GAIA Benchmark for Agentic LLMs
450+ non-trivial questions with single-answer targets for testing tool-using and search-equipped LLMs.
- English
Hugging Face datasets with plain-language descriptions: what's inside and what models to train on them. 4,788 in the catalog so far, with new ones added every hour.
Answered by Gemini 4 · the chat is kept only in this tab
450+ non-trivial questions with single-answer targets for testing tool-using and search-equipped LLMs.
Automatically updated daily digests on the AI market: model and GPU pricing, API latency, news, and benchmarks.
Multi-camera face video recordings and text metadata for contactless photoplethysmography (rPPG) and health metrics assessment.
Speech corpus divided into clean and noisy speech for training and evaluating ASR systems.
Medical image dataset with questions, answers, and semantic annotations in English and Chinese.
Large-scale extraction of texts, images, audio, and video from open datasets in a unified parquet format.
Massive English dataset for Document VQA used in Idefics3 fine-tuning.
A physically verified multimodal benchmark for bird sound attribution.
Parallel texts for machine translation evaluation across numerous languages, including low-resource ones.
A large-scale news corpus (10M to 100M texts) for pre-training and fine-tuning language models on current events.
A text dataset containing between 10K and 100K records without a card description.
Historical US newspaper pages with images, original OCR, and neural text recognition.
Nearly 2M text query and vector graphic element annotation pairs automatically generated using SAM3.
204K prompt-response pairs across 65 languages for training and evaluating multilingual LLMs.
Multilingual benchmark of basic facts about 30 European countries for evaluating QA models.
18k Python task descriptions and solutions in Alpaca format for code generation training.
A GitHub-sourced collection of source code (100M to 1B records) for training code LLMs.
Large-scale 3D medical CT dataset with complete segmentation of 167 anatomical structures from head to knees.
Public collection of 800 simulation-ready kitchen environments with rich metadata for robotics and Embodied AI training.
A complete snapshot of the Mixamo library: 2,317 animations and 114 rigged 3D characters with a unified skeleton for motion generation.
Dataset of 100K–1M examples featuring AI-generated images and human annotations for interpreting and analyzing generative models.
Tabular and text data on all trades, profits, and user behavior on the Polymarket prediction market for financial analysis and ML.
Curated English visual-text tasks for fair evaluation of large multimodal models.
Unnoficial UAVid dataset mirror featuring UAV frames and urban environment segmentation labels.