Math reasoning rollouts
Step-by-step reasoning logs of DeepSeek R1-Distill models solving MATH dataset problems.
- Question answering
- Text generation
- 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
Step-by-step reasoning logs of DeepSeek R1-Distill models solving MATH dataset problems.
Over 10K Codeforces problems for training and evaluating code model reasoning.
A large dataset of image-text pairs without re-compression for training diffusion models.
Problems from the AIME 2025 Olympiad and reference answers to evaluate neural network mathematical capabilities.
100 real operational PDFs and related questions from 10 industries to evaluate multimodal document analysis.
A copy of the PG-19 benchmark based on Project Gutenberg books for training and evaluating language models on long texts.
English question-answer set to test the ability of multimodal models to recognize details in images above 2K.
A subset of The Stack: 10K random code examples for each of 30 languages, 2.6 GB.
A large collection of high-resolution Yahoo Flickr images with English descriptions for generative models.
A text dialogue dataset in English and Chinese using 26,500 different APIs for agent training.
High-quality data selection for training small programming language models.
A multimodal English science question dataset with multiple-choice options and images.
939,344 examples from the dataset mixture used for SFT training of the Tulu 3 model family.
Closed access to ServiceNow test environments for evaluating AI agents within the WorkArena benchmark.
Tabular and text predictors for human speech processing and reading models.
An English dataset for natural language inference (NLI) built with human and model participation.
824 complex multi-hop questions across 2–15 Wikipedia articles to test RAG factuality, retrieval, and reasoning.
Audio recordings of everyday sentences in Irish English.
Updated code text dataset with 100 million to 1 billion examples for pre-training language models.
A large synthetic code task corpus with verifiable solutions and tests for SFT and RL.
Cleaned and fast version of the well-known MMLU benchmark in English for evaluating language model knowledge.
Prepared English text data sequence used to train the open Amber model.
Over 100k first-person CS2 rounds with 720p video, stereo audio, and tick-synchronized data for all 10 players.
1,440 de novo mini-protein binders of 50-120 residues against 16 targets created by Claude agents and verified in CRO.