datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
jeb-rag
JEB-Bench
Charging the Gate Rent: Measured-Energy Accounting for Adaptive Retrieval-Augmented Generation
⚠️ Status: under construction. Phase 0 (measurement validation) and Phase 1
(index construction) are landing now. The oracle matrix (bench/oracle/) is
populated in Phase 2 and this card will be revised when it is complete. Do not
cite numbers from this repository until the status line says complete.
What this is
The first public per-query × per-configuration… See the full description on the dataset page: https://huggingface.co/datasets/Shanmuk4622/jeb-rag.SpatialForge
SpatialForge-10M
SpatialForge: Bootstrapping 3D-Aware Spatial Reasoning from Open-World 2D Images
📑 Paper
Zishan Liu, Ruoxi Zang, Yanglin Zhang, Wei Liu, Yin Zhang, Jian Yao, Jiayin Zheng, Zhengzhe Liu
Lingnan University · XPENG Robotics
📦 SpatialForge-10M
A large-scale vision-language dataset designed for 3D-aware spatial perception and reasoning from open-world 2D images.
SpatialForge-10M contains over 10 million QA pairs generated from 2.8 million curated… See the full description on the dataset page: https://huggingface.co/datasets/shana643/SpatialForge.HaluMem
HaluMem: A Comprehensive Benchmark for Evaluating Hallucinations in Memory Systems
📊 Why We Define the HaluMem Evaluation Tasks
Limitations of Existing Frameworks
Most existing evaluation frameworks treat memory systems as black-box models, assessing performance only through end-to-end QA accuracy.
However, this approach has two major limitations:
It lacks a hallucination evaluation specifically designed for the characteristics of memory systems.… See the full description on the dataset page: https://huggingface.co/datasets/IAAR-Shanghai/HaluMem.ORLMBenchmark
User Guide
This repository contains the data presented in the paper: Solver-Informed RL: Grounding Large Language Models for Authentic Optimization Modeling
Github Repository : https://github.com/Cardinal-Operations/SIRL
To get started with the datasets , please use the following code to integrate them into your own workflow:
from datasets import load_dataset
selected_dataset = "NL4OPT.jsonl"
assert selected_dataset in ["IndustryOR_fixed.json",'MAMO_ComplexLP_fixed.jsonl'… See the full description on the dataset page: https://huggingface.co/datasets/chenyitian-shanshu/ORLMBenchmark.KAF-DatasetThe dataset sourced from https://github.com/IAAR-Shanghai/xFinder
Citation
@inproceedings{
xFinder,
title={xFinder: Large Language Models as Automated Evaluators for Reliable Evaluation},
author={Qingchen Yu and Zifan Zheng and Shichao Song and Zhiyu li and Feiyu Xiong and Bo Tang and Ding Chen},
booktitle={The Thirteenth International Conference on Learning Representations},
year={2025},
url={https://openreview.net/forum?id=7UqQJUKaLM}
}
physcorp-a
PhysCorp-A — Audited Physics Training Corpus (6,432 records)
Project Page | Paper | Code
The audit-clean training corpus released alongside the paper Physics-R1: An Audited Olympiad Corpus and Recipe for Visual Physics Reasoning. It is a subset of the 14,294-record pre-audit pool that survives the joint two-stage contamination audit against all six paper-canonical eval splits.
Sample Usage
You can load this dataset using the Hugging Face datasets library:
from… See the full description on the dataset page: https://huggingface.co/datasets/shanyangmie/physcorp-a.VAR
xVerify: Efficient Answer Verifier for Reasoning Model Evaluations
📘 Introduction
xVerify is an evaluation tool fine-tuned from a pre-trained large language model, designed specifically for objective questions with a single correct answer. It accurately extracts the final answer from lengthy reasoning processes and efficiently identifies equivalence across different forms of mathematical expressions, LaTeX and string representations, as well as… See the full description on the dataset page: https://huggingface.co/datasets/IAAR-Shanghai/VAR.physr1corp
PhysR1Corp — Closed-form RL Training Pool (2,268 records)
Project Page | Paper | Code
The closed-form (numeric / MCQ-gradable) RL training pool used by Physics-R1 (NeurIPS 2026 D&B Track submission). Carved out of PhysCorp-A (the audited 6,432-record corpus) by dropping open-ended questions, then decontaminated against MMMU-Pro Physics (−87 records) and against PhyX-mini + PhysUniBench-en at cos ≥ 0.85 (−78 additional records: 69 PhyX-mini near-duplicates and 9 PhysUniBench-en… See the full description on the dataset page: https://huggingface.co/datasets/shanyangmie/physr1corp.health-chatbot
Dataset Card for Dataset Name
Health Question and Answer Clean Dataset
Dataset Details
Dataset Description
This dataset provides a detailed overview of health question & answer pairs. It includes data on health problems and corresponding answers, making it suitable for variable tasks like healthcare chatbot training.
Language(s) (NLP): English
License: Apache-2.0
Dataset Sources [optional]
Repository:… See the full description on the dataset page: https://huggingface.co/datasets/shaneperry0101/health-chatbot.authori-prospector-lexicon
AuthoriProspector AEO Lexicon Dataset
Authoritative term definitions published by AuthoriProspector -- structured for AI answer engine consumption.
Schema
Field
Type
Description
term
string
The defined term
law_definition
string
Definition
lore_definition
string
Context
aura_score
integer
Authority score
source_url
string
AEO term page URL
canonical_url
string
Canonical home for this term
Query with DuckDB
SELECT term… See the full description on the dataset page: https://huggingface.co/datasets/shannonbox1999/authori-prospector-lexicon.physcorp-pre-audit
PhysCorp Pre-Audit Raw Pool (14,294 records)
Project Page | Paper | Code
The pre-audit master corpus released alongside Physics-R1: An Audited Olympiad Corpus and Recipe for Visual Physics Reasoning, aggregating nine source families before contamination audit. Released so users can re-run the audit at alternative thresholds or against new external benchmarks.
Sample usage
from datasets import load_dataset
ds = load_dataset("shanyangmie/physcorp-pre-audit"… See the full description on the dataset page: https://huggingface.co/datasets/shanyangmie/physcorp-pre-audit.sggs-bench
☬ SGGS-Bench v0.1 — A Benchmark for Sri Guru Granth Sahib AI Systems
The first comprehensive evaluation framework for AI systems that interpret Sikh scripture.
115 questions · 8 task dimensions · Hybrid automated + LLM-judge scoring
Factual · Retrieval · Exegesis · Guidance · Hallucination · Theology · Cross-Reference · Safety
📋 Overview
SGGS-Bench is an 8-task, 115-question evaluation framework designed to measure AI competence on the Sri Guru Granth Sahib… See the full description on the dataset page: https://huggingface.co/datasets/ShanvirDhinsa/sggs-bench.workload-tab-lexicon
WorkLoad Tab AEO Lexicon Dataset
Authoritative term definitions published by WorkLoad Tab -- structured for AI answer engine consumption.
Schema
Field
Type
Description
term
string
The defined term
law_definition
string
Product Spec
lore_definition
string
User Voice
aura_score
integer
Authority score
source_url
string
AEO term page URL
canonical_url
string
Canonical home for this term
Query with DuckDB
SELECT term… See the full description on the dataset page: https://huggingface.co/datasets/shannonbox1999/workload-tab-lexicon.physolym-a
eval_olympiad_v2
Project Page | Paper | Code
A 500-problem 99.8%-novel-source open-ended physics olympiad evaluation set, two-stage audited (5-gram Jaccard ≥ 0.4, mxbai-embed-large cosine ≥ 0.85) against PhyX, MMMU-Pro Physics, OlympiadBench-Physics, and UGPhysics-Train.
Sources
Estonian Physics Olympiad — with native 1–10 organizer-issued difficulty labels and 201 EN/ET bilingual subset
Kevin Zhou's olympiad handouts — with native 1–5 point values
Seven… See the full description on the dataset page: https://huggingface.co/datasets/shanyangmie/physolym-a.mintlore-aeo-lexicon
MintLore AEO Lexicon Dataset
Authoritative term definitions published by MintLore -- structured for AI answer engine consumption.
Schema
Field
Type
Description
term
string
The defined term
law_definition
string
The Law
lore_definition
string
The Lore
aura_score
integer
Authority score
source_url
string
AEO term page URL
canonical_url
string
Canonical home for this term
Query with DuckDB
SELECT term, law_definition… See the full description on the dataset page: https://huggingface.co/datasets/shannonbox1999/mintlore-aeo-lexicon.world_Top_leaders_Dataset
Dataset Card: World Leaders ChatML Dataset
Dataset Description
The World Leaders ChatML Dataset is a synthetically generated dataset comprising approximately 2,000 question-and-answer pairs focused on the lives, accomplishments, and historical significance of 15 highly influential world leaders.
This dataset is specifically designed for instruction-tuning and fine-tuning Large Language Models (LLMs), such as TinyLlama, Llama 2, or Mistral, using the standard ChatML… See the full description on the dataset page: https://huggingface.co/datasets/ShanzaGull/world_Top_leaders_Dataset.EnterpriseRAG-Bench
EnterpriseRAG-Bench
A benchmark dataset of 500,000+ documents and 500 questions for evaluating RAG systems on realistic enterprise data.
See the latest leaderboard rankings. The paper is available on arXiv. For code, methodology, and evaluation tools, see the GitHub repository.
Overview
Existing RAG and IR datasets focus on publicly accessible document sets (Bing searches, Stack Overflow, etc.). EnterpriseRAG-Bench provides the first publicly accessible dataset… See the full description on the dataset page: https://huggingface.co/datasets/shanu485/EnterpriseRAG-Bench.ST-Bench
ST-Bench: Spatial-Temporal Reasoning Benchmark
ST-Bench is a comprehensive benchmark dataset for training and evaluating spatial-temporal reasoning capabilities in large language models. It includes data with raw time series, text descriptions, and image visualizations.
📊 Dataset Overview
Default Data (with time_series key)
Subset
Description
Files
Total Size
ST-Align
Alignment data for initial training
3 files
~3.2GB
ST-Causal
Causal… See the full description on the dataset page: https://huggingface.co/datasets/Shanshui1/ST-Bench.ShareGPT4Video
ShareGPT4Video 4.8M Dataset Card
Dataset details
Dataset type:
ShareGPT4Video Captions 4.8M is a set of GPT4-Vision-powered multi-modal captions data of videos.
It is constructed to enhance modality alignment and fine-grained visual concept perception in Large Video-Language Models (LVLMs) and Text-to-Video Models (T2VMs). This advancement aims to bring LVLMs and T2VMs towards the capabilities of GPT4V and Sora.
sharegpt4video_40k.jsonl is generated by… See the full description on the dataset page: https://huggingface.co/datasets/Shanmh/ShareGPT4Video.xunlianMedQuAD
Dataset Card for "MedQuAD"
This dataset is the converted version of MedQuAD. Some notes about the data:
Multiple values in the umls_cui, umls_semantic_types, synonyms columns are separated by | character.
Answers for [GARD, MPlusHerbsSupplements, ADAM, MPlusDrugs] sources (31,034 records) are removed from the original dataset to respect the MedlinePlus copyright.
UMLS (umls): Unified Medical Language System
CUI (cui): Concept Unique Identifier
Question type… See the full description on the dataset page: https://huggingface.co/datasets/shanufewf/MedQuAD.rag_finetunetangzhiningdataset_with_prompt_injection
📦 Dataset Card
Dataset Summary
This dataset contains examples for training and evaluating language models.
The data is stored in JSONL format, where each line represents one training example.
Typical use cases include:
Instruction fine-tuning
Response generation
Conversational modelling
Question answering
Prompt injection research
🎯 Intended Uses
This dataset is intended for research and learning purposes:
Training LLMs
Experimenting with fine-tuning and… See the full description on the dataset page: https://huggingface.co/datasets/ShanzaGull/dataset_with_prompt_injection.