datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
wikiMIA-2024-hard
WikiMIA-2024 Hard Dataset
Dataset Description
WikiMIA_2024 Hard is a challenging dataset for membership inference attacks intorduced in the paper "The Surprising Effectiveness of Membership Inference with Simple N-Gram Coverage" containing temporal Wikipedia articles with different versions based on date cutoffs.
This dataset is designed to evaluate the robustness of privacy-preserving machine learning models against sophisticated membership inference techniques.
It… See the full description on the dataset page: https://huggingface.co/datasets/hallisky/wikiMIA-2024-hard.turkish-legal-statutory-hallucination-benchmark
Citation
If you use this dataset, please cite the accompanying paper:
@inproceedings{erdoganyilmaz2026statutoryhallucinations,
title = {Measuring Statutory Citation Hallucinations of LLMs in Turkish Law: A Multi-Agent Based Novel Benchmark Dataset and Multi-Dimensional Evaluation Framework},
author = {Cihan Erdoğanyılmaz and Ali Yasir Naç and Gamze Çoskuner},
booktitle = {2026 34th Signal Processing and Communications Applications Conference (SIU)},
year =… See the full description on the dataset page: https://huggingface.co/datasets/LawChatAI/turkish-legal-statutory-hallucination-benchmark.hallucination-bert-spans
Hallucination BERT Span Dataset
Flat, one-row-per-span dataset intended for span/token-classification
(BIO-tagging style) hallucination detection over agent tool-calling traces,
derived from the same judging pipeline as the reasoning-distillation set in
this collection.
File
ds_bert_spans_full.jsonl — 11,942 rows. Already self-contained — no join
needed. Each row is one hallucinated span: span (verbatim text), type
(taxonomy label), avg_iou / exact / n_judges… See the full description on the dataset page: https://huggingface.co/datasets/ssurface/hallucination-bert-spans.plm-hallubench
PLM-HalluBench: A Multi-level Benchmark for Evaluating Hallucinations in Protein Language Models
NeurIPS 2026 Evaluations and Datasets Track submission (double-blind).
PLM-HalluBench is a benchmark for evaluating hallucination in protein language models (PLMs) — outputs that look like proteins but violate basic biophysics. It is organised around a three-level taxonomy that separates sequence-, structure-, and function-level failure modes, and it pairs a Factual track (BPHS against a… See the full description on the dataset page: https://huggingface.co/datasets/plm-hallubench/plm-hallubench.brand-hallucination-and-ai-citation-benchmark
🛡️ Global Brand Hallucination & LLM Citation Benchmark Dataset
Official open dataset by Pixel Office EU tracking empirical brand hallucination rates, stale pricing quotes, and competitor deflection vectors across leading LLMs (ChatGPT GPT-4o, Claude 3.5 Sonnet, Perplexity AI, Google Gemini 2.5 Flash, and DeepSeek V3).
📊 Dataset Summary
Target Problem: Autonomous AI purchasing agents and AI search engines frequently cite outdated pricing tiers, non-existent… See the full description on the dataset page: https://huggingface.co/datasets/pixeloffice/brand-hallucination-and-ai-citation-benchmark.ai-hallucination-trials
AI Hallucination Trials
4,240 hand-coded trials from a research project on why large language models fabricate confident, false information and what reduces it. In each trial a researcher sent one prompt, usually an invented or obscure acronym or a non-existent institution, to one model under one prompting condition. The card's data holds the prompt, the model's full response, and hand-assigned codes for hallucination and hedging.
Models tested: Gemini, ChatGPT, and Claude.… See the full description on the dataset page: https://huggingface.co/datasets/mahashu/ai-hallucination-trials.hallucinationThis is a vendored reupload of the Benchmarking Unfaithful Minimal Pairs (BUMP) Dataset available at https://github.com/dataminr-ai/BUMP
The BUMP (Benchmark of Unfaithful Minimal Pairs) dataset stands out as a superior choice for evaluating hallucination detection systems due to its quality and realism. Unlike synthetic datasets such as TruthfulQA, HalluBench, or FaithDial that rely on LLMs to generate hallucinations, BUMP employs human annotators to manually introduce errors into summaries… See the full description on the dataset page: https://huggingface.co/datasets/GuardrailsAI/hallucination.strix-halo-inference-bench
Strix Halo Local Inference Benchmarks
Measured prefill and decode throughput, and real VRAM cost, for local GGUF models
on AMD Strix Halo (Radeon 8060S / gfx1151) under ROCm.
Why this exists
Strix Halo inverts the usual local-inference trade-off. A discrete 24 GB card gives
you high memory bandwidth and a hard capacity ceiling; Strix Halo gives you the
opposite — up to 64 GiB addressable as VRAM out of 128 GB unified, at substantially
lower bandwidth. That changes… See the full description on the dataset page: https://huggingface.co/datasets/axjns/strix-halo-inference-bench.lm-eval-results-AbacusResearch-haLLawa4-7b-private
Dataset Card for Evaluation run of AbacusResearch/haLLawa4-7b
Dataset automatically created during the evaluation run of model AbacusResearch/haLLawa4-7b
The dataset is composed of 62 configuration(s), each one corresponding to one of the evaluated task.
The dataset has been created from 4 run(s). Each run can be found as a specific split in each configuration, the split being named using the timestamp of the run.The "train" split is always pointing to the latest results.
An… See the full description on the dataset page: https://huggingface.co/datasets/nyu-dice-lab/lm-eval-results-AbacusResearch-haLLawa4-7b-private.toolace-ragtruth-style-hallucinations
ToolACE RAGTruth-style Tool Hallucination Dataset
This dataset was built from ToolACE tool-use dialogues and converted into a RAGTruth-style format for hallucination detection in tool calling.
Task
Given:
query: user query
context: tool response
output: final assistant answer
the goal is to classify whether the answer is grounded in the tool output or belongs to one of three hallucination types.
Labels
clean
tool_output_conflict
overgeneration… See the full description on the dataset page: https://huggingface.co/datasets/Ali-Bhai/toolace-ragtruth-style-hallucinations.halide-training-data
Project Halide Training Data
Film defect detection training data for MiniCPM-V 4.6 fine-tuning.
Dataset
FilmDamageSimulator (Eurographics 2023)
10 film scans (4K resolution)
12,137 defect annotations
5 classes: dust, dirt, short_hair, long_hair, scratch
All bounding boxes normalized to [0.0-1.0]
Format
JSONL with structure:
Classes
Class
Count
Color
dust
7,631
Red
dirt
2,700
Orange
short_hair
1,341
Cyan
long_hair… See the full description on the dataset page: https://huggingface.co/datasets/Lonelyguyse1/halide-training-data.elv-halluc-videos
ELV-Halluc — videos + annotations
A self-contained mirror of the ELV-Halluc benchmark
(CVPR 2026), bundling the raw .mp4 files together with the annotations so the benchmark can be
run without sourcing videos separately.
Paper: arXiv:2508.21496
Original annotations: HLSv/ELV-Halluc (no videos)
Project page: https://elv-halluc.github.io/
This is an unofficial mirror. All credit for the benchmark goes to the original authors; please
cite their paper (below) rather than this… See the full description on the dataset page: https://huggingface.co/datasets/shuzhig/elv-halluc-videos.halueval-llm-judges
HaluEval
Data: https://github.com/RUCAIBox/HaluEval/blob/main/data/dialogue_data.json
Judge prompt: https://github.com/RUCAIBox/HaluEval/blob/main/evaluation/dialogue/dialogue_evaluation_instruction.txt
split: 20,000 examples in total --> 10,000 good responses (i.e., ref="no") + 10,000 hallucinated responses (i.e., ref="yes")
This repository
llm-judges folder contains predictions made by each judge (logits over Yes and No)
Note that "Yes" means there is a… See the full description on the dataset page: https://huggingface.co/datasets/potsawee/halueval-llm-judges.HalCap-Bench
HalCap-Bench
HalCap-Bench dataset.
Columns
model
image_source
image_name
image_type
sentence_index
caption
annotation
error_type
error_words
agreement_ratio
fleiss_Pi
n_correct
n_incorrect
n_unknown
image_url
image_path_in_repo
Notes
Notes
For COCO/CC12M items, the image is referenced by image_url.
For SD/Imagen/data_generation items, the image file is stored under images/ and referenced by image_path_in_repo.
HalfCheetah-v4
HalfCheetah-v4 - Imitation Learning Datasets
This is a dataset created by Imitation Learning Datasets project.
It was created by using Stable Baselines weights from a TD3 policy from HuggingFace.
Description
The dataset consists of 1,000 episodes with an average episodic reward of 9809.9417.
Each entry consists of:
obs (list): observation with length 2.
action (int): action (0 or 1).
reward (float): reward point for that timestep.
episode_starts (bool): if that state was… See the full description on the dataset page: https://huggingface.co/datasets/NathanGavenski/HalfCheetah-v4.HalfCheetah-v2
HalfCheetah-v2 - Continuous Imitation Learning from Observation
This dataset was created for the paper Explorative imitation learning: A path signature approach for continuous environments.
It is based on HalfCheetah-v2, which is an older version for the MuJoCo environment.
If you would like to use newer version, be sure to check: IL-Datasets repository for the updated list.
Description
The dataset consists of 10 episodes with an average episodic reward of 7581.5527.… See the full description on the dataset page: https://huggingface.co/datasets/NathanGavenski/HalfCheetah-v2.Nexesenex__Nemotron_W_4b_Halo_0.1-details
Dataset Card for Evaluation run of Nexesenex/Nemotron_W_4b_Halo_0.1
Dataset automatically created during the evaluation run of model Nexesenex/Nemotron_W_4b_Halo_0.1
The dataset is composed of 38 configuration(s), each one corresponding to one of the evaluated task.
The dataset has been created from 1 run(s). Each run can be found as a specific split in each configuration, the split being named using the timestamp of the run.The "train" split is always pointing to the latest… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard/Nexesenex__Nemotron_W_4b_Halo_0.1-details.lightblue__suzume-llama-3-8B-multilingual-orpo-borda-half-details
Dataset Card for Evaluation run of lightblue/suzume-llama-3-8B-multilingual-orpo-borda-half
Dataset automatically created during the evaluation run of model lightblue/suzume-llama-3-8B-multilingual-orpo-borda-half
The dataset is composed of 38 configuration(s), each one corresponding to one of the evaluated task.
The dataset has been created from 1 run(s). Each run can be found as a specific split in each configuration, the split being named using the timestamp of the run.The… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard/lightblue__suzume-llama-3-8B-multilingual-orpo-borda-half-details.hallucination_detection_transformers
ToolACE Hallucination Dataset
This dataset was generated for the assignment Hallucination Detection in Tool Calling.
It is based on ToolACE tool-calling dialogues and uses a RAGTruth-style schema:
query: user question
context: tool output / grounding evidence
output: assistant final answer
hallucination_labels: character-level hallucination spans
Files:
File
Rows
Description
toolace_clean_ragtruth.jsonl
1347
clean ToolACE tool-use answers… See the full description on the dataset page: https://huggingface.co/datasets/HASSANI8046/hallucination_detection_transformers.gaverfraxz__Meta-Llama-3.1-8B-Instruct-HalfAbliterated-TIES-details
Dataset Card for Evaluation run of gaverfraxz/Meta-Llama-3.1-8B-Instruct-HalfAbliterated-TIES
Dataset automatically created during the evaluation run of model gaverfraxz/Meta-Llama-3.1-8B-Instruct-HalfAbliterated-TIES
The dataset is composed of 38 configuration(s), each one corresponding to one of the evaluated task.
The dataset has been created from 1 run(s). Each run can be found as a specific split in each configuration, the split being named using the timestamp of the run.The… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard/gaverfraxz__Meta-Llama-3.1-8B-Instruct-HalfAbliterated-TIES-details.Quazim0t0__Halo-14B-sce-details
Dataset Card for Evaluation run of Quazim0t0/Halo-14B-sce
Dataset automatically created during the evaluation run of model Quazim0t0/Halo-14B-sce
The dataset is composed of 38 configuration(s), each one corresponding to one of the evaluated task.
The dataset has been created from 1 run(s). Each run can be found as a specific split in each configuration, the split being named using the timestamp of the run.The "train" split is always pointing to the latest results.
An additional… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard/Quazim0t0__Halo-14B-sce-details.comp6881-hal-dataHallucination_Datasetllm-hal-hidden-statesHalfCheetah-Expert-v2Minimal-HalfCheetah-Expert-v5HalfCheetah-Expert-v5