datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
HiFi-KPI
HiFi-KPI: Hierarchical Financial KPI Extraction
Paper: HiFi-KPI: A Dataset for Hierarchical KPI Extraction from Earnings Filings
Dataset Summary
HiFi-KPI is a large-scale dataset designed for financial numerical key performance indicator (KPI) extraction from earnings filings. It is derived from iXBRL filings mandated by the SEC, featuring hierarchical labels structured from the XBRL taxonomy. The dataset consists of ∼1.8M paragraphs and ∼5M entities, each linked… See the full description on the dataset page: https://huggingface.co/datasets/AAU-NLP/HiFi-KPI.hifi-kpi-lite
HiFi-KPI Lite: Expert-Curated Financial KPI Extraction
Dataset Summary
HiFi-KPI Lite is a manually curated subset of the HiFi-KPI dataset, designed for evaluating structured financial KPI extraction. Unlike the full HiFi-KPI dataset, HiFi-KPI Lite maps financial entities to a much reduced, expert-defined label space. The dataset consists of ∼8K paragraphs and ∼25K entities, making it suitable for rapid model evaluation.
Supported Tasks
The dataset is… See the full description on the dataset page: https://huggingface.co/datasets/AAU-NLP/hifi-kpi-lite.effective-performance-measurement
Effective Performance Measurement: KPI Extraction Datasets
This dataset repository accompanies the ACL 2026 (Industry Track) paper: "Effective Performance Measurement: Challenges and Opportunities in KPI Extraction from Earnings Calls".
It contains three novel benchmarks and a prediction set designed to evaluate the extraction of Key Performance Indicators (KPIs) from unstructured financial texts, specifically comparing highly regulated SEC filings to conversational earnings calls.… See the full description on the dataset page: https://huggingface.co/datasets/AAU-NLP/effective-performance-measurement.cryptocurrency-tweets-sentiment
Dataset Card for cryptocurrency-tweets-sentiment
This dataset has been created with distilabel.
Dataset Summary
This dataset contains a pipeline.yaml which can be used to reproduce the pipeline that generated it in distilabel using the distilabel CLI:
distilabel pipeline run --config "https://huggingface.co/datasets/aaurelions/cryptocurrency-tweets-sentiment/raw/main/pipeline.yaml"
or explore the configuration:
distilabel pipeline info --config… See the full description on the dataset page: https://huggingface.co/datasets/aaurelions/cryptocurrency-tweets-sentiment.aau-rainsnow-mtidTRAM-Temporal-ADeLe_battery_v1dot0Silent-Greenalias-game
Instruction Manual: Generating Datasets for the Game 'Alias'
1. Objective
Your primary goal is to generate conversational datasets for the game 'Alias'. In this game, one entity (the user) gives creative clues to another entity (the assistant) to help them guess a secret word. The dataset should be in a structured JSON format.
2. Core Concept of the Game
The user (Clue-Giver): Knows the secret word. Their goal is to provide clues to the assistant without… See the full description on the dataset page: https://huggingface.co/datasets/aaurelions/alias-game.TCP-ADeLe_battery_v1dot0TemporalMix-ADeLe_battery_v1dot0UI
[ICLR 2026] Code Aesthetics with Agentic Reward Feedback
Paper Link👁️
1,2Bang Xiao#,
1,3Lingjie Jiang#,
1Shaohan Huang✉,
1Tengchao Lv,
1Yupan Huang,
1Xun Wu… See the full description on the dataset page: https://huggingface.co/datasets/AAUGS/UI.Tools
Function-calling fine-tune data (xLAM + ToolACE + Java/JS)
Training data for fine-tuning small language models on function calling, built against the
BFCL non-live AST categories. OpenAI/ChatML messages schema; Pythonic [func(arg=val)]
labels.
Configs
Config
File
Examples
Focus
python-headline
train.qwen.messages.jsonl
140,080
xLAM Python backbone + ToolACE + synthetic irrelevance + parallel up-weight
javajs-augmented
train.javajs.messages.jsonl
170… See the full description on the dataset page: https://huggingface.co/datasets/AAUGS/Tools.