datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
TeleQnA
TeleQnA
A Benchmark for Large Language Models in Telecom Knowledge
Developed by the NetOp Team, Huawei Paris Research Center
Ali Maatouk · Fadhel Ayed · Nicola Piovesan · Antonio De Domenico · Merouane Debbah · Zhi-Quan Luo
📄 Read the Paper
🤗 Explore the Dataset
TeleQnA is a comprehensive dataset tailored to assess the knowledge of Large Language Models (LLMs) in the field of telecommunications. It… See the full description on the dataset page: https://huggingface.co/datasets/netop/TeleQnA.unpredictable_bulbapedia-bulbagarden-netThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance. For more details please see the accompanying dataset card.BLiMP-IT
BLiMP-IT
Dataset Summary
BLiMP-IT is a linguistically motivated benchmark for evaluating Italian language models through minimal pairs.
Each example consists of a grammatical sentence paired with a minimally different ungrammatical counterpart that isolates a single morphosyntactic contrast.
The benchmark is designed to evaluate whether language models assign higher probability to the grammatical sentence than to the ungrammatical one.
The benchmark is inspired by… See the full description on the dataset page: https://huggingface.co/datasets/NeTSlab/BLiMP-IT.TeleMath
TeleMath
A Benchmark for Large Language Models in Telecom Mathematical Problem Solving
Developed by the NetOp Team, Huawei Paris Research Center
Vincenzo Colle · Mohamed Sana · Nicola Piovesan · Antonio De Domenico · Fadhel Ayed · Merouane Debbah
📄 Read the Paper
🤗 Explore the Dataset
[!NOTE]
IMPORTANT: Please help us protect the integrity of this benchmark by not publicly sharing, re-uploading, or… See the full description on the dataset page: https://huggingface.co/datasets/netop/TeleMath.wetwijzer_netherlands_legal_corpus
WetWijzer Netherlands Legal Corpus
Hugging Face target: justicedao/wetwijzer_netherlands_legal_corpus.
This unified dataset bundles the quality-audited WetWijzer Netherlands legal corpus stack in one repository for frontend retrieval. It preserves the existing compatibility repositories and does not replace or delete them.
Contents
Laws: 4,999
Articles: 89,737
CID index rows: 94,736
Vector mapping rows: 94,736
BM25 document rows: 94,736
BM25 term rows: 120,521… See the full description on the dataset page: https://huggingface.co/datasets/justicedao/wetwijzer_netherlands_legal_corpus.quantum-networking-and-distributed
Neura Parse — Quantum Networking, Repeaters & Distributed Quantum Computing
A systems-frontier vertical on connecting quantum devices: entanglement distribution and distillation, quantum repeaters, quantum-internet protocol stacks, quantum memories/transduction, and modular/distributed quantum computing (nonlocal gates, circuit knitting across nodes, blind/verifiable delegated computation). Covers protocol and simulation methods used with tools such as NetSquid and SeQUeNCe… See the full description on the dataset page: https://huggingface.co/datasets/Neura-parse/quantum-networking-and-distributed.neteval-examNetEval is a NetOps evaluation suite for foundation models, consisting of 5269 multi-choice questions. Please check our paper for more details about NetEval.
We hope NetEval could help developers track the progress and analyze the NetOps ability of their models.
Citation
Please cite our paper if you use our dataset.
@misc{miao2023empirical,
title={An Empirical Study of NetOps Capability of Pre-Trained Large Language Models},
author={Yukai Miao and Yu Bai and Li Chen and… See the full description on the dataset page: https://huggingface.co/datasets/NASP/neteval-exam.ipfs_netherlands_laws
IPFS Netherlands Laws
Hugging Face target: justicedao/ipfs_netherlands_laws.
This dataset packages Netherlands law records with deterministic IPFS Content IDs. Each row includes a cid and content_address; article rows also include the parent law_cid.
This is a quality-audited catalog-backed Netherlands snapshot from official Dutch government sources. It is not the full Dutch legal corpus: the persistent catalog contains 42,956 discovered BWBR identifiers, of which 5,000 are… See the full description on the dataset page: https://huggingface.co/datasets/justicedao/ipfs_netherlands_laws.CTBench
CTBench
Evaluating Troubleshooting Capabilities of AI Agents in Realistic Telecom Network Operations
📄 Read the Paper
🤗 Explore the Dataset
[!NOTE]
IMPORTANT: Please help us protect the integrity of this benchmark by not publicly sharing, re-uploading, or distributing the dataset.
CTBench
CTBench is an agentic benchmark for evaluating AI agents in realistic telecom Network Operations and Maintenance… See the full description on the dataset page: https://huggingface.co/datasets/netop/CTBench.DoD-Instruction-8010-01-Information-Network-Transport
🌐 DoD Information Network Transport
Maintainer: Terry Eppler
Owner: US Federal Government
Source: DoD Instruction 8010.01
Dataset Size: question-answer records
Source Effective Date: September 10, 2018
Source Organization: Office of the DoD Chief Information Officer
Source Ownership: United States Department of Defense
📋 Overview
Dataset Summary
The DoD Information Network Transport Question-Answer Dataset contains document-grounded… See the full description on the dataset page: https://huggingface.co/datasets/leeroy-jankins/DoD-Instruction-8010-01-Information-Network-Transport.LKValues
LKValues
LKValues is a survey-grounded Sinhala–English resource suite for studying the alignment of large language models with Sri Lankan societal values.
It contains two complementary resources:
LKvaluesIT — an instruction-tuning dataset for training models to generate value-grounded explanations.
LKvaluesBench — an evaluation benchmark for testing controlled value-sensitive judgment.
LKValues is based on 40 societal values retained through a trilingual Sinhala–Tamil–English… See the full description on the dataset page: https://huggingface.co/datasets/Nethmi14/LKValues.ru-instruct-KAN-logic-v1
Russian Instruct KAN-Logic Dataset (v1)
Overview
ru-instruct-KAN-logic-v1 — это специализированный набор данных для instruction tuning (дообучения) языковых моделей на русском языке.
Основной фокус датасета — сложные логические рассуждения (Reasoning), математическое обоснование нейросетевых архитектур нового поколения (KAN - Kolmogorov-Arnold Networks) и теория распределенных вычислений.
Датасет содержит синтетические и курируемые пары instruction - output… See the full description on the dataset page: https://huggingface.co/datasets/K-Net-Labs/ru-instruct-KAN-logic-v1.NetBench
NetBench Dataset
Dataset Overview
The NetBench Dataset is a curated collection of expert-level question-answer pairs designed to benchmark the ability of large language models (LLMs) to achieve network subject matter expert (SME) intelligence across 20 critical telecommunications and network engineering categories. These categories include:
Network Fundamentals & L2 Switching: Basic device access, Layer 2 concepts (VLANs, STP, LAG), L2 security, and interface… See the full description on the dataset page: https://huggingface.co/datasets/NetoAISolutions/NetBench.netherlands-laws-nl-normalized
Netherlands Laws (Dutch, Normalized)
Hugging Face target: justicedao/netherlands-laws-nl-normalized.
This package is a normalized version of the Netherlands laws scrape output.
This is a capped Netherlands scrape, not the full Dutch corpus. The scrape used max_documents=100, parsed 151 law record(s), and discovered 626 unique official BWBR law document(s) before applying the cap. Documents failed: 0.
This refresh includes parser coverage improvements for older/French heading… See the full description on the dataset page: https://huggingface.co/datasets/justicedao/netherlands-laws-nl-normalized.RLVR-Env-Retrieval-Source-code-search-net-javascript
RLVR-Env-Retrieval-Source-code-search-net-javascript
RLVR-ready retrieval environment derived from Nan-Do/code-search-net-javascript.
Author: Aman Priyanshu
What Is This
A 100k-row retrieval QA dataset where each row contains a question, ground-truth chunks, and pre-mined distractor chunks (random + semantically similar). Designed for training and evaluating retrieval agents in an RLVR (Reinforcement Learning with Verifiable Rewards) setup — the agent searches through… See the full description on the dataset page: https://huggingface.co/datasets/AmanPriyanshu/RLVR-Env-Retrieval-Source-code-search-net-javascript.Stack2Graph_VD_vb.net
Vb.Net StackOverflow Vector Dataset
Summary
This Hugging Face dataset repository contains the Vb.Net shard of the Stack2Graph vector-database component as restorable Qdrant artifacts plus portable Parquet fallback files.
Hugging Face uses one dataset repository per programming language, so this repository is directly cloneable without an extra top-level archive wrapper.
The artifacts are intended for semantic and hybrid retrieval, graph entry-point finding, and… See the full description on the dataset page: https://huggingface.co/datasets/Mo7art/Stack2Graph_VD_vb.net.squad-nl-v2.0
SQuAD-NL v2.0 for Sentence Transformers
The SQuAD-NL v2.0 dataset (on Hugging Face: GroNLP/squad-nl-v2.0), modified for use in Sentence Transformers as a dataset of type "Pair with Similarity Score".
Score
We added an extra column score to the original dataset.
The value of score is 1.0 if the question has an answer in the context (no matter where), and 0.0 if there are no answers in the context.
The allows the evaluation of embedding models that aim to pair queries… See the full description on the dataset page: https://huggingface.co/datasets/NetherlandsForensicInstitute/squad-nl-v2.0.RLVR-Env-Retrieval-Source-code-search-net-python
RLVR-Env-Retrieval-Source-code-search-net-python
RLVR-ready retrieval environment derived from Nan-Do/code-search-net-python.
Author: Aman Priyanshu
What Is This
A 100k-row retrieval QA dataset where each row contains a question, ground-truth chunks, and pre-mined distractor chunks (random + semantically similar). Designed for training and evaluating retrieval agents in an RLVR (Reinforcement Learning with Verifiable Rewards) setup — the agent searches through distractors… See the full description on the dataset page: https://huggingface.co/datasets/AmanPriyanshu/RLVR-Env-Retrieval-Source-code-search-net-python.Gpt-5.4-Xhigh-Reasoning-2000x
Gpt-5.4-Xhigh-Reasoning-2000x
A premium-quality reasoning dataset containing 2,007 elite samples distilled from GPT-5.4 XHIGH (the highest reasoning effort tier of GPT-5.4). Each sample features deep, multi-step Chain-of-Thought traces that are significantly longer and more rigorous than standard GPT-5.4 outputs.
This dataset is specifically designed for Supervised Fine-Tuning (SFT) to transform general-purpose language models into powerful reasoning models with explicit thinking… See the full description on the dataset page: https://huggingface.co/datasets/Nettoov/Gpt-5.4-Xhigh-Reasoning-2000x.Stack2Graph_VD_.net
.Net StackOverflow Vector Dataset
Summary
This Hugging Face dataset repository contains the .Net shard of the Stack2Graph vector-database component as restorable Qdrant artifacts plus portable Parquet fallback files.
Hugging Face uses one dataset repository per programming language, so this repository is directly cloneable without an extra top-level archive wrapper.
The artifacts are intended for semantic and hybrid retrieval, graph entry-point finding, and… See the full description on the dataset page: https://huggingface.co/datasets/Mo7art/Stack2Graph_VD_.net.TeleLogsAgent
TeleLogsAgent
A Benchmark for LLM Tool-Use in 5G Network Root Cause Analysis
Developed by the NetOp Team, Huawei Paris Research Center
Mohamed Sana · Nicola Piovesan · Antonio De Domenico · Fadhel Ayed
📄 Read the Paper
🤗 Explore the Dataset
[!NOTE]
IMPORTANT: Please help us protect the integrity of this benchmark by not publicly sharing, re-uploading, or distributing the dataset.
TeleLogsAgent is a benchmark… See the full description on the dataset page: https://huggingface.co/datasets/netop/TeleLogsAgent.netflix-movies_showsqemu_networking
Qemu Networking from Claude Haiku 4.5
A synthetic instruction-tuning dataset covering QEMU networking concepts, generated using Claude Haiku 4.5.
Dataset Summary
Total rows: 75
Topic: QEMU virtual networking
Difficulty distribution: Easy, Intermediate, Advanced
278 unique tags across networking subtopics
Splits
train: 75 rows
Columns
id: Stable entry ID
instruction: Instruction text for fine-tuning
input: Original prompt/question
output:… See the full description on the dataset page: https://huggingface.co/datasets/creeperdatasets/qemu_networking.for_Conceal-NetworkDataset is meant to train OPEN_LLAMA_v2, a converted JSON version is also available
Usage
exemple with llama.cpp & open_llama_3b_v2
Finetune:
finetune --model-base "C:\llama.cpp\models\open_llama_3b_v2_f32.gguf" --train-data "C:\llama.cpp\docs\conceal\conceal56_llama.txt" --lora-out lora-CCX_01.gguf --save-every 0 --threads 16 --ctx 256 --rope-freq-base 10000 --rope-freq-scale 1.0 --batch 1 --grad-acc 1 --adam-iter 256 --adam-alpha 0.00025 --lora-r… See the full description on the dataset page: https://huggingface.co/datasets/Acktarius/for_Conceal-Network.