache 2.0, a novel personalized federated learning architecture that combines dataset distillation with knowledge caching. The primary contributions are: (1) a new FEL architecture design, (2) a server-side knowledge cache mechanism, and (3) a device-centric cache sampling strategy. The paper ... | false | "HIGH" | "The paper uses five existing datasets across different modalities but does not create, curate, or release any new dataset, annotated or otherwise." | "INTRODUCES_NEW_DATASET: NO \nCONFIDENCE: HIGH \nREASON: The paper uses five existing datasets across different modalities but does not create, curate, or release any new dataset, annotated or otherwise." | true | "not_new_dataset" |
"2201.00814" | "Vision Transformer Slimming: Multi-Dimension Searching in Continuous\n Optimization Space" | " This paper explores the feasibility of finding an optimal sub-model from a\nvision transformer and introduces a pure vision transformer slimming (ViT-Slim)\nframework. It can search a sub-structure from the original model end-to-end\nacross multiple dimensions, including the input tokens, MHSA and MLP modules\nwith ... | "cs.CV cs.AI cs.LG" | 2022-04-26T00:00:00 | "random" | "not_new_dataset" | false | 0.95 | "This paper introduces a framework (ViT-Slim) for vision transformer slimming and model compression, not a dataset. The paper proposes a method for finding optimal sub-models through multi-dimension searching in continuous optimization space. While it mentions evaluating on ImageNet and downstream datasets, these are e... | false | "HIGH" | "The paper proposes a method for model compression and architecture search in Vision Transformers but only uses existing datasets like ImageNet and downstream datasets without creating or releasing a new data collection." | "INTRODUCES_NEW_DATASET: NO \nCONFIDENCE: HIGH \nREASON: The paper proposes a method for model compression and architecture search in Vision Transformers but only uses existing datasets like ImageNet and downstream datasets without creating or releasing a new data collection." | true | "not_new_dataset" |
"2206.07699" | "Write and Paint: Generative Vision-Language Models are Unified Modal\n Learners" | " Recent advances in vision-language pre-training have pushed the\nstate-of-the-art on various vision-language tasks, making machines more capable\nof multi-modal writing (image-to-text generation) and painting (text-to-image\ngeneration). However, few studies investigate if these two essential\ncapabilities can be le... | "cs.CV cs.CL cs.LG" | 2023-02-20T00:00:00 | "benchmark" | "not_new_dataset" | false | 0.95 | "This paper proposes a new unified vision-language model called DaVinci with a symmetric generative pre-training framework. The primary contributions are: (1) the model architecture, (2) the prefix multi-modal modeling training objective, and (3) benchmarking results across 27 tasks. While the paper mentions using 'pre... | false | "HIGH" | "The paper uses existing datasets for pre-training and benchmarking, and does not create or release a new annotated dataset, corpus, or data resource." | "INTRODUCES_NEW_DATASET: NO \nCONFIDENCE: HIGH \nREASON: The paper uses existing datasets for pre-training and benchmarking, and does not create or release a new annotated dataset, corpus, or data resource." | true | "not_new_dataset" |
"2507.17978" | "MeAJOR Corpus: A Multi-Source Dataset for Phishing Email Detection" | "Phishing emails continue to pose a significant threat to cybersecurity by exploiting human vulnerabilities through deceptive content and malicious payloads. While Machine Learning (ML) models are effective at detecting phishing threats, their performance largely relies on the quality and diversity of the training data... | "cs.CR cs.AI cs.HC" | 2025-11-07T00:00:00 | "corpus_annotation" | "new_dataset" | true | 0.98 | "The paper explicitly introduces MeAJOR Corpus as a new dataset that integrates 135,894 samples from multiple open-source repositories. The dataset is the primary contribution (stated in title as 'Corpus' and emphasized throughout abstract). Key indicators: 'presents...a novel, multi-source phishing email dataset', 'in... | true | "HIGH" | "The paper presents MeAJOR Corpus, a newly created and released multi-source dataset integrating 135,894 phishing and legitimate email samples with engineered features, explicitly designed as a novel benchmark for phishing detection research." | "INTRODUCES_NEW_DATASET: YES \nCONFIDENCE: HIGH \nREASON: The paper presents MeAJOR Corpus, a newly created and released multi-source dataset integrating 135,894 phishing and legitimate email samples with engineered features, explicitly designed as a novel benchmark for phishing detection research." | true | "new_dataset" |
"1902.10526" | "DiscoFuse: A Large-Scale Dataset for Discourse-Based Sentence Fusion" | " Sentence fusion is the task of joining several independent sentences into a\nsingle coherent text. Current datasets for sentence fusion are small and\ninsufficient for training modern neural models. In this paper, we propose a\nmethod for automatically-generating fusion examples from raw text and present\nDiscoFuse,... | "cs.CL" | 2019-03-19T00:00:00 | "dataset_in_title" | "new_dataset" | true | 0.95 | "The paper explicitly 'presents DiscoFuse, a large scale dataset for discourse-based sentence fusion' as a primary contribution. The authors developed rules to identify discourse phenomena and decompose text into fusion examples, then applied this to raw text collections (Wikipedia, Sports articles) to generate 60 mill... | true | "HIGH" | "The paper presents DiscoFuse, a large-scale, automatically generated dataset with 60 million fusion examples annotated with discourse information, which is newly created and released for discourse-based sentence fusion." | "INTRODUCES_NEW_DATASET: YES \nCONFIDENCE: HIGH \nREASON: The paper presents DiscoFuse, a large-scale, automatically generated dataset with 60 million fusion examples annotated with discourse information, which is newly created and released for discourse-based sentence fusion." | true | "new_dataset" |
"2207.10141" | "AudioScopeV2: Audio-Visual Attention Architectures for Calibrated\n Open-Domain On-Screen Sound Separation" | " We introduce AudioScopeV2, a state-of-the-art universal audio-visual\non-screen sound separation system which is capable of learning to separate\nsounds and associate them with on-screen objects by looking at in-the-wild\nvideos. We identify several limitations of previous work on audio-visual\non-screen sound separ... | "cs.SD cs.CV eess.AS" | 2022-07-22T00:00:00 | "corpus_annotation" | "new_dataset" | true | 0.85 | "The paper introduces a new dataset as a significant contribution. While the primary focus is the AudioScopeV2 system, the authors explicitly state they 'collected new human annotations of onscreen sounds from a large database of in-the-wild videos (YFCC100M)' and describe 'This new dataset is more diverse and challeng... | true | "HIGH" | "The paper explicitly states that it collected new human annotations of onscreen sounds from YFCC100M, creating a more diverse and challenging dataset for training and evaluation." | "INTRODUCES_NEW_DATASET: YES \nCONFIDENCE: HIGH \nREASON: The paper explicitly states that it collected new human annotations of onscreen sounds from YFCC100M, creating a more diverse and challenging dataset for training and evaluation." | true | "new_dataset" |
"2301.06057" | "Hawk: An Industrial-strength Multi-label Document Classifier" | " There are a plethora of methods and algorithms that solve the classical\nmulti-label document classification. However, when it comes to deployment and\nusage in an industry setting, most, if not all the contemporary approaches fail\nto address some of the vital aspects or requirements of an ideal solution: i.\nabili... | "cs.CL cs.LG" | 2023-01-18T00:00:00 | "benchmark" | "not_new_dataset" | false | 0.95 | "The paper proposes a neural network architecture (hydranet-like) for multi-label document classification and evaluates it on existing benchmark datasets (Web of Science - 5763, Web of Science - 11967, BBC Sports, BBC News). The primary contribution is the architecture design addressing industrial requirements (catastr... | false | "HIGH" | "The paper benchmarks its model on existing datasets (Web of Science - 5763, Web of Science - 11967, BBC Sports, and BBC News) without creating or releasing any new data collection." | "INTRODUCES_NEW_DATASET: NO \nCONFIDENCE: HIGH \nREASON: The paper benchmarks its model on existing datasets (Web of Science - 5763, Web of Science - 11967, BBC Sports, and BBC News) without creating or releasing any new data collection." | true | "not_new_dataset" |
"2303.05345" | "TGDataset: Collecting and Exploring the Largest Telegram Channels\n Dataset" | " Telegram is one of the most popular instant messaging apps in today's digital\nage. In addition to providing a private messaging service, Telegram, with its\nchannels, represents a valid medium for rapidly broadcasting content to a large\naudience (COVID-19 announcements), but, unfortunately, also for disseminating\... | "cs.CY" | 2025-03-04T00:00:00 | "dataset_in_title" | "new_dataset" | true | 0.98 | "This paper explicitly introduces TGDataset as a new dataset containing 120,979 Telegram channels and over 400 million messages. The dataset is clearly the primary contribution - the paper is named after it, describes the data collection process, analyzes the data, and releases both the raw dataset and analysis scripts... | true | "HIGH" | "The paper presents TGDataset, a newly collected and released dataset comprising 120,979 Telegram channels and over 400 million messages, which is explicitly described as the largest such collection to date and includes original data curation and release." | "INTRODUCES_NEW_DATASET: YES \nCONFIDENCE: HIGH \nREASON: The paper presents TGDataset, a newly collected and released dataset comprising 120,979 Telegram channels and over 400 million messages, which is explicitly described as the largest such collection to date and includes original data curation and release." | true | "new_dataset" |
"2403.07012" | "A PID-Controlled Non-Negative Tensor Factorization Model for Analyzing\n Missing Data in NILM" | " With the growing demand for energy and increased environmental awareness,\nNon-Intrusive Load Monitoring (NILM) has become an essential tool in smart grid\nand energy management. By analyzing total power load data, NILM infers the\nenergy usage of individual appliances without the need for separate sensors,\nenablin... | "cs.LG" | 2024-10-22T00:00:00 | "recent" | "not_new_dataset" | false | 0.95 | "This paper introduces a new model (PNLF - Proportional-Integral-Derivative Controlled Non-Negative Latent Factorization of Tensor) for handling missing data in Non-Intrusive Load Monitoring (NILM). The abstract mentions 'NILM datasets' only in the context of existing data suffering from sensor failures and missing val... | false | "HIGH" | "The paper proposes a new model (PNLF) for handling missing data in existing NILM datasets but does not create or release a new dataset, corpus, or annotated collection." | "INTRODUCES_NEW_DATASET: NO \nCONFIDENCE: HIGH \nREASON: The paper proposes a new model (PNLF) for handling missing data in existing NILM datasets but does not create or release a new dataset, corpus, or annotated collection." | true | "not_new_dataset" |
"2502.10881" | "CiteCheck: Towards Accurate Citation Faithfulness Detection" | " Citation faithfulness detection is critical for enhancing retrieval-augmented\ngeneration (RAG) systems, yet large-scale Chinese datasets for this task are\nscarce. Existing methods face prohibitive costs due to the need for manually\nannotated negative samples. To address this, we introduce the first large-scale\nC... | "cs.CL" | 2025-02-18T00:00:00 | "random" | "new_dataset" | true | 0.95 | "The paper explicitly introduces CiteCheck as a new large-scale Chinese dataset for citation faithfulness detection. Key indicators: (1) 'we introduce the first large-scale Chinese dataset CiteCheck' - clear statement of dataset creation; (2) 'constructed via a cost-effective approach using two-stage manual annotation'... | true | "HIGH" | "The paper explicitly states it introduces CiteCheck, the first large-scale Chinese dataset for citation faithfulness detection, which is constructed via a novel two-stage manual annotation approach and publicly released." | "INTRODUCES_NEW_DATASET: YES \nCONFIDENCE: HIGH \nREASON: The paper explicitly states it introduces CiteCheck, the first large-scale Chinese dataset for citation faithfulness detection, which is constructed via a novel two-stage manual annotation approach and publicly released." | true | "new_dataset" |
"2503.07603" | "Should VLMs be Pre-trained with Image Data?" | " Pre-trained LLMs that are further trained with image data perform well on\nvision-language tasks. While adding images during a second training phase\neffectively unlocks this capability, it is unclear how much of a gain or loss\nthis two-step pipeline gives over VLMs which integrate images earlier into the\ntraining... | "cs.CV" | 2025-03-11T00:00:00 | "recent" | "not_new_dataset" | false | 0.95 | "This paper does not introduce a new dataset. The abstract describes experiments using 'various datasets' (existing ones) to study training strategies for Vision-Language Models. The primary contribution is empirical findings about when to introduce visual tokens during pre-training, not dataset creation. The paper eva... | false | "HIGH" | "The paper uses existing datasets for training and evaluation but does not create or release a new dataset, corpus, or annotated collection of data." | "INTRODUCES_NEW_DATASET: NO \nCONFIDENCE: HIGH \nREASON: The paper uses existing datasets for training and evaluation but does not create or release a new dataset, corpus, or annotated collection of data." | true | "not_new_dataset" |
"2508.16994" | "GRADE: Generating multi-hop QA and fine-gRAined Difficulty matrix for RAG Evaluation" | "Retrieval-Augmented Generation (RAG) systems are widely adopted in knowledge-intensive NLP tasks, but current evaluations often overlook the structural complexity and multi-step reasoning required in real-world scenarios. These benchmarks overlook key factors such as the interaction between retrieval difficulty and re... | "cs.CL cs.AI" | 2025-12-16T00:00:00 | "benchmark" | "new_dataset" | true | 0.95 | "The paper explicitly states 'We construct a synthetic multi-hop QA dataset from factual news articles' - this is a clear statement of creating a new dataset as a primary contribution. The authors collect data from factual news articles, extract knowledge graphs, augment them through semantic clustering to recover miss... | true | "HIGH" | "The paper explicitly constructs a synthetic multi-hop QA dataset from factual news articles using knowledge graphs and semantic clustering, and releases it as part of the GRADE framework for fine-grained RAG evaluation." | "INTRODUCES_NEW_DATASET: YES \nCONFIDENCE: HIGH \nREASON: The paper explicitly constructs a synthetic multi-hop QA dataset from factual news articles using knowledge graphs and semantic clustering, and releases it as part of the GRADE framework for fine-grained RAG evaluation." | true | "new_dataset" |
"2507.07135" | "FACap: A Large-scale Fashion Dataset for Fine-grained Composed Image Retrieval" | "The composed image retrieval (CIR) task is to retrieve target images given a reference image and a modification text. Recent methods for CIR leverage large pretrained vision-language models (VLMs) and achieve good performance on general-domain concepts like color and texture. However, they still struggle with applicat... | "cs.LG" | 2025-07-11T00:00:00 | "dataset_in_title" | "new_dataset" | true | 0.95 | "The paper explicitly introduces FACap as a new large-scale fashion-domain composed image retrieval dataset. The abstract states: 'To address these challenges, we introduce FACap, a large-scale, automatically constructed fashion-domain CIR dataset.' The dataset is constructed using a two-stage annotation pipeline with ... | true | "HIGH" | "The paper presents FACap, a large-scale, automatically constructed fashion-domain CIR dataset with detailed and relevant annotations generated via a novel two-stage pipeline using VLMs and LLMs, which is explicitly released and used to train and evaluate a new model." | "INTRODUCES_NEW_DATASET: YES \nCONFIDENCE: HIGH \nREASON: The paper presents FACap, a large-scale, automatically constructed fashion-domain CIR dataset with detailed and relevant annotations generated via a novel two-stage pipeline using VLMs and LLMs, which is explicitly released and used to train and evaluate a new... | true | "new_dataset" |
"1705.04513" | "GRID Storage Optimization in Transparent and User-Friendly Way for LHCb\n Datasets" | " The LHCb collaboration is one of the four major experiments at the Large\nHadron Collider at CERN. Many petabytes of data are produced by the detectors\nand Monte-Carlo simulations. The LHCb Grid interware LHCbDIRAC is used to make\ndata available to all collaboration members around the world. The data is\nreplicate... | "cs.DC" | 2017-12-06T00:00:00 | "dataset_in_title" | "not_new_dataset" | false | 0.95 | "This paper presents a new approach/methodology for data replication and distribution strategy based on data popularity prediction. While it uses machine learning techniques and time series analysis on existing LHCb data access history and metadata, the primary contribution is the optimization approach itself, not the ... | false | "HIGH" | "The paper focuses on optimizing data replication and distribution in the LHCb Grid using existing data access history and metadata, without creating or releasing a new dataset, annotated or otherwise." | "INTRODUCES_NEW_DATASET: NO \nCONFIDENCE: HIGH \nREASON: The paper focuses on optimizing data replication and distribution in the LHCb Grid using existing data access history and metadata, without creating or releasing a new dataset, annotated or otherwise." | true | "not_new_dataset" |
"2107.01327" | "VinDr-RibCXR: A Benchmark Dataset for Automatic Segmentation and\n Labeling of Individual Ribs on Chest X-rays" | " We introduce a new benchmark dataset, namely VinDr-RibCXR, for automatic\nsegmentation and labeling of individual ribs from chest X-ray (CXR) scans. The\nVinDr-RibCXR contains 245 CXRs with corresponding ground truth annotations\nprovided by human experts. A set of state-of-the-art segmentation models are\ntrained o... | "eess.IV cs.CV" | 2021-07-06T00:00:00 | "benchmark" | "new_dataset" | true | 1 | "This paper explicitly introduces a new benchmark dataset (VinDr-RibCXR) as its primary contribution. The abstract states 'We introduce a new benchmark dataset' and describes collecting 245 chest X-rays with expert annotations. The dataset is released for the research community to serve as a benchmark and baseline for ... | true | "HIGH" | "The paper explicitly states that VinDr-RibCXR is a new benchmark dataset with 245 chest X-rays and expert-annotated ground truth for individual rib segmentation and labeling, which is newly created and released." | "INTRODUCES_NEW_DATASET: YES \nCONFIDENCE: HIGH \nREASON: The paper explicitly states that VinDr-RibCXR is a new benchmark dataset with 245 chest X-rays and expert-annotated ground truth for individual rib segmentation and labeling, which is newly created and released." | true | "new_dataset" |
"2601.13304" | "CausalSpatial: A Benchmark for Object-Centric Causal Spatial Reasoning" | "Humans can look at a static scene and instantly predict what happens next -- will moving this object cause a collision? We call this ability Causal Spatial Reasoning. However, current multimodal large language models (MLLMs) cannot do this, as they remain largely restricted to static spatial perception, struggling to ... | "cs.CV" | 2026-01-21T00:00:00 | "recent" | "new_dataset" | true | 0.95 | "The paper explicitly introduces CausalSpatial as a new diagnostic benchmark for evaluating models on causal spatial reasoning. The abstract states 'We introduce CausalSpatial, a diagnostic benchmark' and mentions they 'make the dataset and code publicly available.' This is a primary contribution - a newly created benc... | true | "HIGH" | "The paper introduces CausalSpatial, a new diagnostic benchmark with four novel tasks and a curated dataset of 3D scene dynamics designed to evaluate causal spatial reasoning, which is explicitly released alongside the code." | "INTRODUCES_NEW_DATASET: YES \nCONFIDENCE: HIGH \nREASON: The paper introduces CausalSpatial, a new diagnostic benchmark with four novel tasks and a curated dataset of 3D scene dynamics designed to evaluate causal spatial reasoning, which is explicitly released alongside the code." | true | "new_dataset" |
End of preview. Expand in Data Studio
No dataset card yet
- Downloads last month
- 37