datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
ultrafeedback-binarized-preferences-cleaned
UltraFeedback - Binarized using the Average of Preference Ratings (Cleaned)
This dataset represents a new iteration on top of argilla/ultrafeedback-binarized-preferences,
and is the recommended and preferred dataset by Argilla to use from now on when fine-tuning on UltraFeedback.
Read more about Argilla's approach towards UltraFeedback binarization at argilla/ultrafeedback-binarized-preferences/README.md.
Differences with argilla/ultrafeedback-binarized-preferences… See the full description on the dataset page: https://huggingface.co/datasets/argilla/ultrafeedback-binarized-preferences-cleaned.text-2-video-human-preferences
Rapidata Video Generation Preference Dataset
This dataset was collected in ~12 hours using the Rapidata Python API, accessible to anyone and ideal for large scale data annotation.
The data collected in this dataset informs our text-2-video model benchmark. We just started so currently only two models are represented in this set:
Sora
Hunyouan
Pika 2.0
Runway ML Alpha
Luma Ray 2
Explore our latest model rankings on our website.
If you get value from this dataset and would… See the full description on the dataset page: https://huggingface.co/datasets/Rapidata/text-2-video-human-preferences.text-2-video-human-preferences-wan2.1
Rapidata Video Generation Alibaba Wan2.1 Human Preference
If you get value from this dataset and would like to see more in the future, please consider liking it.
This dataset was collected in ~1 hour total using the Rapidata Python API, accessible to anyone and ideal for large scale data annotation.
Overview
In this dataset, ~45'000 human annotations were collected to evaluate Alibaba Wan 2.1 video generation model on our benchmark. The up to date benchmark… See the full description on the dataset page: https://huggingface.co/datasets/Rapidata/text-2-video-human-preferences-wan2.1.UltraFeedback-truthfulness-preferences
Dataset Card for "UltraFeedback-truthfulness-preferences"
More Information needed
text-2-video-human-preferences-seedance-1-pro
Rapidata Video Generation Seedance 1 Pro Human Preference
In this dataset, ~60k human responses from ~20k human annotators were collected to evaluate Seedance 1 Pro video generation model on our benchmark. This dataset was collected in roughtly 30 min using the Rapidata Python API, accessible to anyone and ideal for large scale data annotation.
Explore our latest model rankings on our website.
If you get value from this dataset and would like to see more in the future, please… See the full description on the dataset page: https://huggingface.co/datasets/Rapidata/text-2-video-human-preferences-seedance-1-pro.UltraFeedback-honesty-preferences
Dataset Card for "UltraFeedback-honesty-preferences"
More Information needed
text-to-speech-human-preferences-315k
Text-to-speech human preferences: 315K votes across 15 models
This gated dataset contains the evaluation record behind Datapoint Audio
Bench: 315,000 eligible pairwise votes comparing 15 text-to-speech
models in a complete round-robin over 300 English prompts. The prompt set
covers eight practical voice-agent categories, and every generated sample is
included as a typed audio record.
The source evaluation collected 357,651 completed responses. The published
benchmark excluded… See the full description on the dataset page: https://huggingface.co/datasets/datapointai/text-to-speech-human-preferences-315k.text-2-video-human-preferences-moonvalley-marey
Rapidata Video Generation Marey Pro Human Preference
In this dataset, ~75k human responses from ~15k human annotators were collected to evaluate Marey video generation model on our benchmark. This dataset was collected in roughtly 30 min using the Rapidata Python API, accessible to anyone and ideal for large scale data annotation.
Explore our latest model rankings on our website.
If you get value from this dataset and would like to see more in the future, please consider… See the full description on the dataset page: https://huggingface.co/datasets/Rapidata/text-2-video-human-preferences-moonvalley-marey.Capybara-Preferences
Dataset Card for Capybara-Preferences
This dataset has been created with distilabel.
Dataset Summary
This dataset is built on top of LDJnr/Capybara, in order to generate a preference
dataset out of an instruction-following dataset. This is done by keeping the conversations in the column conversation but splitting
the last assistant turn from it, so that the conversation contains all the turns up until the last user's turn, so that it can be reused… See the full description on the dataset page: https://huggingface.co/datasets/argilla/Capybara-Preferences.text-2-video-human-preferences-veo3
Rapidata Video Generation Veo 3 Human Preference
In this dataset, ~46k human responses from ~20k human annotators were collected to evaluate Veo3 video generation model on our benchmark. This dataset was collected in roughly 35 minutes using the Rapidata Python API, accessible to anyone and ideal for large scale data annotation.
Explore our latest model rankings on our website.
If you get value from this dataset and would like to see more in the future, please consider liking… See the full description on the dataset page: https://huggingface.co/datasets/Rapidata/text-2-video-human-preferences-veo3.ultrafeedback-binarized-preferences
Ultrafeedback binarized dataset using the mean of preference ratings
Introduction
This dataset contains the result of curation work performed by Argilla (using Argilla 😃).
After visually browsing around some examples using the sort and filter feature of Argilla (sort by highest rating for chosen responses), we noticed a strong mismatch between the overall_score in the original UF dataset (and the Zephyr train_prefs dataset) and the quality of the chosen response.
By… See the full description on the dataset page: https://huggingface.co/datasets/argilla/ultrafeedback-binarized-preferences.text-2-image-human-preferences-2m
Text-to-image human preferences: 2M votes across 30 models
This dataset contains the complete voting record behind the
Datapoint Image Bench
leaderboard: 2,161,160 validated pairwise votes — exactly 10 for each of
216,116 image pairs. The votes compare 30 text-to-image models in a complete
round-robin on 500 prompts, judged by annotators from over 200 countries.
Every vote includes the annotator's trust score at the time the vote was
cast.
Built on the Datapoint annotation… See the full description on the dataset page: https://huggingface.co/datasets/datapointai/text-2-image-human-preferences-2m.pairwise_preferencesv2text-2-video-human-preferences-veo3.1
Rapidata Video Generation Veo 3.1 Human Preference
In this dataset, ~74k human responses from ~23k human annotators were collected to evaluate the Veo 3.1 video generation model on our benchmark. This dataset was collected using the Rapidata Python API, accessible to anyone and ideal for large scale data annotation.
Explore our latest model rankings on our website.
If you get value from this dataset and would like to see more in the future, please consider liking it ❤️… See the full description on the dataset page: https://huggingface.co/datasets/Rapidata/text-2-video-human-preferences-veo3.1.text-2-video-human-preferences-veo2
Rapidata Video Generation Google DeepMind Veo2 Human Preference
If you get value from this dataset and would like to see more in the future, please consider liking it.
This dataset was collected in ~1 hour total using the Rapidata Python API, accessible to anyone and ideal for large scale data annotation.
Overview
In this dataset, ~45'000 human annotations were collected to evaluate Google DeepMind Veo2 video generation model on our benchmark. The up to… See the full description on the dataset page: https://huggingface.co/datasets/Rapidata/text-2-video-human-preferences-veo2.ultrafeedback-multi-binarized-preferences-cleaned
UltraFeedback - Multi-Binarized using the Average of Preference Ratings (Cleaned)
This dataset represents a new iteration on top of argilla/ultrafeedback-binarized-preferences-cleaned,
and has been created to explore whether DPO fine-tuning with more than one rejection per chosen response helps the model perform better in the
AlpacaEval, MT-Bench, and LM Eval Harness benchmarks.
Read more about Argilla's approach towards UltraFeedback binarization at… See the full description on the dataset page: https://huggingface.co/datasets/argilla/ultrafeedback-multi-binarized-preferences-cleaned.text-2-video-human-preferences-genmo-mochi-1
Rapidata Video Generation Genmo Mochi-1 Human Preference
In this dataset, ~60k human responses from ~20k human annotators were collected to evaluate mochi-1 video generation model on our benchmark. This dataset was collected in roughtly 30 min using the Rapidata Python API, accessible to anyone and ideal for large scale data annotation.
Explore our latest model rankings on our website.
If you get value from this dataset and would like to see more in the future, please… See the full description on the dataset page: https://huggingface.co/datasets/Rapidata/text-2-video-human-preferences-genmo-mochi-1.text-2-video-human-preferences-pika2.2
Rapidata Video Generation Pika 2.2 Human Preference
In this dataset, ~756k human responses from ~29k human annotators were collected to evaluate Pika 2.2 video generation model on our benchmark. This dataset was collected in ~1 day total using the Rapidata Python API, accessible to anyone and ideal for large scale data annotation.
Explore our latest model rankings on our website.
If you get value from this dataset and would like to see more in the future, please consider… See the full description on the dataset page: https://huggingface.co/datasets/Rapidata/text-2-video-human-preferences-pika2.2.fold_pants_preferences
fold_pants — pairwise preferences on a Franka Panda
Real-robot trajectories for "fold the shorts" with human pairwise preference
labels on multiple judgment axes. Built for reward-model / preference-learning
research: every label is a comparison of two trajectories on one named axis, not a
scalar score.
The trajectory data is a standard LeRobot
v2.1 dataset, so it also loads directly as an imitation-learning dataset.
Contents
Episodes
536
Frames
530… See the full description on the dataset page: https://huggingface.co/datasets/MarcelTorne/fold_pants_preferences.setup_table_preferences
setup_table — pairwise preferences on a Franka Panda
Real-robot trajectories for "set up the table" with human pairwise preference
labels on multiple judgment axes. Built for reward-model / preference-learning
research: every label is a comparison of two trajectories on one named axis, not a
scalar score.
The trajectory data is a standard LeRobot
v2.1 dataset, so it also loads directly as an imitation-learning dataset.
Contents
Episodes
467
Frames… See the full description on the dataset page: https://huggingface.co/datasets/MarcelTorne/setup_table_preferences.Capybara-Preferences-Filtered
Dataset Card for Capybara-Preferences-Filtered
This dataset has been created with distilabel, plus some extra post-processing steps described below.
Dataset Summary
This dataset is built on top of argilla/Capybara-Preferences, but applies a further in detail filtering.
The filtering approach has been proposed and shared by @LDJnr, and applies the following:
Remove responses from the assistant, not only in the last turn, but also in intermediate… See the full description on the dataset page: https://huggingface.co/datasets/argilla/Capybara-Preferences-Filtered.gigaverbo-v2-preferences
GigaVerbo-v2 Preferences: A Hybrid-Reasoning Portuguese Preference Dataset
Dataset Summary
GigaVerbo-v2 Preferences is a preference dataset designed for Direct Preference Optimization (DPO) and other direct alignment algorithms. The dataset comprises approximately 27.8 million tokens across 28,437 preference pairs, organized into 4 distinct subsets covering both quality-focused and safety-focused alignment. It is entirely composed of high-quality, LLM-generated data… See the full description on the dataset page: https://huggingface.co/datasets/Polygl0t/gigaverbo-v2-preferences.UltraFeedback-instruction_following-preferences
Dataset Card for "UltraFeedback-instruction_following-preferences"
More Information needed
dialect-preferences
DiaLLM — Pooled Preference Dataset (Implicit Thread)
Part of DiaLLM: An Investigation into the Robustness-Generation Gap in
English Dialect Adaptation (EMNLP 2026 Main).
45,690 preference pairs, pooling all three variety-specific sets
(Australian,
Northern British,
Indian) without
variety targeting. Used for implicit-thread DPO training, where the three
varieties are pooled rather than targeted individually, preserving the
variety-agnostic objective of that thread.… See the full description on the dataset page: https://huggingface.co/datasets/surrey-nlp/dialect-preferences.ultrafeedback-multi-binarized-quality-preferences-cleanedplate_toast_preferences
plate_toast — pairwise preferences on a Franka Panda
Real-robot trajectories for "put the toast in the plate" with human pairwise preference
labels on multiple judgment axes. Built for reward-model / preference-learning
research: every label is a comparison of two trajectories on one named axis, not a
scalar score.
The trajectory data is a standard LeRobot
v2.1 dataset, so it also loads directly as an imitation-learning dataset.
Contents
Episodes
271… See the full description on the dataset page: https://huggingface.co/datasets/MarcelTorne/plate_toast_preferences.DPO-tldr-summarisation-preferences
Dataset Card for DPO-tldr-summarisation-preferences
Reformatted from openai/summarize_from_feedback dataset.
The LION-series are trained using an empirically optimized pipeline that consists of three stages: SFT, DPO, and online preference learning (online DPO). We find simple techniques such as sequence packing, loss masking in SFT, increasing the preference dataset size in DPO, and online DPO training can significantly improve the performance of language models. Our best models… See the full description on the dataset page: https://huggingface.co/datasets/Columbia-NLP/DPO-tldr-summarisation-preferences.text-2-video-human-preferences-sora-2
Rapidata Video Generation Sora 2 Human Preference
In this dataset, ~75k human responses from ~15k human annotators were collected to evaluate the Sora 2 video generation model on our benchmark. This dataset was collected in roughtly 30 min using the Rapidata Python API, accessible to anyone and ideal for large scale data annotation.
Explore our latest model rankings on our website.
If you get value from this dataset and would like to see more in the future, please… See the full description on the dataset page: https://huggingface.co/datasets/Rapidata/text-2-video-human-preferences-sora-2.text-2-video-human-preferences-kling-v2.1-master
Rapidata Video Generation Kling v2.1 Master Human Preference
In this dataset, ~60k human responses from ~20k human annotators were collected to evaluate Kling v2.1 Master video generation model on our benchmark. This dataset was collected in roughtly 30 min using the Rapidata Python API, accessible to anyone and ideal for large scale data annotation.
Explore our latest model rankings on our website.
If you get value from this dataset and would like to see more in the future… See the full description on the dataset page: https://huggingface.co/datasets/Rapidata/text-2-video-human-preferences-kling-v2.1-master.text-2-video-human-preferences-sora-2-pro
Rapidata Video Generation Sora 2 Pro Human Preference
In this dataset, ~75k human responses from ~15k human annotators were collected to evaluate the Sora 2 Pro video generation model on our benchmark. This dataset was collected in roughtly 30 min using the Rapidata Python API, accessible to anyone and ideal for large scale data annotation.
Explore our latest model rankings on our website.
If you get value from this dataset and would like to see more in the future, please… See the full description on the dataset page: https://huggingface.co/datasets/Rapidata/text-2-video-human-preferences-sora-2-pro.