datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
shot2story
Shot2Story: A New Benchmark for Comprehensive Understanding of Multi-shot Videos
Please download the multi-shot videos from OneDrive or HuggingFace.
We are excited to release a new video-text benchmark for multi-shot video understanding. This release contains a 134k version of our dataset. It includes detailed long summaries (human annotated + GPTV generated) for 134k videos and shot captions (human annotated) for 188k video shots.
Annotation Format
Our 134k multi-shot… See the full description on the dataset page: https://huggingface.co/datasets/mhan/shot2story.One-Shot-CFT-Data
One-Shot-CFT: Unleashing the Reasoning Potential of Pre-trained LLMs by Critique Fine-Tuning on One Problem
💻 Code |
📄 Paper |
📊 Dataset |
🤗 Model |
🌐 Project Page
🧠 Overview
One-Shot Critique Fine-Tuning (CFT) is a simple, robust, and compute-efficient training paradigm for unleashing the reasoning capabilities of pretrained LLMs in both mathematical and logical domains. By leveraging critiques on just one problem, One-Shot CFT enables models… See the full description on the dataset page: https://huggingface.co/datasets/TIGER-Lab/One-Shot-CFT-Data.Shot2Story-20K
Shot2Story: A New Benchmark for Comprehensive Understanding of Multi-shot Videos
We have a more recent release of 134K version here. Please have a look.
For video data downloading, please have a look at this issue.
We are excited to release a new video-text benchmark for multi-shot video understanding. This release contains a 134k version of our dataset. It includes detailed long summaries (human annotated + GPTV generated) for 134k videos and shot captions (human annotated) for… See the full description on the dataset page: https://huggingface.co/datasets/mhan/Shot2Story-20K.shot2story
Shot2Story: A New Benchmark for Comprehensive Understanding of Multi-shot Videos
Please download the multi-shot videos from OneDrive or HuggingFace.
We are excited to release a new video-text benchmark for multi-shot video understanding. This release contains a 134k version of our dataset. It includes detailed long summaries (human annotated + GPTV generated) for 134k videos and shot captions (human annotated) for 188k video shots.
Annotation Format
Our 134k multi-shot… See the full description on the dataset page: https://huggingface.co/datasets/huankguan2/shot2story.capybara-sharegpt
capybara-sharegpt
LDJnr/Capybara converted to ShareGPT format for use in common training repositories.
Please refer to the original repository's dataset card for more information. All credit goes to the original creator.
reasoning-sft-One-Shot-CFT-Data-4.7K
One-Shot-CFT-Data (converted)
Converted version of TIGER-Lab/One-Shot-CFT-Data, merging all 10 splits into 4,751 rows.
Format
Each row has three columns:
input — list of dicts [{"role": "user", "content": "..."}, ...] (conversation turns ending on the last user turn)
response — critique response string (includes <think> reasoning block followed by conclusion)
source — fixed as One-Shot-CFT-Data
Conversion
All 10 splits (4 DSR + 6 BBEH) merged into a… See the full description on the dataset page: https://huggingface.co/datasets/AmanPriyanshu/reasoning-sft-One-Shot-CFT-Data-4.7K.one-shot-grpo-bias-flipped
GRPO-Bias: One-Shot Flipped-Label Training Data
⚠️ Content warning. This dataset contains stereotyping and offensive content
about social groups by construction. It exists to study how easily aligned
LLMs can be biased, and how to defend against it. It does not reflect the views
of the authors or the University of Michigan.
This is the derived, flipped-label training data for the paper "It Takes One
to Bias Them All: Breaking Bad with One-Shot GRPO." These are the single (and… See the full description on the dataset page: https://huggingface.co/datasets/MichiganNLP/one-shot-grpo-bias-flipped.