datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
dexgarmentlab-lift-correspondence-20260822
DexGarmentLab Lift Correspondence 2026-08-22
This is the point-cloud correspondence dataset used by the UniGarmentManip
flat/deformation representation experiment pinned at commit
0701bc475877da68bc07f727a1995fd78a3de511, together with the full-state
trajectory HDF5 used to train UniClothDiff models.
It contains 1,800 bimanual lift episodes (1,687 successful, 113 recorded
failures) spanning tops, dresses, and trousers with garment-disjoint
train/validation/test splits (see… See the full description on the dataset page: https://huggingface.co/datasets/Cloth-splatters/dexgarmentlab-lift-correspondence-20260822.enron-correspondence-dedup
Enron Correspondence Deduplicated (Enriched GT, Agent-Blind Default)
The deduplicated, ground-truth-enriched Enron correspondence benchmark:
exact-duplicate bodies removed from the cleaned CMU Enron corpus (517,390
rows in → 247,523 unique-text rows out, 269,867 duplicates dropped; first
occurrence wins on maildir-path order; empty bodies never deduped against
each other). This dataset is the core evaluation corpus for the
LLM Mailroom agent-sorting stack.
⚠️… See the full description on the dataset page: https://huggingface.co/datasets/Lucius-Morningstar/enron-correspondence-dedup.numerology-cosmos-correspondences
Numerology–Cosmos Correspondences, Provenance-Tagged
Each numerology number (1–9, 11, 22, 33) mapped to its traditional planet, weekday, metal, element and constellation — with every claim tagged by confidence tier.
A reference table mapping each numerology number to its classical correspondences (planet, weekday, metal, element, constellation). What makes it unusual: every cell is tagged with a provenance tier — A (factual), B (classical tradition), C (modern convention), D… See the full description on the dataset page: https://huggingface.co/datasets/alexdrago/numerology-cosmos-correspondences.sdg-bertopic-correspondence-v0.1
SDG BERTopic Input-Intermediate-Output Correspondence (v0.1)
Status: v0.1, peer-review preview. Schema and methodology may evolve.
Curator: @zndx
Project: Aegir — https://github.com/<...> (link pending public release)
A research dataset capturing the full computational trace from
constrained-decode-generated ontology compositions through BERTopic
intermediate state to verifier scoring. Built for reproducibility,
verifier-methodology validation, and as a paired data resource for… See the full description on the dataset page: https://huggingface.co/datasets/zndx/sdg-bertopic-correspondence-v0.1.robinson_schensted_knuth_correspondence_9
The Robinson-Schensted-Knuth Correspondence for Permutations of Size 9
The Robinson-Schensted-Knuth (RSK) algorithm [1,2] gives a bijection between
pairs of semistandard Young tableau of the same shape and matrices with non-negative
integer entries. The special case we consider (which is sometimes called the
Robinson-Schensted algorithm) restricts to a bijection between pairs of standard Young
tableaux and permutations in SnS_nSn. This correspondence is highly significant in… See the full description on the dataset page: https://huggingface.co/datasets/ACDRepo/robinson_schensted_knuth_correspondence_9.clinical-regulatory-evidence-correspondence-v0.1
What this dataset tests
Regulatory claims must map to evidence scope.
Population boundaries matter.
Why it exists
Regulatory language can drift.
Indications expand.
Subgroups disappear.
This set detects when claims exceed the evidence base.
Data format
Each row contains
evidence_base
regulatory_claim
correspondence_pressure
constraints
failure_modes_to_avoid
target_behaviors
gold_checklist
Feed the model
evidence_base
regulatory_claim
Score… See the full description on the dataset page: https://huggingface.co/datasets/ClarusC64/clinical-regulatory-evidence-correspondence-v0.1.clinical-guideline-strength-correspondence-v0.1
What this dataset tests
Guideline strength must track evidence strength.
Authority must not exceed data.
Why it exists
Guidelines often harden too early.
Language outruns certainty.
This set checks whether recommendation force matches evidence quality.
Data format
Each row contains
evidence_profile
guideline_recommendation
strength_pressure
constraints
failure_modes_to_avoid
target_behaviors
gold_checklist
Feed the model
evidence_profile… See the full description on the dataset page: https://huggingface.co/datasets/ClarusC64/clinical-guideline-strength-correspondence-v0.1.robinson_schensted_knuth_correspondence_8
The Robinson-Schensted-Knuth Correspondence for Permutations of Size 8
The Robinson-Schensted-Knuth (RSK) algorithm [1,2] gives a bijection between
pairs of semistandard Young tableau of the same shape and matrices with non-negative
integer entries. The special case we consider (which is sometimes called the
Robinson-Schensted algorithm) restricts to a bijection between pairs of standard Young
tableaux and permutations in SnS_nSn. This correspondence is highly significant in… See the full description on the dataset page: https://huggingface.co/datasets/ACDRepo/robinson_schensted_knuth_correspondence_8.robinson_schensted_knuth_correspondence_10
The Robinson-Schensted-Knuth Correspondence for Permutations of Size 10
The Robinson-Schensted-Knuth (RSK) algorithm [1,2] gives a bijection between
pairs of semistandard Young tableau of the same shape and matrices with non-negative
integer entries. The special case we consider (which is sometimes called the
Robinson-Schensted algorithm) restricts to a bijection between pairs of standard Young
tableaux and permutations in SnS_nSn. This correspondence is highly significant in… See the full description on the dataset page: https://huggingface.co/datasets/ACDRepo/robinson_schensted_knuth_correspondence_10.Real_World_Written_German_Correspondence_Echter_deutscher_SchriftverkehrSynthetic-Visual-Correspondence-Data
Synthetic Visual Correspondence Data
Dataset for the BLINK benchmark visual correspondence task. Contains 50,000 image pairs where Image 1 has a reference point and Image 2 has 4 candidate points (A-D). The task is to identify which candidate point matches the reference point.
Dataset Summary
Total Samples: 50,000
Image Size: Height 1024 (aspect ratio preserved)
Format: Parquet files (HuggingFace Datasets compatible)
Composition
Multi-Illumination-Dataset:… See the full description on the dataset page: https://huggingface.co/datasets/mavleo96/Synthetic-Visual-Correspondence-Data.task281_points_of_correspondence
Dataset Card for Natural Instructions (https://github.com/allenai/natural-instructions) Task: task281_points_of_correspondence
Additional Information
Citation Information
The following paper introduces the corpus in detail. If you use the corpus in published work, please cite it:
@misc{wang2022supernaturalinstructionsgeneralizationdeclarativeinstructions,
title={Super-NaturalInstructions: Generalization via Declarative Instructions on 1600+ NLP Tasks}… See the full description on the dataset page: https://huggingface.co/datasets/Lots-of-LoRAs/task281_points_of_correspondence.momagraph_correspondenceflan_combined_task281_points_of_correspondence