datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
phase3-dataset
Dataset
Generated dataset with 120 configurations.
Configuration
dataset_name: phase3-dataset
graphs_path: hf://CSE472-blanket-challenge/phase3-graphs
output_path: data/datasets/${dataset_name}
n_samples: 1000
n_datasets: 1
scm_type: linear, nonlinear
coeff_range: 1.0
noise_std: 0.5
env_type: iid, covariate, label
projection: pca
shift_mean: 0.8
shift_std: 0.2
train_fraction: 0.8
seed: 42
overwrite: true
Load data
from huggingface_hub import snapshot_download… See the full description on the dataset page: https://huggingface.co/datasets/CSE472-blanket-challenge/phase3-dataset.LMA_phase3_data
LMA Phase 3 — transitive reasoning dataset (Hindi / Nepali)
Synthetic transitive-comparison problems over people, in Hindi and Nepali, with
chain-of-thought targets. Built for a study of data scaling and domain
generalization in small from-scratch language models.
The models finetuned on it are at
Prateek-Tiwari10/LMA_phase3.
Layout
manifest_scaling.json machine-readable index of every file below
train/{hi,ne}/train_{10,30,50,60}k.jsonl nested… See the full description on the dataset page: https://huggingface.co/datasets/Prateek-Tiwari10/LMA_phase3_data.phase3-review-cookies-friedrice2phase3-graphs
Phase 3 Graphs
Graphs used in Phase 3
Data Card
Field
Type
Description
graph_id
string
Unique identifier for the graph
adjacency_matrix
List[List[int]]
NxN binary adjacency matrix where A[i,j]=1 means i→j
num_nodes
int
Number of nodes in the DAG
num_edges
int
Number of edges in the DAG
density
float
Graph density (edges / max_possible_edges)
method
string
Graph generation method (e.g., "PA", "ER")
Graph Generation Settings
Generation… See the full description on the dataset page: https://huggingface.co/datasets/CSE472-blanket-challenge/phase3-graphs.visrl-14b-phase3e-step3814-resumesublim-phase3-teacher-dataorh-exp1-phase3clinicaltrials-completed-phase3
clinicaltrials-completed-phase3-full — Data Card
Version: 1.1.0 | Generated: 2026-06-14T06:18:47.568731+00:00 | Records: 29,633
Dataset Summary
clinicaltrials-completed-phase3-full contains structured clinical trial records from ClinicalTrials.gov, the U.S. federal registry of clinical studies operated by the National Library of Medicine (NIH). Each record includes trial status, design parameters, eligibility criteria, sponsor information, enrollment counts, and… See the full description on the dataset page: https://huggingface.co/datasets/RubyIntelligence/clinicaltrials-completed-phase3.phase3phase3_train_datasetphase3_train_dataset_0.05phase_3finetuning_phase3
Curated Datasets for Collusion Experiments
Non-overlapping datasets for fine-tuning and evaluation with full provenance tracking.
Files
File
Purpose
Rows
Unique Problem IDs
finetuning_1500.json
Primary fine-tuning dataset
1500
1000
finetuning_sample_500.json
Original 500 with all columns (archive)
500
500
finetuning_1000.json
Previous version (superseded)
1000
1000
evaluation_v7.json
Evaluate collusion resistance
1000
700
llama70b_detectable_128.json… See the full description on the dataset page: https://huggingface.co/datasets/jprivera44/finetuning_phase3.phase3_train_dataset_hard_maxphase3_train_dataset_0.5phase3_train_dataset_0.0phase3_train_dataset_hard_sample_0.7Phase3aiws5.3_phase3