datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
asr-ser-quechua-collao-embeddings
ASR-SER embeddings for Quechua Collao
This repository contains embeddings only. It does not contain raw audio.
These embeddings were extracted for ASR-to-SER transfer experiments on the Quechua Collao emotional speech corpus. The release is intended for reproducible feature-extraction experiments and downstream analysis.
Dataset contents
One PyTorch tensor per utterance stored as an embedding file under embeddings/
A sanitized metadata table describing utterance… See the full description on the dataset page: https://huggingface.co/datasets/QuechuaBase/asr-ser-quechua-collao-embeddings.gemma4-serving-bench-data
Gemma 4 12B (QAT-Q4_0) — Serving-Behavior Test Data
Test data, charts, and the running research log from an autonomous research
loop characterizing and tuning a Gemma 4 12B QAT-Q4_0 model served via
llama.cpp/llamafile on a single RTX 3080 Ti. Every ~30 min the loop
summarizes findings, proposes a goal, tests it end-to-end, documents success or
failure, and publishes here + to GitHub.
Model under test: gemma-4-12b-it-qat-q4_0.gguf (Google, June 2026), 128K
ctx, f16 KV, MTP… See the full description on the dataset page: https://huggingface.co/datasets/SEBK4C/gemma4-serving-bench-data.drbench
DRBench: A Realistic Benchmark for Enterprise Deep Research
📄 Paper | 💻 GitHub | 💬 Discord
DRBench is the first of its kind benchmark designed to evaluate deep research agents on complex, open-ended enterprise deep research tasks. It tests an agent's ability to conduct multi-hop, insight-driven research across public and private data sources, just like a real enterprise analyst.
✨ Key Features
🔎 Real Deep Research Tasks: Not simple fact lookups. Tasks… See the full description on the dataset page: https://huggingface.co/datasets/ServiceNow/drbench.raw_friends_series_transcriptRaw transcript from friends tv series, chunk into ~1000 token lines. Text tagged by character.
language:
- en
size_categories:
- n<1K
AgentJudgeBench
AgentJudgeBench: Evaluating LLM Judge Reliability on Agentic Tool-Calling
A benchmark for systematically evaluating how reliably LLM judges assess
agentic tool-calling workflows across structured, dependency-driven tasks.
Why this benchmark?
AgentJudgeBench measures how reliably LLM judges assess agentic tool-calling outputs. It provides 3,808 benchmark records spanning six DAG topologies and three difficulty… See the full description on the dataset page: https://huggingface.co/datasets/ServiceNow-AI/AgentJudgeBench.SP500-Financial-News-Articles-Time-SeriesTextual Time Series Dataset for finetuning / pretraining.
Json version of original dataset.
Original Dataset : https://www.kaggle.com/datasets/skywalker290/financial-news-article-and-stock-trend-dataset?select=stock_data_articles.csv
Dr-CiK
Dr-CiK: A Testbed for Foresight-Driven Agents
Dr-CiK is a benchmark for evaluating whether agents can retrieve
forecasting-relevant context from a noisy document corpus, filter out
distractors, distill the retrieved context into forecast-useful evidence, and
produce forecasts grounded in that evidence.
Real-world time-series forecasting often depends not only on historical
observations but also on external context that must be actively discovered
from heterogeneous, noisy… See the full description on the dataset page: https://huggingface.co/datasets/ServiceNow/Dr-CiK.uranium-series
☢️ Uranium Research Series
By Artifact Virtual — Ali A. Shakil & Ava Shakil
"Hardware is algorithmic. Binary weights learn. Gradients are optional. Self-conditioning is the universal failure mode."
The Uranium Series is a sequence of research papers exploring the fundamental physics of neural computation — from treating GPUs as algorithmic substrates, through binary-weight learning without gradients, to the discovery that autoregressive models inevitably poison themselves through… See the full description on the dataset page: https://huggingface.co/datasets/amuzetnoM/uranium-series.pelican-svg-drawings
Pelican SVG: frontier model drawings, scored
139 SVGs produced by seven frontier models answering Simon Willison's prompt,
"generate an SVG of a pelican riding a bicycle", each with the score the
pelican_svg_env
OpenEnv environment gave it. Four of the seven are open weights and three are closed.
Simon has run that prompt against nearly every model release since early 2025, but the
results live as embedded images across 129 blog posts and scattered gists. This dataset
exists… See the full description on the dataset page: https://huggingface.co/datasets/sergiopaniego/pelican-svg-drawings.Sera-4.5A-Full-T1This dataset contains 72118 trajectories. Data was generated from the first rollout of SVG on 121 SWE-smith codebases using GLM-4.5-Air as teacher and includes three SVG runs per function.
Schema:
messages: Generated trajectory
instance_id: ID of trajectory
rollout_patch: Created patch to the codebase from the current trajectory
func_name: Name of function sampled from codebase to start the pipeline
func_path: File path to the sampled function
problem_statement: Problem statement provided to… See the full description on the dataset page: https://huggingface.co/datasets/allenai/Sera-4.5A-Full-T1.SERA-4.6-Lite-Best-SubsetThis dataset contains 47498 high-quality samples from Sera-4.6-Lite-T1 and Sera-4.6-Lite-T2. Training leads to a open-source SoTA 50.67% +/- 1.86% performance on SWE-Bench Verified at 32K context length, outperforming Devstral-Small-2 and GLM-4.5-Air.
Method:We keep only model-submitted train samples and then filter by truncation ratio at 32K tokens until a threshold ratio of 0.88.
Schema:
messages: Generated trajectory
instance_id: ID of trajectory
rollout_patch: Created patch to the codebase… See the full description on the dataset page: https://huggingface.co/datasets/allenai/SERA-4.6-Lite-Best-Subset.orc-bench
ORC-bench
Task 1: Topological Path Finding
Task 2: Topological Connectivity
Task 3: Linear Power Flow
Task 4: Contingency Analysis
Task 5: Power Grid ControlTask 6: Power Flow Optimization
Task 1: Topological Path Finding
Problem Formulation
This task assesses the spatial reasoning ability of the model by asking it to determine the shortest path between two specific buses in a given power grid state. The grid state… See the full description on the dataset page: https://huggingface.co/datasets/serval-uni-lu/orc-bench.Sera-4.6-Lite-T2This dataset contains 36083 trajectories. A 25000 subset was used to train SERA-32B. Data was generated from the second rollout of SVG on 121 SWE-smith codebases using GLM-4.6 as teacher.
Schema:
messages: Generated trajectory
instance_id: ID of trajectory
rollout_patch: Created patch to the codebase from the current trajectory
func_name: Name of function sampled from codebase to start the pipeline
func_path: File path to the sampled function
problem_statement: Problem statement provided to… See the full description on the dataset page: https://huggingface.co/datasets/allenai/Sera-4.6-Lite-T2.SerendibLLM-PoemSong-Dataset
SerendibLLM Poem and Song Dataset
Sinhala poem and song lyrics dataset for fine-tuning Sinhala LLMs on creative generation.
Built as part of the Serendib LLM Honours project (UCLan 2025-2026).
19,184 instruction-response entries
Sources: kawmuthu.blogspot.com (poems) + lyrics-lk.com (songs)
6 instruction variants per entry (Sinhala + English prompts)
8 themes: love, nature, sadness, joy, spring, country, religion, general
NavLock-BreachSynth
NavLock-BreachSynth
NavLock-BreachSynth is a fixed, test-only multimodal counterfactual benchmark
for navigation-lock warning reliability assessment. It pairs each generated
hazard scene with its factual source observation in
NavLock-HY through
the stable sample_idx field. Project code is available at
https://github.com/serendipitylxd/NavLock-BreachSynth.
Release Scope
123 fixed test scenes and 1186 synchronized frames;
eight camera images per generated frame;… See the full description on the dataset page: https://huggingface.co/datasets/serendipitylxd/NavLock-BreachSynth.online_terms_of_service
Dataset Card for A Corpus for Multilingual Analysis of Online Terms of Service
Dataset Summary
"We present the first annotated corpus for multilingual analysis of potentially unfair clauses in online Terms of
Service [=ToS]. The data set comprises a total of 100 contracts, obtained from 25 documents annotated in four different
languages: English, German, Italian, and Polish. For each contract, potentially unfair clauses for the consumer are
annotated, for nine different… See the full description on the dataset page: https://huggingface.co/datasets/joelniklaus/online_terms_of_service.Sera-4.5A-Full-T2This dataset contains 66337 trajectories. Data was generated from the second rollout of SVG on 121 SWE-smith codebases using GLM-4.5-Air as teacher and includes three SVG runs per function. Sera-4.5-Lite-T2 is a subset of this dataset and was used to train SERA-32B-GA.
Schema:
messages: Generated trajectory
instance_id: ID of trajectory
rollout_patch: Created patch to the codebase from the current trajectory
func_name: Name of function sampled from codebase to start the pipeline
func_path:… See the full description on the dataset page: https://huggingface.co/datasets/allenai/Sera-4.5A-Full-T2.retail-bank-servicing-alignment-sft
Retail Bank Servicing Alignment SFT
The training corpus for the Granite retail-bank servicing agent. It is the
released tool-use SFT corpus merged with a servicing-alignment continuation
curriculum that teaches multi-turn behaviours the base corpus does not: what to
do when the customer says "that one", when a policy question interrupts a
transfer, when the agent's own previous turn was wrong, and when the honest
answer is that the agent cannot see what it was asked about.
Every… See the full description on the dataset page: https://huggingface.co/datasets/spkc83/retail-bank-servicing-alignment-sft.brazilian-customer-service-conversations
Brazilian Customer Service Conversations
Dataset de conversas de atendimento ao cliente em portugues brasileiro (PT-BR).
De um like me apoie em manter esse dataset!
Descricao
Conversas sinteticas de alta qualidade simulando interacoes reais entre clientes e atendentes em diversos setores da economia brasileira. Util para treinar e avaliar modelos de:
Chatbots de atendimento
Classificacao de intencao (intent classification)
Analise de sentimento em conversas
Geracao de… See the full description on the dataset page: https://huggingface.co/datasets/RichardSakaguchiMS/brazilian-customer-service-conversations.bible-reference
Bible Reference Corpus
Thirteen aligned reference datasets for study of the biblical text: Greek and
Hebrew lexicons keyed to Strong's numbers, an interlinear word map, the critical
apparatus of eight Greek editions, cross-reference and topical indexes, and
geolocated places.
Published by SermonIndex.
Everything in this repository is public domain or CC BY 4.0. Sources with
share-alike terms are kept in a separate repository,
sermonindex/bible-reference-sa,
so that a share-alike… See the full description on the dataset page: https://huggingface.co/datasets/sermonindex/bible-reference.requests-pr-diff
requests-pr-diff
Generated by Repo2RLEnv.
💡 Browse this dataset in your browser — click the badge above or open
HuggingFaceH4/harbor-visualiser
to inspect every task's spec, instruction, oracle patch, test script, and Dockerfile.
Source repo: psf/requests
Pipeline: pr_diff
Tasks: 47
Visibility: public
Spec: Harbor task format with [metadata.repo2env] extension
Reward kinds
This dataset emits diff_similarity rewards. Each task ships an oracle diff at… See the full description on the dataset page: https://huggingface.co/datasets/sergiopaniego/requests-pr-diff.ConflictGUI
ConflictGUI
ConflictGUI is a benchmark for evaluating conflict awareness in GUI agents.
Dataset Splits
Split
Feasible
Conflict1
Conflict2
Total
Calibration
564
300
300
1,164
Test
1,800
822
874
3,496
Sources
ConflictGUI is constructed from AMEX, AndroidControl, and AITZ. The conflict instructions and labels are newly annotated, while screenshots and original instructions remain subject to their respective upstream terms.… See the full description on the dataset page: https://huggingface.co/datasets/serein356/ConflictGUI.Sera-4.6-Lite-T1This dataset contains 36825 trajectories. Data was generated from the first rollout of SVG on 121 SWE-smith codebases using GLM-4.6 as teacher.
Schema:
messages: Generated trajectory
instance_id: ID of trajectory
rollout_patch: Created patch to the codebase from the current trajectory
func_name: Name of function sampled from codebase to start the pipeline
func_path: File path to the sampled function
problem_statement: Problem statement provided to the model
target_patch: Ground truth patch… See the full description on the dataset page: https://huggingface.co/datasets/allenai/Sera-4.6-Lite-T1.bible
The Bible in 1,004 Languages
14,497,397 verses across 1,253 translations in 1,004 languages, every verse
keyed to the same chapter-and-verse address so that any two languages can be
aligned by joining on book, chapter and verse.
The Bible is the most widely translated text in existence, and for several
hundred of the languages here it is the largest — sometimes the only —
substantial digitised text. That makes this corpus unusually useful for
low-resource machine translation… See the full description on the dataset page: https://huggingface.co/datasets/sermonindex/bible.eva-bench
EVA-Bench: A New End-to-end Framework for Evaluating Voice Agents
EVA-Bench is an end-to-end evaluation framework for conversational voice agents that orchestrates bot-to-bot audio conversations and scores them on both task accuracy and interaction experience.
About
No existing benchmark jointly addresses the two core evaluation challenges for voice agents: generating realistic simulated conversations, and measuring quality across the full scope of… See the full description on the dataset page: https://huggingface.co/datasets/ServiceNow-AI/eva-bench.Sera-4.5A-Lite-T2This dataset contains 35615 trajectories. Data was generated from the second rollout of SVG on 121 SWE-smith codebases using GLM-4.5-Air as teacher and includes one SVG run per function. 16000 samples from the dataset were used to train SERA-32B-GA. Sera-4.5-Full-T2 is a superset of this dataset with three SVG runs per function.
Schema:
messages: Generated trajectory
instance_id: ID of trajectory
rollout_patch: Created patch to the codebase from the current trajectory
func_name: Name of… See the full description on the dataset page: https://huggingface.co/datasets/allenai/Sera-4.5A-Lite-T2.Sera-4.5A-Lite-T1This dataset contains 36607 trajectories. Data was generated from the first rollout of SVG on 121 SWE-smith codebases using GLM-4.5-Air as teacher and includes one SVG runs per function.
Schema:
messages: Generated trajectory
instance_id: ID of trajectory
rollout_patch: Created patch to the codebase from the current trajectory
func_name: Name of function sampled from codebase to start the pipeline
func_path: File path to the sampled function
problem_statement: Problem statement provided to… See the full description on the dataset page: https://huggingface.co/datasets/allenai/Sera-4.5A-Lite-T1.customer-service
Customer Service Conversations Dataset
This dataset contains 100 realistic customer service conversations between customers and support agents. Each dialogue is 11 turns long and covers a variety of common issues such as late deliveries, billing errors, account problems, and more. It is ideal for training and evaluating AI assistants, chatbots, and customer support models.
Dataset Structure
Each conversation is stored as a JSON object with the following fields:… See the full description on the dataset page: https://huggingface.co/datasets/chenlei123/customer-service.Minecraft-Server-Chat
Minecraft Server Chat
Important Info: This dataset contains swears. I filtered out as much racism as possible. People who were racist were banned from the server. I am not affiliated with the server in any way.
A collection of 2,000,000 messages said across two years in a minecraft server. The minecraft semi-anarchy server logged all of its messages to discord between 2020 and 2023. I downloaded all of them and made them into a json in chronological order. I also cleaned the… See the full description on the dataset page: https://huggingface.co/datasets/declip/Minecraft-Server-Chat.pharma-serialized-events
ZigoTrace Pharma — Serialized Events (synthetic)
Feature vectors extracted from a synthetic DSCSA-style serialized medicine supply
chain, generated by packages/intelligence/src/synthetic.ts in the
zigo-pharma engine and exported via
hf/generate_fixtures.mjs. Used to train and validate the diversion-detection model
(zigotrace/pharma-authenticity-model).
⚠️ Synthetic data notice
This is entirely synthetic — a seeded generator (generateChain), not real
distributor or… See the full description on the dataset page: https://huggingface.co/datasets/AsamAce/pharma-serialized-events.