datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
NYT-Connections
🧩 NYT-Connections
This repository contains the NYT-Connections dataset proposed in the work NYT-Connections: A Deceptively Simple Text Classification Task that Stumps System-1 Thinkers. This work was published at the 31st International Conference on Computational Linguistics (COLING 2025) and was honored with the Best Dataset Paper award.
Authors: Angel Yahir Loredo Lopez, Tyler McDonald, Ali Emami
📜 Paper Abstract
Large Language Models (LLMs) have shown impressive… See the full description on the dataset page: https://huggingface.co/datasets/tm21cy/NYT-Connections.create_interpret
create_interpret — results and raw data
Every artefact behind ManyaWadhwa/create_interpret:
raw generations, parsed paths, judge caches, scores, and the two analyses built on them. Code,
method and write-ups live in the GitHub repo (Task A,
Task B); this repo is the data
those documents are computed from.
Task A — how CREATE creative utility moves with model and sampling temperature, and how diverse
16 samples of one query actually are. 50 instances of wadhma/CREATE
x 16 samples… See the full description on the dataset page: https://huggingface.co/datasets/connections-dev/create_interpret.hydraulic-pipe-connection-types
Hydraulic and pipe connection type identification
Canonical, always-current version: https://referencesource.org/hydraulic-pipe-connection-types/
Machine-readable: https://referencesource.org/hydraulic-pipe-connection-types/data.json — this mirror is a point-in-time copy.
Last verified: 2026-08-13
Stale after: 2028-08-12 (past this date, prefer the canonical copy —
it re-verifies on a cadence this snapshot does not)
Records: 21
Identifying characteristics of each hydraulic… See the full description on the dataset page: https://huggingface.co/datasets/referencesource/hydraulic-pipe-connection-types.nyt-connections-cot
Connections & Linguistic Reasoning Dataset
Dataset Summary
A comprehensive dataset for training models on word association puzzles and linguistic reasoning, featuring two complementary components:
NYT Connections-style puzzles with human-like solving narratives
Linguistic pattern puzzles focusing on sophisticated wordplay and etymology
Both datasets include detailed chain-of-thought reasoning demonstrating step-by-step problem-solving approaches.
Dataset… See the full description on the dataset page: https://huggingface.co/datasets/nickting/nyt-connections-cot.nyt-connections-predictions
NYT Connections Predictions
This dataset contains model predictions from fine-tuned language models evaluated on NYT Connections puzzles. These predictions correspond to the experiments described in the NYT Connections Experiments Dataset.
Dataset Overview
This dataset contains prediction outputs from 11 experimental runs testing different training configurations:
831 NYT Puzzles: 673 training, 74 validation, 84 test
200 Synthetic Puzzles: 162 training, 18 validation, 20… See the full description on the dataset page: https://huggingface.co/datasets/nickting/nyt-connections-predictions.