datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
Arabic-Optimized-Reasoning-Dataset
Arabic Optimized Reasoning Dataset
Dataset Name: Arabic Optimized ReasoningLicense: Apache-2.0Formats: CSVSize: 1600 rowsBase Dataset: cognitivecomputations/dolphin-r1Libraries Used: Datasets, Dask, Croissant
Overview
The Arabic Optimized Reasoning Dataset helps AI models get better at reasoning in Arabic. While AI models are good at many tasks, they often struggle with reasoning in languages other than English. This dataset helps fix this problem by:
Using fewer tokens… See the full description on the dataset page: https://huggingface.co/datasets/Jr23xd23/Arabic-Optimized-Reasoning-Dataset.project-llm-dataset-sft-citation-optimized-v3
SF-SFT-v3：引用净化后的正式 90/10 SFT
首先选择正确 config
目标
应使用的 config
原因
复现九个正式 SFT v3 run
official_90_10
真正按监督 token 冻结为约 90% 项目域 + 10% 通用
分析所有合格通用候选、重新设计配比
canonical_full_pool
保留完整 general_diverse 池
比较 v2/v3 引用净化
canonical_full_pool + SF-SFT-v2
记录身份对应最完整
不要因为名字里有 canonical 就直接用 canonical_full_pool 复现正式训练。引用净化缩短了项目域回答，而通用池未变短，使 full pool 的通用监督 token 比例升至 train 16.0396%、eval 16.0663%。official_90_10 才是最终训练视图。
两个 config、两个 split 的精确规模… See the full description on the dataset page: https://huggingface.co/datasets/vosldtgbj/project-llm-dataset-sft-citation-optimized-v3.cbt-optimized-zips
CBT JEE Optimized Dataset (Transparent Lossless WebP)
Ground-truth verified question and solution packages for JEE Advanced & JEE Main Computer-Based Tests.
All question and solution images have been optimized to transparent lossless WebP (66.7% smaller than raw PNGs) for high-speed client-side loading in the CBT Exam Engine.
adaption-marketing-optimized-neural-titans
Adaption Marketing Optimized Dataset - Neural Titans
Competition: Adaption AutoScientist Challenge ($50,000 Prize Pool)Track: MarketingTeam: Neural Titans (HackIndia)
Dataset Details
Metric
Value
Rows
5,000
Size
22.5 MB
Format
JSONL (instruction-tuning)
Pipeline Configuration
Recipes Applied
Deduplication - Removes duplicate and near-duplicate entries
Prompt Rephrasing - Diversifies prompt formulations for robust… See the full description on the dataset page: https://huggingface.co/datasets/rishini/adaption-marketing-optimized-neural-titans.