datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
BrainBench
BrainBench
BrainBench is a benchmark for evaluating AI agents on practical EEG and polysomnography reasoning tasks. This repository hosts the fixed case JSON files used by the benchmark. Each JSON file contains the agent input, parsing instructions, ground truth, and evaluation metrics for one benchmark instance.
No original EEG or PSG recordings are distributed in this repository. Users must obtain the required source datasets from their official providers and prepare them… See the full description on the dataset page: https://huggingface.co/datasets/xbb083/BrainBench.TopoSense-Bench
TopoSense-Bench: A Campus-Scale Benchmark for Semantic-Spatial Sensor Scheduling
TopoSense-Bench is a large-scale, rigorous benchmark designed to evaluate Large Language Models (LLMs) and agents on the Semantic-Spatial Sensor Scheduling (S³) problem. It features a realistic digital twin of a university campus equipped with 2,510 cameras and contains 5,250 natural language queries grounded in physical topology.
This dataset is the official benchmark for the ACM MobiCom 2026 paper:… See the full description on the dataset page: https://huggingface.co/datasets/IoT-Brain/TopoSense-Bench.BrainTRACE
BrainTRACE — Brain MRI Tracking, Reasoning, Annotation & Comparison Evaluation
A vision-language benchmark of 6,923 task definitions (7,273 scored VQA instances) over the upstream MR-RATE longitudinal brain MRI dataset.
⚠️ What BrainTRACE redistributes (and what it does not). BrainTRACE is not a re-publication of MR-RATE. The contributions released here are the task definitions — questions, ground-truth values, multi-slot rubrics, per-step chain rubrics, and per-item pointers to… See the full description on the dataset page: https://huggingface.co/datasets/BrainTRACE-anon/BrainTRACE.MedQADataFrenchWilliams
MedQAData-FR
French-focused medical Q&A dataset derived from MedQAData-v2.
Field
Description
context_question
Full patient narrative
question
Short direct question reformulated from context (1 line, ends with ?)
answer
Concise doctor answer (filler removed)
language
French
urgency
low / medium / high / critical
speciality
Medical specialty
article_title
Reference article title
entities
Dict with keys age, medicament, sympt, medical_field, disease, Test (each… See the full description on the dataset page: https://huggingface.co/datasets/BrainHealthAI/MedQADataFrenchWilliams.code-training-il
Code-Training-IL
A 40,330-example instruction-tuning dataset for code: 20K Python (NVIDIA OpenCodeInstruct, test-filtered) + 20K TypeScript + 330 hand-written bilingual identity examples.
Overview
code-training-il is a curated, filtered instruction-tuning corpus for training small coding assistants. It is the dataset used to fine-tune code-il-E4B, a 4B on-device model.
The dataset was designed around a thesis: less data, better filtered, beats more data. The… See the full description on the dataset page: https://huggingface.co/datasets/BrainboxAI/code-training-il.MedCortex-v1
MedCortex — Bilingual Medical Reasoning + Consultation Corpus
86,006 provenance-tracked, decontaminated medical examples in one uniform schema, fusing two
complementary strengths without flattening either:
task_type
Rows
What it is
Why it is here
reasoning
44,736
Verified chain-of-thought (KG-grounded + verified CoT), English
Drives exam-style benchmark reasoning — the source of the MedReason paper's measured gains
consultation
41,270
Bilingual (EN/FR) clinical Q&A… See the full description on the dataset page: https://huggingface.co/datasets/BrainHealthAI/MedCortex-v1.medical-training-il
Medical-Training-IL
A bilingual (Hebrew / English) medical instruction-tuning corpus — curated for training small, on-device medical models for Israeli residents preparing for Stage A exams.
Overview
medical-training-il is a curated, bilingual medical instruction-tuning dataset designed to fine-tune language models for Israeli clinical reasoning. It combines high-quality English medical QA (USMLE-style, basic sciences, research-grounded) with ~5,000 Hebrew-native… See the full description on the dataset page: https://huggingface.co/datasets/BrainboxAI/medical-training-il.MedQADataEnglishWilliams
MedQAData-EN
English-focused medical Q&A dataset derived from MedQAData-v2.
Field
Description
context_question
Full patient narrative
question
Short direct question reformulated from context (1 line, ends with ?)
answer
Concise doctor answer (filler removed)
language
English
urgency
low / medium / high / critical
speciality
Medical specialty
article_title
Reference article title
entities
Dict with keys age, medicament, sympt, medical_field, disease, Test… See the full description on the dataset page: https://huggingface.co/datasets/BrainHealthAI/MedQADataEnglishWilliams.brainly
brainly.co.id dataset
Data Structure
The keys in each JSONL object include:
"id": An integer value representing the page of task from url (e.g. brainly.co.id/tugas/117).
"subject": A string indicating the subject of the question (e.g., "Fisika", "Matematika", "Sejarah").
"author": A string representing the author of the question.
"instruction": A string providing the instruction or prompt for the question.
"answerer_1", "answer_2": Strings representing the answerers for… See the full description on the dataset page: https://huggingface.co/datasets/lesserfield/brainly.BrainMedCoT
BrainMedCoT — Trilingual Medical Chain-of-Thought Dataset
BrainMedCoT is a trilingual (French / English / Arabic Darija) medical Q&A dataset enriched
with structured chain-of-thought reasoning (<think> block), grounded in real biomedical
sources (PubMed / RxNorm / DailyMed / MedlinePlus). It is the CoT fine-tuning stage of the
HELIX-FT medical-LLM curriculum (SFT → SASR/GRPO → CoT).
Stat
Value
Total examples
3458
Splits (train / val / test)
2768 / 345 / 345
Source… See the full description on the dataset page: https://huggingface.co/datasets/BrainHealthAI/BrainMedCoT.MedQA-Darija-MCQ-500
MedQA-Darija-MCQ-500 — Moroccan Darija Medical MCQ Benchmark
A 500-question multiple-choice benchmark for medical reasoning in Moroccan Arabic
Darija. To our knowledge this is the first publicly released medical MCQ benchmark in
Darija — built specifically because no equivalent existed and Darija-medical LLM
evaluation had nowhere to land.
Stat
Value
Items
500
Language
Moroccan Arabic Darija
Format
4-option MCQ (A/B/C/D, single correct)
Source stems… See the full description on the dataset page: https://huggingface.co/datasets/BrainHealthAI/MedQA-Darija-MCQ-500.MedQADataEnglishSaad
Health QA English — Medical Question Answering Dataset
Dataset Description
A curated dataset of 13,812 medical question-answer pairs sourced from real patient-doctor consultations. Each entry contains a patient's clinical scenario, a focused medical question, and a doctor's professional response, enriched with named medical entities (symptoms, diseases, medications, tests).
Key Features
13,812 high-quality entries across 15 medical specialties
Structured… See the full description on the dataset page: https://huggingface.co/datasets/BrainHealthAI/MedQADataEnglishSaad.MedQA_Multi
MedQA_Multi
A high-quality, cleaned, and deduplicated bilingual (English and French) medical QA dataset based on MedQA, enriched with medical specialty metadata.
Dataset Structure
The dataset contains two splits:
train: 36,909 examples
test: 4,645 examples
Schema
Each entry contains the following fields:
question (string): The clinical query or question.
context_question (string): Additional clinical context or patient case details.
answer… See the full description on the dataset page: https://huggingface.co/datasets/BrainHealthAI/MedQA_Multi.Brainteaser
The Brainteaser dataset
This is a dataset concurrently released with our paper
Creativity or Brute Force? Using Brainteasers as a Window into the Problem-Solving Abilities of Large Language Models, NeurIPS 2025.
Citation
If you use this dataset, please cite our paper:
@article{han2025creativity,
title={Creativity or Brute Force? Using Brainteasers as a Window into the Problem-Solving Abilities of Large Language Models},
author={Han, Simeng and Dai, Howard and… See the full description on the dataset page: https://huggingface.co/datasets/ChenLiu1996/Brainteaser.legal-training-il
Legal-Training-IL
A 17,613-example bilingual instruction-tuning corpus for Israeli legal reasoning — covering rulings, statutes, citizen-rights pages, and contract clauses.
Overview
legal-training-il is a curated, bilingual (Hebrew / English) instruction-tuning dataset designed to adapt general-purpose language models to Israeli legal work. It was built to train law-il-E2B, a 2B-parameter on-device legal assistant.
The dataset is not a scraped dump. Every example… See the full description on the dataset page: https://huggingface.co/datasets/BrainboxAI/legal-training-il.MedUnified
MedUnified — Trilingual Medical SFT Mix (HELIX-FT v2)
MedUnified is the continued-SFT data mix for the second iteration of the HELIX-FT
medical LLM (BrainHealthAI/MedQA-Llama3.1-8B-HELIX-v2). It unifies five complementary
medical sources — real and synthetic, English / French / Moroccan Darija — into one
single-language-per-row training corpus, decontaminated against the standard medical eval
benchmarks.
Stat
Value
Total rows
23 000 (22 505 train + 495 validation)… See the full description on the dataset page: https://huggingface.co/datasets/BrainHealthAI/MedUnified.livenewsbench-search-arms
LiveNewsBench Search Arms
Paired measurements of four language models answering the same 1,329 news
questions under four retrieval conditions. Every question was run in every
condition, so each row pairs with 13 others on task_key.
The release answers one question: how much of an agent's answer quality comes
from the model, and how much from the search system wrapped around it.
This is a derivative evaluation-results dataset, not the original
LiveNewsBench benchmark. The… See the full description on the dataset page: https://huggingface.co/datasets/BraintrustDataDev/livenewsbench-search-arms.MedQADataDarijaSaad
Health QA Darija — Medical QA in Moroccan Arabic (الدارجة المغربية)
Dataset Description
A curated dataset of 8,129 medical question-answer pairs in Moroccan Darija (الدارجة المغربية). Each entry contains a patient scenario, a focused medical question, and a doctor's response — all in authentic Darija. Enriched with named medical entities (symptoms, diseases, medications, tests).
🇲🇦 First large-scale medical QA dataset in Moroccan Darija — addressing the critical gap in… See the full description on the dataset page: https://huggingface.co/datasets/BrainHealthAI/MedQADataDarijaSaad.BatVenom
BatVenom: Dual-Personality Roleplay Dataset 🦇🕷️
This dataset contains over 200+ hand-crafted and AI-assisted roleplay scenarios designed to fine-tune Large Language Models (LLMs) into the "BatVenom" persona—a hybrid of Batman (Bruce Wayne) and the Venom Symbiote.
📊 Dataset Structure
The data is provided in the Alpaca/LLaMA-Factory format:
instruction: The context or setup of the scene.
input: The specific user prompt or dialogue.
output: The formatted response showing… See the full description on the dataset page: https://huggingface.co/datasets/BrainDelay/BatVenom.hermes-brain-bench-v0.1
Hermes Brain Bench v0.1
A small, diverse, graded-rubric benchmark for picking the agent brain behind
a local-only Spark deployment of an OpenAI-compatible tool-using assistant
(developed against Hermes Agent, but
the suite is harness-agnostic — any OpenAI-tool-format runner works).
The bench answers a question that single-stream throughput benchmarks can't:
which local serving lane actually produces the more correct agent — under the
same rubric, run-to-run?
Companion artifacts… See the full description on the dataset page: https://huggingface.co/datasets/Orionfold/hermes-brain-bench-v0.1.brain-vqa-radBrain_Helix