datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
audio_alltimeTIME-OutputThis repository contains the extracted time series features (tsfeatures) for each variate and the detailed forecasting results for every experiment.
Note: These files are for building leaderboard and visualization; users do not need to download this directory.
features/: Statistical Features (tsfeatures)
Each dataset's features are saved to: output/features/{dataset}/{freq}/.
This directory stores the computed tsfeatures for the variates in the dataset. The folder contains a CSV file… See the full description on the dataset page: https://huggingface.co/datasets/Real-TSF/TIME-Output.code_alltime
RealTimeData Monthly Collection - Github Code
This datasets provides the monthly screenshots of the 500 cherry-picked open source projects on GitHub from 2017 to current.
To access articles in a specific month, simple run the following:
ds = datasets.load_dataset('RealTimeData/code_alltime', '2020-02')
This will give you the 2020-02 version of the 500 selected GitHub repos that were just updated in 2020-02.
Want to crawl the data by your own?
Please head to… See the full description on the dataset page: https://huggingface.co/datasets/RealTimeData/code_alltime.bbc_news_alltime
RealTimeData Monthly Collection - BBC News
This datasets contains all news articles from BBC News that were created every months from 2017 to current.
To access articles in a specific month, simple run the following:
ds = datasets.load_dataset('RealTimeData/bbc_news_alltime', '2020-02')
This will give you all BBC news articles that were created in 2020-02.
Want to crawl the data by your own?
Please head to LatestEval for the crawler scripts.
Credit… See the full description on the dataset page: https://huggingface.co/datasets/RealTimeData/bbc_news_alltime.bbc_images_alltime
RealTimeData Monthly Collection - BBC News Images
This datasets contains all news articles head images from BBC News that were created every months from 2017 to current.
To access articles in a specific month, simple run the following:
ds = datasets.load_dataset('RealTimeData/bbc_images_alltime', '2020-02')
This will give you all BBC news head images that were created in 2020-02.
Want to crawl the data by your own?
Please head to LatestEval for the crawler… See the full description on the dataset page: https://huggingface.co/datasets/RealTimeData/bbc_images_alltime.arxiv_alltime
RealTimeData Monthly Collection - ArXiv
This datasets contains selected papers from arXiv that were created every months from 2017 to current.
To access papers in a specific month, simple run the following:
ds = datasets.load_dataset('RealTimeData/arxiv_alltime', '2020-02')
This will give you about 1k selected papers that were created in 2020-02.
Want to crawl the data by your own?
Please head to LatestEval for the crawler scripts.
Credit
This is… See the full description on the dataset page: https://huggingface.co/datasets/RealTimeData/arxiv_alltime.wikitext_alltime
RealTimeData Monthly Collection - Wikipedia
This datasets contains different versions of the 500 selected wikipedia articles from Wikipedia that were updated every months from 2017 to current.
To access articles in a specific month, simple run the following:
ds = datasets.load_dataset('RealTimeData/wikitext_alltime', '2020-02')
This will give you the 2020-02 version of the 500 selected wiki pages that were just updated in 2020-02.
Want to crawl the data by your own?… See the full description on the dataset page: https://huggingface.co/datasets/RealTimeData/wikitext_alltime.math_alltime
RealTimeData Monthly Collection - Math
This datasets contains selected math question from Math Stackoverflow that were created every months from 2017 to current.
To access questions in a specific month, simple run the following:
ds = datasets.load_dataset('RealTimeData/arxiv_alltime', '2020-02')
This will give youquestions that were created in 2020-02.
Want to crawl the data by your own?
Please head to LatestEval for the crawler scripts.
Credit
This is… See the full description on the dataset page: https://huggingface.co/datasets/RealTimeData/math_alltime.TIME
TIME
We present TIME, a task-centric benchmark comprising 50 fresh datasets and 98 forecasting tasks, where configurations are aligned with real-world operational requirements. By archiving window-level prediction results, our benchmark enables pattern-level evaluation and visualization. This repository contains the ready-to-use Hugging Face dataset for running experiments.
🏆Leaderboard
📂Processed csv (Optional: For reference or custom pipelines only)
📊Reuslts & tsfeatures… See the full description on the dataset page: https://huggingface.co/datasets/Real-TSF/TIME.TIME-ProcessedCSVThis repository contains the processed CSVs for the TIME benchmark. These files serve as the foundational source used to generate the official Hugging Face dataset and compute time series features.
These files are intended for users who require access to the raw data for custom extensions or flexible data integration. For the ready-to-use dataset, please visit Real-TSF/TIME.
MOSS-VL-Realtime-Demo-VideosRealtimeGym
Real-Time Reasoning
Real-time reasoning traces for agents in evolving environments. This dataset accompanies the Real-Time Reasoning project page and demos. It provides step-by-step agent states, actions, scores, and (optionally) "thinking" content for three environments: Freeway, Snake, and Overcooked, under varying cognitive loads and time-pressure budgets, across multiple seeds and agent paradigms.
Project page: https://realtimegym.saltlab.stanford.edu/
Dataset repo:… See the full description on the dataset page: https://huggingface.co/datasets/SALT-NLP/RealtimeGym.real-time-reasoning
Real-Time Reasoning
Real-time reasoning traces for agents in evolving environments. This dataset accompanies the Real-Time Reasoning project page and demos. It provides step-by-step agent states, actions, scores, and (optionally) "thinking" content for three environments: Freeway, Snake, and Overcooked, under varying cognitive loads and time-pressure budgets, across multiple seeds and agent paradigms.
Project page: https://bleaves.github.io/real-time-reasoning/
Dataset repo:… See the full description on the dataset page: https://huggingface.co/datasets/BLeaves/real-time-reasoning.shona1bbc_alltimeThis dataset contains BBC News articles from 2017 to 2022. The articles are arraged by month. Access the specific month by using the format "YYYY-MM" as config. Such as load_dataset("RealTimeData/bbc_alltime", "2021-1").magenta-realtime-mlx-cpp
Magenta RealTime — C++ MLX runtime bundle
This dataset is a re-packaging of
Google's Magenta RealTime weights
for the C++ MLX runtime in
rhymeswithlion/magenta-realtime-mlx-cpp.
It contains exactly what mlx-stream needs at startup; nothing more, nothing
less. The upstream .pt / .npy checkpoints are intentionally not
mirrored here — they're only useful for the (Python) re-export tooling on the
project's main distribution.
Contents
.
├──… See the full description on the dataset page: https://huggingface.co/datasets/rhymeswithlion/magenta-realtime-mlx-cpp.US-Real-time-gun-detection-in-CCTV-An-open-problem-datasetRealtime-QA-100K
Realtime-QA-100K
📄 Tech Report |
💻 GitHub
Realtime-QA-100K is a 100K-sample realtime video question answering dataset
constructed from YouTube videos. Each sample contains a multimodal
conversation and frame timestamp metadata that aligns every <|video|> token in
the assistant text with one video frame timestamp.
Open-source training subset.
Realtime-QA-100K is the open-source subset of the real-time training data for
MOSS-Video-Preview… See the full description on the dataset page: https://huggingface.co/datasets/OpenMOSS-Team/Realtime-QA-100K.RealTimeVideo-Instruct-112K
RealTimeVideo-Instruct-112K
A real-time video QA corpus of 112,102 instruction samples. Each question is
posed at the moment its answer first becomes visible in the video, so a
model must answer from the current scene rather than from the whole clip. It
was used, together with offline long-video QA from LLaVA-Video-178K, to train
StreamTTT.
Annotations only. No video is redistributed. Download each source video
set from its original provider (see Video sources).… See the full description on the dataset page: https://huggingface.co/datasets/zeyun-zhong/RealTimeVideo-Instruct-112K.osworld-realtime-gameswikitext_alltime_backup
Wikipedia for All Times
You could find the history of 419 selected Wikipedia pages for every month between 2017 to 2022.
Use this to download the historical version of Wikipedia articles in a specific month:
ds = datasets.load_dataset('RealTimeData/wikitext_alltime', '2017-8')
The time stamp follows the format of "YYYY-MM".
nyc-subway-realtime
NYC Subway Realtime Archive
Continuous capture of the New York City subway's public realtime feeds, decoded
into analysis-ready tables — plus the derived service-quality panels, learned
"normal" baselines, and disruption-prediction track record built on top of them.
Collected every 30 seconds since 2026-04-16 across all nine MTA GTFS-RT
feeds, by the pipeline behind subway.fyi.
Source: github.com/digitalhen/subway-data.
This archive exists because the source data disappears.… See the full description on the dataset page: https://huggingface.co/datasets/digitalhen/nyc-subway-realtime.real-time-library-occupancyaero-realtime-data
Aero Realtime Data
Training data for Aero Realtime, a native proactive audio-video language model whose input and output share one aligned autoregressive stream.
⚠️ This is only part of the data
This repository hosts only the self-packaged assets that are not available on an
official Hugging Face release:
qaego4d videos and TTS audio
all TTS audio for EgoIT-99K
the lmms_engine_parquet / sft_parquet manifests used by the training configs
The… See the full description on the dataset page: https://huggingface.co/datasets/kcz358/aero-realtime-data.real-time-traffic-video-dataset
Traffic Dataset - 500 Videos
Dataset comprises 500 videos of urban traffic captured by surveillance cameras, providing real-time traffic data enriched with bounding box annotations for vehicles and pedestrians. Designed for traffic monitoring and safety research, the dataset supports tasks like vehicle detection, traffic flow analysis, and accident prediction.
By leveraging this dataset, researchers and engineers can advance real-time object detection, traffic surveillance… See the full description on the dataset page: https://huggingface.co/datasets/UniDataPro/real-time-traffic-video-dataset.NEW_qwen2_5_MATH_1_5b_grpo_reg_beta_0.1_ccpo_bce_third_time_realrealtime-turn-detection-test-data
Realtime speech test recordings
Synthetic speech recordings for black-box Realtime API behavior tests in
Speaches. Each WAV file is the unmodified output of OpenAI
text-to-speech. Tests are responsible for adding silence, combining recordings, and choosing streaming chunk
boundaries for their scenarios.
metadata.jsonl follows the Hugging Face AudioFolder layout. Each record contains the generation inputs, file
digest, expected text, transcription, and word/speech intervals from… See the full description on the dataset page: https://huggingface.co/datasets/speaches-ai/realtime-turn-detection-test-data.github_latest
Latest GitHub Repositories
You could always access the latest Github repos via this dataset.
We update the dataset weekly, on every Sunday. So the dataset always provides the latest Github repos from the last week.
The current dataset on main branch contains the latest Github Repos submitted from 2024-08-26 to 2024-09-02.
The data collection is conducted on 2024-09-09.
Use the dataset via:
ds = datasets.load_dataset('RealTimeData/github_latest')
Previsou versions
You… See the full description on the dataset page: https://huggingface.co/datasets/RealTimeData/github_latest.audio-speech-realtime-voice-agents-2026
🎙️ Audio, Speech Foundation Models & Real-Time Voice Agents Dataset (2026 Edition)
A structured research dataset featuring 1,722 domain-verified research papers and 298 official code repositories focused on Full-Duplex Speech-to-Speech LLMs, Real-Time Voice Agents (<200ms Latency), Zero-Shot TTS, Voice Cloning, OpenAI Whisper-v3, Neural Audio Codecs (EnCodec/DAC/SNAC), and Generative Music (2023–2026).
Built with Universal Scientific Engine V18.1 Diamond, providing 48 schema… See the full description on the dataset page: https://huggingface.co/datasets/beatsprom/audio-speech-realtime-voice-agents-2026.realtime-conversational-voice-agent-duplex-2026
🎙️ Real-Time Conversational Voice Agent, Turn-Taking, Full-Duplex & Prosody SFT/DPO Dataset (2026)
This repository contains the 100-Sample Production Teaser for the Real-Time Conversational Voice Agent & Full-Duplex Prosody Suite (2026) by BeatsProm AI Research Lab.
The dataset is engineered to train open-weights language models (Qwen-2.5-Audio, Llama-3.1-Voice, Moshi, Mini-Omni, Whisper-LLM) into ultra-low latency, real-time conversational voice agents featuring sub-150ms… See the full description on the dataset page: https://huggingface.co/datasets/beatsprom/realtime-conversational-voice-agent-duplex-2026.