datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
si_for_sdSIFT-50M
Dataset Card for SIFT-50M
SIFT-50M (Speech Instruction Fine-Tuning) is a 50-million-example dataset designed for instruction fine-tuning and pre-training of speech-text large language models (LLMs). It is built from publicly available speech corpora containing a total of 14K hours of speech and leverages LLMs and off-the-shelf expert models. The dataset spans five languages, covering diverse aspects of speech understanding and controllable speech generation instructions. SIFT-50M… See the full description on the dataset page: https://huggingface.co/datasets/amazon-agi/SIFT-50M.SWE-V-SIFsift-audio
SIFT Audio Dataset
Self-Instruction Fine-Tuning (SIFT) dataset for training audio understanding models.
Dataset Description
This dataset contains audio samples paired with LLM-generated responses following the
AZeroS multi-mode approach. Each audio sample is processed in three different modes
to train models that can both respond conversationally AND describe/analyze audio.
SIFT Modes
Each audio sample generates three training samples with different behaviors:… See the full description on the dataset page: https://huggingface.co/datasets/mazesmazes/sift-audio.ptb-sifosift-128-euclidean
Dataset Overview
dataset: sift-128-euclidean
Metadata
Creation Time: 2025-01-07 11:37:52+0000
Update Time: 2025-01-07 11:38:07+0000
Source: https://github.com/erikbern/ann-benchmarks
Task: N/A
Train Samples: N/A
Test Samples: N/A
License: DISCLAIMER AND LICENSE NOTICE:
This dataset is intended for benchmarking and research purposes only.
The source data used in this dataset retains its original license and copyright. Users must comply with the respective licenses of… See the full description on the dataset page: https://huggingface.co/datasets/open-vdb/sift-128-euclidean.banglish_bench
BanglishBench
A smoke test for Banglish models. It answers one question: did this build break?
700 prompts, 7 categories, a floor per category, and an exit code. The same job
pytest does before you demo a feature.
It does not rank models and it does not measure quality. It tells you whether a
build is worth the time it takes to read its answers. Whether the answers are
any good still takes a person who reads Banglish.
Run it
pip install huggingface_hub
hf… See the full description on the dataset page: https://huggingface.co/datasets/sifat-febo/banglish_bench.smoleval
SmolEval
Pick the right base model before you fine-tune
Which small base model is worth your training time? There are dozens under
2B, and fine-tuning the wrong one costs hours. This scores one in a few
minutes, on what base models actually do: continue text.
90 prompts, 3-run average
coherence
relevance
diversity
SmolLM2-135M
███░░░░░░░ 34%
█░░░░░░░░░ 7%
█░░░░░░░░░ 8%
SmolLM2-360M
██░░░░░░░░ 21%
░░░░░░░░░░ 0%
█░░░░░░░░░ 14%
SmolLM2-1.7B
████░░░░░░… See the full description on the dataset page: https://huggingface.co/datasets/sifat-febo/smoleval.sift-archive
Sift — 研究数据归档
Sift 是一个 CPU/DDR-primary + GPU-assisted 的分层内存 MoE + 长上下文推理系统研究项目
（用便宜的大容量 DDR/CXL 承载放不进 HBM 的大型稀疏 MoE + 长上下文；头条指标是 tokens-per-dollar / tokens-per-Joule）。
本仓是该项目自产实验数据的归档，用于把数据从本地磁盘卸下来。
这里没有模型权重 —— 模型是上游公开 GGUF，见 MODELS.manifest.json + restore_models.sh。
取数据
hf download yil384/sift-archive --repo-type dataset fetch_archive.sh --local-dir .
bash fetch_archive.sh # 列出仓里有什么
bash fetch_archive.sh ssd2/traces/v2lite #… See the full description on the dataset page: https://huggingface.co/datasets/yil384/sift-archive.SIF-VLM-Fingerprint-Triggers
SIF and AGDI VLM Fingerprint Triggers
This public repository contains 12,000 model-specific visual fingerprint
trigger images for research on fingerprint transfer and robustness in Large
Vision-Language Models:
9,000 SIF baseline triggers generated with Ordinary, RNA, and PLA;
3,000 AGDI triggers generated for the same three base models.
Dataset configs
Config
Training model
Method
Rows
qwen2.5-vl-7b
Qwen/Qwen2.5-VL-7B-Instruct
Ordinary, RNA, PLA
3… See the full description on the dataset page: https://huggingface.co/datasets/autoRiver/SIF-VLM-Fingerprint-Triggers.Kabyle-French
French - Kabyle (Tatoeba)
This dataset contains translation pairs for French (fr) and Kabyle (kab). The data was collected from the Tatoeba Project, a free collaborative online database of example sentences.
⚠️ Important Note on Quality
Disclaimer: This dataset has been exported automatically and has not been manually verified. While Tatoeba relies on community contributions, errors or inconsistencies in translation pairs may exist. Use with appropriate caution.
sifta-document-forgery-datasetDetectAIRevDetectAIRev, an AI-generated review detection dataset curated from human- written and LLM-generated reviews across diverse
domains and diverse- LLMs
eq-esconv-sifted
EQ-ESConv-Sifted: Elo-Ranked Emotional Support Conversations
The ESConv dataset (Liu et al., ACL 2021) ranked by empathetic quality via Swiss-style Elo tournament. All 1,300 conversations scored and sorted.
Why this exists
ESConv is a widely-used emotional support dataset but quality varies significantly — some conversations have excellent empathetic support, others are low-effort or off-topic. This dataset adds Elo rankings so you can filter by quality.
For… See the full description on the dataset page: https://huggingface.co/datasets/nivvis/eq-esconv-sifted.sift1bKabyleWikipediaDMSD-ood
Debiasing Multimodal Sarcasm Detection with Contrastive Learning
This is a replication of the DMSD-ood dataset for easier access.
Reference
Jia, M., Xie, C., & Jing, L. (2024). Debiasing Multimodal Sarcasm Detection with Contrastive Learning. Proceedings of the AAAI Conference on Artificial Intelligence, 38(16), 18354-18362.
electricity-productionThis dataset is used for demo purposes to illustrate using the time series forecasting models present in the Transformers library.
Source: https://www.kaggle.com/datasets/shenba/time-series-datasets
Sourjayon__DeepSeek-R1-8b-Sify-details
Dataset Card for Evaluation run of Sourjayon/DeepSeek-R1-8b-Sify
Dataset automatically created during the evaluation run of model Sourjayon/DeepSeek-R1-8b-Sify
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… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard/Sourjayon__DeepSeek-R1-8b-Sify-details.sifahane-turkish-medical-complaintsRedEval-ood
Leveraging Generative Large Language Models with Visual Instruction and Demonstration Retrieval for Multimodal Sarcasm Detection
This is a replication of the RedEval-ood dataset for easier access.
Reference
Binghao Tang, Boda Lin, Haolong Yan, and Si Li. 2024. Leveraging Generative Large Language Models with Visual Instruction and Demonstration Retrieval for Multimodal Sarcasm Detection. In Proceedings of the 2024 Conference of the North American Chapter of the… See the full description on the dataset page: https://huggingface.co/datasets/sifan077/RedEval-ood.english_voice_512english_voice_256Sify_CoT_Datasetbook_dataenglish_voice_dummytestsift_vivoice700k_qwen27Bsi-follow-dummysifito