datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
EpiBench-NeurIPS2026
EpiBench
Anonymous release for NeurIPS 2026 Evaluations & Datasets Track review (paper ID 1899). All methodology, ablations, and analyses are in the companion paper; this card lists only what reviewers and downstream users need to load the data.
A 25,737-patient ILAE-aligned multimodal epilepsy benchmark derived from PubMed Central case reports + 192 EpiRAG textbook vignettes.
6 tasks: epilepsy_type, seizure_type, ez_localization, aed_response, surgery_outcome, status_epilepticus… See the full description on the dataset page: https://huggingface.co/datasets/NeurIPS-1899-ED-2026/EpiBench-NeurIPS2026.NeurIPS_2026_BLV
BLV Object Recognition: Synthetic + Real-World
A dataset for training and evaluating object recognition and segmentation
models on infrastructure relevant to blind and low-vision (BLV) navigation
in urban environments. Three configurations plus a flat tree of 3D assets:
Config / tree
Splits
Purpose
syn
train
Photorealistic IsaacSim renders for training / pretraining.
real_ours
train / validation / test
Real photographs we captured. real_ours/test is the canonical… See the full description on the dataset page: https://huggingface.co/datasets/NavAble/NeurIPS_2026_BLV.neurips-2026-evals
NeurIPS 2026 Agent Evaluation Dataset
This dataset contains evaluation results for various AI agents across multiple benchmarks.
Dataset Structure
The dataset is organized by model (as configs) with each benchmark as a split.
Each model/benchmark folder contains:
Main results file (.jsonl or .parquet format)
Summary statistics (.summary.json) - for models with metadata
Configuration file (.toml) - for models with metadata
Traces folder with execution traces… See the full description on the dataset page: https://huggingface.co/datasets/akenginorhun/neurips-2026-evals.IGF-Bench
IGF-Bench: Indoor Geometric Fidelity Benchmark
Anonymous mirror for NeurIPS 2026 Evaluations and Datasets Track double-blind review.
The de-anonymised author/maintainer information will replace this header at camera-ready.
IGF-Bench is the first benchmark for evaluating structural-level geometric fidelity of conditionally generated indoor scene images, going beyond perceptual metrics like FID and LPIPS. It pairs 3,600 calibrated synthetic ground-truth views with 21,600 generated… See the full description on the dataset page: https://huggingface.co/datasets/igfbench-neurips2026/IGF-Bench.ARES-Bench
ARES-Bench
ARES-Bench is the open audit substrate released with the paper
Auditing LLM User Simulators for Recommender A/B Testing (NeurIPS 2026, ED
Track, under review). It turns the ARES reliability-audit view — the LLM
backbone is the measurement instrument under test, not an interchangeable
implementation detail — into a reproducible protocol over structured
behavioral logs, a portable visual sandbox, and a screenshot cache.
This release hosts the 17,000-session core corpus that… See the full description on the dataset page: https://huggingface.co/datasets/neurips2026-ares-authors/ARES-Bench.PRISM-Dataset
PRISM: Polarimetric Road-surface Intelligent Sensing and Measurement Dataset
Anonymous submission to NeurIPS 2026 Evaluations & Datasets Track.
Author identities and the camera-ready release URL will be revealed at the
camera-ready stage.
PRISM is a polarimetric road-surface dataset and benchmark of 47,098
time-synchronized frames combining trichromatic linear polarization,
co-boresighted RGB, 128-channel LiDAR, and RTK-GNSS/INS, captured from an
in-vehicle sensor rig across… See the full description on the dataset page: https://huggingface.co/datasets/NeurIPS-2026-PRISM/PRISM-Dataset.causalverify-neurips2026
🎯 CausalVerify
An Execution-Grounded Benchmark for LLM Causal Inference Workflows
NeurIPS 2026 — Evaluations and Datasets Track · double-blind review · frozen at tag neurips2026-submission
💡 TL;DR
A benchmark of 259 published economics papers (Experiment A — real-paper text-agreement diagnostic) and 100 fixed-seed synthetic data-generating processes (Experiment B — execution-grounded coefficient recovery), evaluating 7 frontier LLMs. The central… See the full description on the dataset page: https://huggingface.co/datasets/causalverify/causalverify-neurips2026.NeurIPS2026-DermVerse-500K
DermVerse-500K
Paper: DermVerse-500K: A Large-Scale Expert-Annotated Dataset with VLM Assisted Clinical Feature ExtractionSubmitted to: 40th Conference on Neural Information Processing Systems (NeurIPS 2026)Paper ID: 4127
Dataset Summary
DermVerse-500K is a large-scale dermatology dataset comprising 500,000 clinical images paired with fine-grained, structured clinical descriptors. It is designed to address critical limitations in existing dermatology AI resources:… See the full description on the dataset page: https://huggingface.co/datasets/BBAnonymous/NeurIPS2026-DermVerse-500K.crychic-dafny-acsl
CRYCHIC Dafny-to-ACSL-C Verified Translation Benchmark
This anonymized review artifact accompanies the NeurIPS 2026 Evaluations and Datasets submission:
CRYCHIC: A Universal Framework for Cross-Language Verified Code Translation.
CRYCHIC translates verified Dafny programs into C programs annotated with ACSL specifications, then checks the generated artifacts with Frama-C WP. This release contains the 1,679 fully verified Dafny/C+ACSL pairs used as the positive benchmark corpus.
The… See the full description on the dataset page: https://huggingface.co/datasets/neurips2026-crychic/crychic-dafny-acsl.BenchRep-T
BenchRep-T
BenchRep-T is a benchmark for TCRβ (T-cell receptor β-chain) immune-repertoire classification.
It bundles the multi-disease Mal-ID cohort together with five external disease cohorts
(sequenced on the Adaptive Biotechnologies immunoSEQ platform), providing per-specimen
repertoire files, harmonized sample metadata, and auxiliary files used for the
sequencing-depth scaling-law and driver-sequence experiments.
Each specimen (one sequenced repertoire) is one example; the… See the full description on the dataset page: https://huggingface.co/datasets/neurips-2026-dataset/BenchRep-T.tdbench-review
TDBench: Benchmarking Vision-Language Models on Top-Down Images
Note (Anonymous Review Version). This dataset card accompanies a NeurIPS 2026 Evaluations & Datasets double-blind submission. Author identifiers, institutional affiliations, project pages, and external repository links have been removed for the review period. The full set of public artifacts and the final citation will be restored upon decision.
Overview
TDBench is a benchmark for evaluating… See the full description on the dataset page: https://huggingface.co/datasets/neurips2026-tdbench/tdbench-review.example_difficulty_scores
Precomputed scores from Dataset Difficulty and the Role of Inductive Bias
checkpoints end with .pt
all scores are saved in .npz files
scores with variances computed over training are stored under the keys 'mean' and 'var', otherwise they are under the key 'arr_0'
precomputed contains scores from Carlini, Erlingsson & Papernot (2019) [1], Feldman & Zhang (2020) [2], and Sorscher et al. (2022) [3]. This directory can be placed in src/difficulty if using code from… See the full description on the dataset page: https://huggingface.co/datasets/anonymous-neurips2026-0383/example_difficulty_scores.VSTAT
VSTAT: Visual State Tracking Benchmark
VSTAT is a video-based benchmark for evaluating the visual state tracking
capability of Multimodal Large Language Models (MLLMs). It contains 813 video
clips paired with 1,479 questions whose answers cannot be inferred from any
single keyframe or short segment.
Dataset Composition
Split
Videos
Questions
synthetic
450
550
self_recorded
80
100
youtube
283
830
Total
813
1,479
Files… See the full description on the dataset page: https://huggingface.co/datasets/VSTAT-NeurIPS2026/VSTAT.NeurIPS_2026_BLV_Subset
4 GB stratified preview. Full dataset: NavAble/NeurIPS_2026_BLV.
BLV Object Recognition: Synthetic + Real-World
A dataset for training and evaluating object recognition and segmentation
models on infrastructure relevant to blind and low-vision (BLV) navigation
in urban environments. Three configurations plus a flat tree of 3D assets:
Config / tree
Splits
Purpose
syn
train
Photorealistic IsaacSim renders for training / pretraining.
real_ours
train / validation / test… See the full description on the dataset page: https://huggingface.co/datasets/NavAble/NeurIPS_2026_BLV_Subset.RegressionPRISM-Dataset-Sample
PRISM Sample: Polarimetric Road-surface Intelligent Sensing and Measurement Dataset
Anonymous submission to NeurIPS 2026 Evaluations & Datasets Track.
This is a representative sample of the PRISM dataset, designed to enable reviewers and researchers to inspect data quality without downloading the full ~1.6 TB dataset.
Why a sample dataset?
The full PRISM dataset contains 47,098 time-synchronized frames across 41 sessions. This sample provides:
Quick quality inspection:… See the full description on the dataset page: https://huggingface.co/datasets/NeurIPS-2026-PRISM/PRISM-Dataset-Sample.NeurIPS2026-Genericaudit-release
Structural T2I Verifier Protocol Audit Dataset
This artifact contains the generated images, human annotations, prompt metadata, verifier outputs, and detector artifacts needed by the companion code repository. Mount or copy this directory as ./dataset when running the code.
The dataset is intended for research and audit of structural text-to-image
evaluation protocols. It should be used with the companion code artifact or with
independent analysis code that preserves the same… See the full description on the dataset page: https://huggingface.co/datasets/neurips2026-ed443/audit-release.LEDE-dataset
LEDE : A large-scale benchmark for AI-generated news detection
Anonymous submission to NeurIPS 2026 Evaluations & Datasets Track. Author identities and the camera-ready release URL will be revealed at the camera-ready stage.
LEDE is a large-scale, AI-generated news detection benchmark dataset comprising 337K articles and over 4.3M sentences. It addresses the limitations of existing benchmarks by providing broader generator diversity and news-specific coverage across 21… See the full description on the dataset page: https://huggingface.co/datasets/NeurIPS-2026-LEDE/LEDE-dataset.opentouch-samplesob
The Structured Output Benchmark (SOB)
A multi-source benchmark for evaluating structured-output quality in LLMs.
(Anonymous submission — links to code, paper, and leaderboard withheld during double-blind review.)
Dataset summary
SOB evaluates how accurately LLMs produce schema-compliant and value-correct JSON from unstructured or semi-structured context — across three source modalities:
Config
Source
Context delivered as
Records
default
HotpotQA (multi-hop QA)… See the full description on the dataset page: https://huggingface.co/datasets/neurips2026sob/sob.LR-Transfer-Trajectory
Dataset Documentation
Overview
This dataset captures per-step training and validation metrics from training runs of a 12-layer GPT-style decoder-only transformer. Each run is stored as a single .csv file in which every row corresponds to one logged step, and several columns hold per-parameter measurements encoded as JSON.
The dataset is designed to support post-hoc analysis of:
Loss curves (train / val)
Throughput and step latency
Per-layer / per-parameter dynamics:… See the full description on the dataset page: https://huggingface.co/datasets/neurips2026lrtransfer/LR-Transfer-Trajectory.signalbench-openapps
SignalBench OpenApps Dense Signal Dataset
OpenApps browser-task evaluation points, Set-of-Marks screenshots, and Monte Carlo dense-signal labels under a scripted policy.
This repository contains one SignalBench dataset with two synchronized views:
runtime/dataset.pkl is the executable artifact used by the SignalBench benchmark code.
data/examples.parquet has exactly one row per benchmark example, with
state/action/next-state text, renderable state_image and
next_state_image columns… See the full description on the dataset page: https://huggingface.co/datasets/neurips2026-anonymous/signalbench-openapps.signalbench-alfworld-bkp
SignalBench ALFWorld Dense Signal Dataset
ALFWorld household-task evaluation points, visual state observations, and Monte Carlo dense-signal labels under a scripted policy.
This repository contains one SignalBench dataset with two synchronized views:
runtime/dataset.pkl is the executable artifact used by the SignalBench benchmark code.
data/examples.parquet has exactly one row per benchmark example, with
state/action/next-state text, renderable state_image and
next_state_image… See the full description on the dataset page: https://huggingface.co/datasets/neurips2026-anonymous/signalbench-alfworld-bkp.DataShare_NeurIPS2026_LLMsRiskAttitude
Risk Attitude in LLMs: A Cross-Task Behavioral Dataset
Dataset accompanying the NeurIPS 2026 paper:
"[ANONYMIZED FOR REVIEW]"
Authors: [ANONYMIZED FOR REVIEW]
This dataset contains LLM response data and human baseline data collected across
three structured decision tasks designed to measure risk attitude and
behavioral reliability of contextual belief and risk decisions.
Tasks
Abbreviation
Full Name
Description
CTD
Clinical Triage Decision
Model acts… See the full description on the dataset page: https://huggingface.co/datasets/LLMsRiskAttitudeDataShare/DataShare_NeurIPS2026_LLMsRiskAttitude.formal-anytime-valid-stats
Formal-AVS: A Lean Benchmark for Anytime-Valid Confidence-Sequence Theorem Proving
60 Lean 4 theorem targets on anytime-valid confidence sequences across four families (Howard-Ramdas, betting, Whitehouse vector, asymptotic CLT).
Benchmark Structure
60 targets grouped into tiers T0-T3 (pre-evaluation) and categories T4-T5 (empirical)
7 drafters evaluated across single-shot, agentic, and unbounded modes
14 Aristotle sessions (unbounded refinement)
Headline Results… See the full description on the dataset page: https://huggingface.co/datasets/neurips-2026-avs-bench/formal-anytime-valid-stats.opentouchKGQA
Compressed Multi-Hop KGQA Benchmark
Dataset Summary
This dataset is a multiple-choice benchmark for evaluating compressed multi-hop reasoning over knowledge graphs.
Files
data/benchmark.jsonl: main benchmark file.
data/question_graphs.jsonl: candidate graph/path evidence for each question.
Data Fields
Each benchmark example contains:
question_id: unique question identifier
type: question type
question: natural language question
given_entity:… See the full description on the dataset page: https://huggingface.co/datasets/neurips2026-ed-dataset/KGQA.ThinkSafe-R1-Distill-7Bsignalbench-frozenlake-bkp
SignalBench FrozenLake Dense Signal Dataset
FrozenLake 8x8 evaluation points, rendered state images, and Monte Carlo dense-signal labels under a scripted policy.
This repository contains one SignalBench dataset with two synchronized views:
runtime/dataset.pkl is the executable artifact used by the SignalBench benchmark code.
data/examples.parquet has exactly one row per benchmark example, with
state/action/next-state text, renderable state_image and
next_state_image columns when… See the full description on the dataset page: https://huggingface.co/datasets/neurips2026-anonymous/signalbench-frozenlake-bkp.