Loading...
Loading...
Text classification, translation, QA, summarization, dialogue, sentiment analysis, language modeling, text corpora
46,168 datasets
UltraBeacon is a high-quality English question-answering dataset prepared by Dl26 for text-generation and chat-style model training. The dataset is converted from TriviaQA into a consistent assistant format with prompt, response, and messages fields. It is intended for supervised fine-tuning of general assistant models that need strong question answering behavior.
UltraAtlas is a high-quality English question-answering dataset prepared by Dl26 for text-generation and chat-style model training. The dataset is converted from SQuAD v2 into a consistent assistant format with prompt, response, and messages fields. It is intended for supervised fine-tuning of general assistant models that need strong question answering behavior.
UltraBeacon is a high-quality English question-answering dataset prepared by developer Dl26. The dataset is converted from TriviaQA into a consistent assistant format with prompt, response, and messages fields. It is intended for supervised fine-tuning of general assistant models that need strong question answering behavior.
Feeding guild counts aggregated from species-level infauna data collected during a 2008 marine survey on the Lord Howe Island shelf. Geoscience Australia acquired the data using the Research Vessel Southern Surveyor, merging bathymetry to cover a 1034 sq km grid. The underlying species data was aggregated on October 27, 2010.
18.9 MB of experimental data in XLSX format, supporting research on photoelectrochemical water splitting. The dataset was authored by Mengya Yang and last updated on 2026-05-14. It likely contains measurements related to photogenerated hole dynamics in modified hematite photoanodes.
Five distinct acoustic seabed facies were defined for the Great Australian Bight based on 3.5kHz echo-sounding and seismic profiles from a 2000/2001 study. This geospatial dataset, produced by Geoscience Australia, delineates four major geomorphological features—continental shelf, terraces, slope, and rise—to support regional marine planning. The data captures the distribution of layered, disturbed, and topographically varied sediments across the shelf, slope, and abyssal plain.
A 5-meter resolution multibeam sonar survey of the seabed near Forster Pacific Palms, acquired by the New South Wales government between March 2019 and August 2022. The dataset provides 32-bit floating point geotiff files of bathymetry and backscatter, processed through Hypack, R2Sonic GUI, POSView, POSPac, Qimera, and FMGT software. It was created to establish a baseline and map seabed type distribution as part of the SeabedNSW program.
A 2017 survey collected 285 seabed sediment samples from inner Darwin Harbour and shallow water areas in and around Bynoe Harbour. The project was led by the Northern Territory Government and involved Geoscience Australia and the Australian Institute of Marine Science, funded by the INPEX-led Ichthys LNG Project. The data, including sediment samples and seagrass observations, is intended for creating baseline habitat maps to support marine resource management.
Global one-degree gridded data provides daily atmospheric profiles and surface properties derived from microwave observations. The dataset contains retrieved geophysical parameters, including temperature and water vapor profiles, surface skin temperature, and sea surface temperature, using the RAMSES II algorithm on Suomi NPP ATMS instrument data. It is an all-weather product without cloud clearing, and includes quality control flags for each variable.
Daily near-real-time model results from January 2024 and earlier for the Great Barrier Reef. This dataset contains outputs from version 3.2 of a 1km-resolution biogeochemistry and sediments model, forced by hydrodynamics and river inputs. The dataset was officially retired in February 2026 due to a model error producing unrealistically high Chlorophyll-a levels, and its authors advise discontinuing all use.
Original energy consumption data published on figshare by jiaqi Li. The dataset is available as a 10.7 KB XLSX file and was last updated on 2026-05-28. The specific source, time range, and geographic scope of the data are not detailed in the provided metadata.
Kimi-K2.6-Reasoning-3300x-WandB is a synthetic reasoning dataset containing 3,303 accepted rows. It was generated using the Kimi-K2.6 model through Weights & Biases Inference and is a subset of a larger planned 8,000-example distillation run. The dataset was created by author trjxter and last updated on 2026-05-18.
OpenS2V-Nexus is a detailed benchmark and dataset for subject-to-video generation. The released OpenS2V-5M subset contains approximately 5 million samples after filtering. It was created by author peterlrm and last updated on May 28, 2026.
ESMAtlas provides more than 700 million predicted protein structures generated by Meta AI's ESMFold language model. The dataset mirrors the ESM Metagenomic Atlas, built from protein sequences in the MGnify database. The v2023_02 release corresponds to the MGnify 2023_02 snapshot and was created using the esm.pretrained.esmfold_v1() model.
KairosQA is a question-answering dataset designed to evaluate the temporal reasoning capabilities of Large Language Models. It focuses on subject–relation–object triplets from Wikidata that changed at least twice between 2018 and 2025. The dataset was created by author kyutai and was last updated on Hugging Face in May 2026.
The NSW Department of Planning and Environment acquired this multibeam survey data from the Research Vessel Bombora between 20/APR/2021 and 30/APR/2021. The dataset contains 32-bit floating point geotiff files of bathymetry and backscatter at a 5-meter resolution for the Crowdy Head Offshore area. Data processing utilized Hypack, R2Sonic GUI, POSView, POSPac, Qimera, and FMGT software.
A 2026 study by Miguel Carrasco compares human and Vision Transformer (ViT) attention patterns for 20 artisanal objects. The dataset contains results from 1,152,000 distance evaluations across four metrics, based on eye-tracking data from 30 participants and attention maps from a 12-head ViT model. It is a 5.5 KB XLS file shared under a CC-BY-4.0 license.
Areas of Interest analysis results comparing human visual attention with Vision Transformer attention mechanisms for 20 artisanal objects. The dataset contains 1,152,000 distance evaluations from four metrics, derived from eye-tracking data of 30 participants and a pre-trained DINO ViT model. Author Miguel Carrasco uploaded the 5.5 KB XLS file to figshare in April 2026.
30 participants viewed 20 artisanal objects while their gaze patterns were recorded with a Pupil Labs eye-tracker. Miguel Carrasco published this dataset in 2026, which compares these human visual attention heatmaps to attention maps extracted from a pre-trained Vision Transformer model using four metrics across 1,152,000 distance evaluations. The analysis identifies specific attention heads, such as head #12, that show the strongest alignment with human visual patterns.
A historical geography dataset reconstructing settlement evolution in Xilingol League, a pastoral region in northern China, between 1915 and 1980. The data was created by Sen Mu using declassified statistics and historical documents, employing kernel density estimation and hotspot analysis. It was last updated on April 10, 2026.