datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
ChinaTravel
ChinaTravel Query Dataset
This dataset is licensed under Creative Commons Attribution 4.0 International (CC BY 4.0).
ChinaTravel is an open-ended travel-planning benchmark with compositional
constraint validation for language agents. See the
paper,
Hugging Face paper page,
code, and
bilingual sandbox database
(ModelScope mirror)
for the complete benchmark resources.
Introduction
For a given query, a language agent uses the sandbox tools to collect
information and… See the full description on the dataset page: https://huggingface.co/datasets/LAMDA-NeSy/ChinaTravel.LAMBDA
Dataset Summary
LAMDBA is a long term ad memorability dataset, featuring data from 1749 participants and 2205 ads across 276 brands.
Dataset Structure
from datasets import load_dataset
ds = load_dataset("behavior-in-the-wild/LAMBDA")
ds
DatasetDict({
train: Dataset({
features: ['video_id', 'recall_score', 'youtube_id', 'ad_details'],
num_rows: 1964
})
test: Dataset({
features: ['video_id', 'recall_score', 'youtube_id', 'ad_details']… See the full description on the dataset page: https://huggingface.co/datasets/behavior-in-the-wild/LAMBDA.lambda-chat
lambda-chat
lambda-chat is a Japanese instruction-following dataset for supervised fine-tuning of chat models. It combines openly available datasets into one consistent chat format for easier use.
Purpose
The dataset is intended for training and evaluating Japanese chat and instruction-following models. Each example uses a list of messages with role and content fields.
Source Data
The dataset contains data from the following sources.… See the full description on the dataset page: https://huggingface.co/datasets/KeisukeMiyamoto/lambda-chat.BeamRL-TrainData
BeamRL-TrainData
BeamRL-TrainData is a synthetic dataset of beam mechanics question-answer pairs used to train the BeamPERL model via Group Relative Policy Optimization (GRPO) with verifiable reward signals. Each row corresponds to a unique simply supported beam configuration solved symbolically, paired with natural-language questions and ground-truth reaction force answers.
Dataset Details
Property
Value
Rows
180
Beam type
Simply supported (pin at x=0… See the full description on the dataset page: https://huggingface.co/datasets/lamm-mit/BeamRL-TrainData.qwen3.5-moe-awq-calibration
Qwen3.5 MoE AWQ Calibration Dataset
Calibration dataset for AWQ (Activation-Aware Weight Quantization) of
Qwen/Qwen3.5-35B-A3B and
Qwen/Qwen3.5-35B-A3B-Base.
Designed for MoE expert routing diversity: Qwen3.5-35B-A3B has 256 experts with 8
active per token, so calibration data needs broad domain coverage to exercise as many
routing paths as possible.
Sampling methodology
Source: PleIAs/common_corpus
(open multi-domain corpus with labeled collections)
Filtering:
Token… See the full description on the dataset page: https://huggingface.co/datasets/Lambent/qwen3.5-moe-awq-calibration.post-cutoff-2024-2026-bundles
post-cutoff-2024-2026-bundles
12 research briefings (53,685 words / ~70K tokens) covering events from April 2024 through May 2026. Built as source material for context-distillation SFT of a pre-April-2024 base model, and usable directly as a small CPT-style corpus.
Format
{
"text": "<full markdown bundle>",
"topic": "ai_ml_2024_2026",
"word_count": 4950,
"char_count": 37474
}
Each bundle is markdown with ###-level entries (typically 10–18 entries per bundle)… See the full description on the dataset page: https://huggingface.co/datasets/Lambent/post-cutoff-2024-2026-bundles.BeamRL-EvalData
BeamRL-EvalData
BeamRL-EvalData is a synthetic dataset of beam mechanics question-answer pairs used to evaluate the BeamPERL model. It is the companion evaluation set to tphage/BeamRL-TrainData, and is deliberately designed with harder, more varied configurations to test out-of-distribution generalization: a fixed beam length (9*L) and load magnitude (-13*P) are used, but configurations span 1–3 simultaneous point loads and variable support positions (not just pin at x=0 and roller… See the full description on the dataset page: https://huggingface.co/datasets/lamm-mit/BeamRL-EvalData.silkome-masp
Silkome MaSp
lamm-mit/silkome-masp is the major ampullate spidroin (MaSp) sequence-property subset used for the
SilkomeGPT study:
Wei Lu, David L. Kaplan, and Markus J. Buehler, "Generative Modeling, Design, and Analysis of
Spider Silk Protein Sequences for Enhanced Mechanical Properties", Advanced Functional
Materials 34, 2311324 (2024).
The dataset is curated from lamm-mit/silkome-full
by selecting rows whose category1 is one of:
MaSp, MaSp1, MaSp2, MaSp2B, MaSp3, MaSp3B… See the full description on the dataset page: https://huggingface.co/datasets/lamm-mit/silkome-masp.lampung-pixelgpt
Lampung PixelGPT Dataset
This dataset contains preprocessed Lampung text data for training PixelGPT models.
Dataset Statistics
Language: Lampung (lampung)
Total samples: 1,029
Train samples: 945
Test samples: 84
Tokenizers
Grapheme tokenizer: izzako/sunda-llama-tokenizer
LLaMA tokenizer: ernie-research/DualGPT
Features
text_id: Document identifier
chunk_id: Chunk identifier within document
pixel_values: Rendered pixel representation of aksara… See the full description on the dataset page: https://huggingface.co/datasets/izzako/lampung-pixelgpt.NeuroDivBench
NeuroDivBench: Measuring LLM Behavioral Bias Toward Neurodivergent Users
Do LLMs stereotype disability? Here's the data to test that.
Tell an LLM "you are autistic" and its output changes in measurable, stereotyped ways: shorter sentences, more off-topic drift, literal interpretation of sarcasm (46% vs. 10% baseline). Tell it "you have OCD" and you get anxious, fragmented prose (effect size d = 2.76). Tell it "you have ADHD" and you get ALL CAPS enthusiasm and self-narrated… See the full description on the dataset page: https://huggingface.co/datasets/Lamir007/NeuroDivBench.Euskal-liburu-datasetaEuskerazko liburuak osaturiko dataseta. Booktegi webgunetik aterata.