datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
ruler-300-seed42
Frozen RULER 300, seed 42
This dataset freezes the exact RULER inputs used by the
short-long-pretraining native evaluation suite.
Repository: bicycleman15/ruler-300-seed42
Rows: 6,300
Tasks: s-niah-1, s-niah-2, s-niah-3, mk1, mk2, mv, mq
Context lengths: 1024, 2048, 4096
Samples per task/length: 300
Seed: 42
Dataset SHA-256: 4d82df6f9b1f2d9c45c0a0bda8c734032e62f517b746c6351bf9c2f38335ab3d
Tokenizer SHA-256: 1f186971e25f7bda3dd6f93a100bb8fa2a6801cf8dc3807c8a8c4e45f296ab90… See the full description on the dataset page: https://huggingface.co/datasets/bicycleman15/ruler-300-seed42.RuleWorldRuleWorld
RuleWorld is a large-scale benchmark for evaluating whether language models can retrieve and apply a shared repository of externally provided procedural rules. Its rules are abstract, globally reusable, and intentionally non-commonsense, so a model cannot answer reliably from world knowledge alone. Each rule is supplied in aligned natural-language (NL) and first-order logic (FOL) forms.
The benchmark covers three reasoning settings: Single-Rule QA, Parallel Multi-Rule QA… See the full description on the dataset page: https://huggingface.co/datasets/SharkSpicy/RuleWorld.DHSA_RULER
RULER Evaluation Data
This dataset contains pre-generated JSONL files for the RULER long-context evaluation benchmark, used in Long-Context Modeling with Dynamic Hierarchical Sparse Attention for Memory-Constrained LLM Inference (ICML 2026 Spotlight). RULER is designed to evaluate effective context length and long-context behavior beyond simple retrieval, covering retrieval, multi-hop tracing, aggregation, and question answering style tasks.
The files are organized by target… See the full description on the dataset page: https://huggingface.co/datasets/sxiong/DHSA_RULER.RuleWeaverRuleWeaver
This directory contains the public evaluation set for the paper RuleWeaver:
Benchmarking Rule-Centered Scenario Reasoning for Large Language Models.
To evaluate models on this benchmark, use the evaluation code provided in the
RuleWeaver GitHub repository.
File
Contents
scenario_qa.jsonl
96 scenario-based QA cases: 48 same-source and 48 cross-source.
rules.jsonl
The organized pool of 200 root rules, 50 per source dataset, with four final variants per… See the full description on the dataset page: https://huggingface.co/datasets/SharkSpicy/RuleWeaver.cross_rulings_hts_dataset_for_tariffs
CROSS Rulings HTS Dataset for Tariff Classification
Maintained by Flexify.AI Inc. as part of the ATLAS trade intelligence research program.
Paper: ATLAS: Benchmarking and Adapting LLMs for Global Trade via Harmonized Tariff Code Classification
Project Page: https://tariffpro.flexify.ai/
This dataset is constructed from the U.S. Customs and Border Protection (CBP) Rulings Online Search System (CROSS).It contains rulings where importers sought clarification on the correct Harmonized… See the full description on the dataset page: https://huggingface.co/datasets/Dayanand314Krishna/cross_rulings_hts_dataset_for_tariffs.ruler-niah-multilength-eval-benchmark
📌 Fixed Multi-Length RULER NIAH Benchmark (1K, 2K, 4K, 8K)
Deterministic synthetic Needle-In-A-Haystack (NIAH) benchmark splits for reproducible long-context evaluation.
Dataset Specifications:
Tasks (4):
niah_single_1: Repeat haystack, single word needle, number value.
niah_single_2: Essay haystack, single word needle, number value.
niah_single_3: Essay haystack, single word needle, UUID value.
niah_multikey_1: Essay haystack, 4 keys needle, number value.… See the full description on the dataset page: https://huggingface.co/datasets/gyung/ruler-niah-multilength-eval-benchmark.court-rules
US Court Rules — federal and state, held word for word
Code & tools: github.com/docketx — legal-scrambler pseudonymises a case file on your own hardware before a frontier model sees it; claude-for-legal is the Claude Code plugin (docketx-open-law) that loads these datasets and checks citations against them.
21,062 court rules across 28 states and the federal courts, in one uniform record.
Rules of civil and appellate procedure, evidence, criminal procedure, professional… See the full description on the dataset page: https://huggingface.co/datasets/docketx/court-rules.ruler-100-nemotron
RULER-100 — Nemotron-Nano-v3 tokenized
RULER long-context evaluation data, regenerated with the
nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16 (instruct) tokenizer so the labeled context
lengths are exact for that model — instead of drifting, as they do when RULER data tokenized for a
different model (e.g. Qwen3) is fed to Nemotron.
What's here
7 context lengths: 4096, 8192, 16384, 32768, 65536, 131072, 262144 (the model's max).
13 RULER tasks: niah_single_1/2/3… See the full description on the dataset page: https://huggingface.co/datasets/jet-ai/ruler-100-nemotron.ruler-500-llama2
RULER-500 — Llama-2 tokenized
RULER long-context evaluation data, regenerated with the
Llama-2 tokenizer (NousResearch/Llama-2-7b-hf — the ungated mirror of
meta-llama/Llama-2-7b; LlamaTokenizer, SentencePiece, vocab_size=32000) so the labeled
context lengths are exact for Llama-2-family models instead of drifting, as they do when RULER
data tokenized for a different model (e.g. Qwen3) is fed to Llama-2.
Built with RULER's current ("binary-search") generators — so records carry… See the full description on the dataset page: https://huggingface.co/datasets/tturing/ruler-500-llama2.ruler-2m-niah-external
RULER-2M NIAH Eval — external handoff
50 × single_needle_uuid samples at ~2 M tokens per sample. One of four
length variants (1 M / 2 M / 6 M / 12 M) prepared for the external
long-context retrieval handoff.
field
value
samples
50
tasks
{single_needle_uuid: 50}
target tokens
2,000,000
seed
1344
negative_rate
0.0
eval/heldout/data.jsonl is the chat-templated form, ready for
model.forward(); eval/heldout/raw.jsonl is the pre-template form… See the full description on the dataset page: https://huggingface.co/datasets/ryansubq/ruler-2m-niah-external.uae-banking-rulebook-qa
UAE Banking Rulebook QA Dataset
Dataset Summary
The UAE Banking Rulebook QA dataset is a question-answering dataset generated from the official UAE Central Bank Rulebook. The dataset is designed to support the development of retrieval-augmented generation (RAG) systems, question-answering models, and other generative AI use cases in the legal and financial domains.
The dataset includes 490,000+ QA pairs, each grounded in a relevant context chunk from the official banking… See the full description on the dataset page: https://huggingface.co/datasets/rajeshthangaraj1/uae-banking-rulebook-qa.ruler-1m-niah-external
RULER-1M NIAH Eval — external handoff
50 × single_needle_uuid samples at ~1 M tokens per sample. One of four
length variants (1 M / 2 M / 6 M / 12 M) prepared for the external
long-context retrieval handoff.
field
value
samples
50
tasks
{single_needle_uuid: 50}
target tokens
1,000,000
seed
1344
negative_rate
0.0
eval/heldout/data.jsonl is the chat-templated form, ready for
model.forward(); eval/heldout/raw.jsonl is the pre-template form… See the full description on the dataset page: https://huggingface.co/datasets/ryansubq/ruler-1m-niah-external.ruler-12m-niah-external
RULER-12M NIAH Eval — external handoff
50 × single_needle_uuid samples at ~12 M tokens per sample. One of four
length variants (1 M / 2 M / 6 M / 12 M) prepared for the external
long-context retrieval handoff.
field
value
samples
50
tasks
{single_needle_uuid: 50}
target tokens
12,000,000
seed
1344
negative_rate
0.0
eval/heldout/data.jsonl is the chat-templated form, ready for
model.forward(); eval/heldout/raw.jsonl is the pre-template form… See the full description on the dataset page: https://huggingface.co/datasets/ryansubq/ruler-12m-niah-external.Iranian-Court-Rulings-25K
Iranian Court Rulings Dataset
A structured Persian-language corpus containing 25,617 Iranian judicial rulings collected from publicly accessible pages of the Iranian National Judicial Opinions database.
The dataset is intended for research and development in Persian Legal NLP, Information Retrieval, Retrieval-Augmented Generation (RAG), semantic search, legal document understanding, and related areas.
Dataset Overview
Number of records: 25,617
Language: Persian… See the full description on the dataset page: https://huggingface.co/datasets/sinamahallati/Iranian-Court-Rulings-25K.mtg-rules-qapolish-court-rulings-sample
Polish Court Rulings — Sample (korpus-pl)
A production-grade, PII-hardened corpus of Polish court rulings — free evaluation sample.
Full corpus: 505,611 rulings · ~3.18B tokens, licensed commercially.
Contact: licensing@aioil.ai · aioil.ai
What this is
This sample contains 500 Polish court rulings drawn from the full korpus-pl dataset — a cleaned, deduplicated and PII-audited corpus of Polish jurisprudence built for AI training, evaluation and legal RAG… See the full description on the dataset page: https://huggingface.co/datasets/aioil-ai/polish-court-rulings-sample.alimony-rules-by-state-2026
Alimony Rules By State 2026
Alimony/spousal support rules for 13 states with mortgage impact.
Details
Records: 13
Format: JSONL
License: CC-BY-4.0
Last Updated: March 2026
Verified By: Wendy Thompson, CPA, CDLP, NMLS #504814
Publisher: Wendy Thompson Lending Team
Thompson Alpha Logic
State-by-state alimony duration and calculation methods mapped to mortgage qualification impact. Shows how alimony income qualifies (or disqualifies) for FHA, VA, and… See the full description on the dataset page: https://huggingface.co/datasets/Wendy-Thompson-Lending-Team/alimony-rules-by-state-2026.ruler-6m-niah-external
RULER-6M NIAH Eval — external handoff
50 × single_needle_uuid samples at ~6 M tokens per sample. One of four
length variants (1 M / 2 M / 6 M / 12 M) prepared for the external
long-context retrieval handoff.
field
value
samples
50
tasks
{single_needle_uuid: 50}
target tokens
6,000,000
seed
1344
negative_rate
0.0
eval/heldout/data.jsonl is the chat-templated form, ready for
model.forward(); eval/heldout/raw.jsonl is the pre-template form… See the full description on the dataset page: https://huggingface.co/datasets/ryansubq/ruler-6m-niah-external.mining_rules_datatot-cwq-plan-sft-outputs34-rule-full-pw4-expand-labels-v2
ToT CWQ Plan SFT - outputs34_rule_full_pw4_expand_labels_v2
Merged SFT output from local run outputs34_rule_full_pw4_expand_labels_v2.
Version ID
local output dir: tot/sft/outputs34_rule_full_pw4_expand_labels_v2
file: cwq_train_plan.no_mid.jsonl
dataset: CWQ
grouping backend: TOT_REL_GROUPING_BACKEND=rules
parallel workers: 4
strict expand parity: enabled
nested expand labels: enabled
Main difference from earlier runs
This version renders nested Expand… See the full description on the dataset page: https://huggingface.co/datasets/YF0808/tot-cwq-plan-sft-outputs34-rule-full-pw4-expand-labels-v2.baggageitems_rules_llama2ru-linux-sysadmin-dialogues
Russian Linux Sysadmin Dialogues (Датасет для обучения ИИ)
Высококачественный структурированный набор данных (датасет) на русском языке, содержащий профессиональные инструкции, разборы технических проблем и сценарии общения в сфере системного администрирования операционных систем семейства Linux.
Этот датасет разработан специально для тонкой настройки (fine-tuning) больших языковых моделей (LLM), обучения диалоговых агентов, умных помощников технической поддержки и наполнения… See the full description on the dataset page: https://huggingface.co/datasets/CBERX/ru-linux-sysadmin-dialogues.padel-rules-sft
padel-rules-sft
1,496 supervised fine-tuning examples teaching a small model padel rule fidelity:
answers accurate to the FIP regulations that import nothing from tennis or squash.
Used to train https://huggingface.co/vevag/padel-qwen3-1.7b-lora — 19.4% → 45.2% spec
adherence on a held-out set of 31 scenarios.
Format
One JSON object per line, chat format:
{"messages": [
{"role": "system", "content": "You are a helpful assistant for padel players. Answer… See the full description on the dataset page: https://huggingface.co/datasets/vevag/padel-rules-sft.boardgame_rules_qa_dataset_rufrc-rules-syntheticFRC manual specific synthetic dataset!