datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
chess-puzzlesPuzzleWorld
Dataset Card for PuzzleWorld
PuzzleWorld is a benchmark of 667 real-world puzzlehunt–style problems designed to evaluate open-ended, multimodal reasoning capabilities of AI models. Curated from Puzzled Pint’s Creative Commons–licensed archives (2010–2025), each puzzle combines text, visual, and structured inputs with no explicitly stated instructions. Solvers must first infer the hidden problem structure from ambiguous clues and then execute a multi-step, creative reasoning… See the full description on the dataset page: https://huggingface.co/datasets/hzli1202/PuzzleWorld.prop_logic_puzzlevisual-puzzlespuzzle-hle-filteredchess-puzzles
Dataset Card for Lichess Puzzles
Dataset Description
6,100,960 puzzles, rated and tagged. See them in action on Lichess.
This dataset is updated monthly, and was last updated on September 7th, 2026.
Dataset Creation
Generating the initial dataset chess puzzles took more than 50 years of CPU time. We went through 300,000,000 analyzed games from the Lichess database, and re-analyzed interesting positions with Stockfish 12/13/14/15 NNUE at 40… See the full description on the dataset page: https://huggingface.co/datasets/Lichess/chess-puzzles.GridCorpus_9M_Sudoku_Puzzles_Enriched
╔══════════════════════════════════════════════════════════════════════╗
║ ║
║ G R I D C O R P U S ║
║ ║
║ "004300209005009001070060043..." ║
║ │ ║
║ ▼… See the full description on the dataset page: https://huggingface.co/datasets/beta3/GridCorpus_9M_Sudoku_Puzzles_Enriched.Puzzle-KD-Nemotron-Post-Training-Dataset-v2
Puzzle-KD-Nemotron-Post-Training-Dataset-v2 Release
Dataset Overview
The Puzzle-KD-Nemotron-Post-Training-Dataset-v2 dataset is a curated and filtered subset of NVIDIA’s Nemotron-Post-Training-Dataset-v2.
The original dataset was released by NVIDIA in August 2025 as part of the NVIDIA Nemotron Nano 9B model family and was designed to
improve post-training alignment for reasoning, math, code, STEM, and chat capabilities across multiple languages.
It contains synthetic… See the full description on the dataset page: https://huggingface.co/datasets/nvidia/Puzzle-KD-Nemotron-Post-Training-Dataset-v2.nvarc-artifacts-puzzlespuzzle-4x4-play-10m-v0
Puzzle-4x4-Play 10M Dataset
This dataset contains 10M transitions for the puzzle-4x4-play task from OGBench. It was used for the experiments in the paper Trust Region Q Adjoint Matching.
Project Page | Code | Paper
Dataset Structure
10 files of ~1M transitions each (split for easier loading)
Each file includes:
Training set: 1,000 episodes × 1,001 steps
Validation set: 100 episodes × 1,001 steps
Observation dim: 83, Action dim: 5
Files… See the full description on the dataset page: https://huggingface.co/datasets/yonghoon96/puzzle-4x4-play-10m-v0.PuzzleVQAPaper | Code | Dataset
About
Large multimodal models extend the impressive capabilities of large language models by integrating multimodal
understanding abilities. However, it is not clear how they can emulate the general intelligence and reasoning ability of
humans. As recognizing patterns and abstracting concepts are key to general intelligence, we introduce PuzzleVQA, a
collection of puzzles based on abstract patterns. With this dataset, we evaluate large multimodal models with… See the full description on the dataset page: https://huggingface.co/datasets/declare-lab/PuzzleVQA.8-puzzle
8-puzzle / 3x3 sliding puzzle
Fixed 3x3 / 8-puzzle evaluation data and teacher-rollout datasets for sliding-puzzle reasoning experiments.
Configs
eval: 31 fixed evaluation puzzles, with exactly one puzzle at every optimal distance from 1 through 31.
sft-source: 200 fresh boards, balanced with 20 boards at each exact depth 10-19. It excludes the reserved evaluation boards and deliberately contains no solution-action labels.
sft: 3,745 replay-verified, fully… See the full description on the dataset page: https://huggingface.co/datasets/saad1926q/8-puzzle.chess-puzzles-images-mini
Dataset Card for Chess Puzzles Images (mini)
This dataset contains 124,999 chess board positions in JPG format, derived from Lichess puzzles. Each image is accompanied by a shortened FEN string, indication for the color to play as, castling and en passant availability, and best moves in standard algebraic notation.
The fields are as follows:
image: image, A visual representation of the chess board showing the current piece arrangement.
board_state: string, A shortened FEN… See the full description on the dataset page: https://huggingface.co/datasets/bingbangboom/chess-puzzles-images-mini.puzzlescript-gists
PuzzleScript Human-Authored Games (Full Gist Corpus)
35,704 human-authored PuzzleScript games — the
complete source text of each — collected from public GitHub gists.
This is the full corpus: every distinct gist is kept, and each row is tagged
with its deduplication cluster so you can reduce to a unique set with a one-line
filter. The deduplication is reproducible from the shipped dedup_master.json +
dedup_master.py; non-vanilla PuzzleScript-Plus files are excluded (listed in… See the full description on the dataset page: https://huggingface.co/datasets/smearle/puzzlescript-gists.chess_puzzle_benchmark
Chess Puzzle Benchmark
Chess puzzle evaluation sets at five difficulty tiers, B1 (easiest) through
B5 (hardest). Each example is a chess game given in PGN move notation; the
model must produce the next move(s).
Two prompt variants
The same puzzles are released in two forms that differ only in the prompt suffix:
think/ — the prompt ends with a special <T> token. <T> is a
reasoning trigger: it tells the model to think (produce a chain of reasoning)
before… See the full description on the dataset page: https://huggingface.co/datasets/pavelslab-nyu/chess_puzzle_benchmark.puzzle-map
Puzzle-Map Dataset
The Puzzle-Map Dataset is a collection of images and annotations for research on computer vision applied to jigsaw puzzles.
Directory Structure
masks/
*.png | *.jpg
masks-raw/
*.png | *.jpg
pieces/
*.png | *.jpg
annotations.json
puzzles/
*.png | *.jpg
annotations.json
Description
masks/ – Processed masks used for data augmentation.
masks-raw/ – Original masks used to generate the processed masks.… See the full description on the dataset page: https://huggingface.co/datasets/pablo-moreira/puzzle-map.Sudoku-1M-Puzzles
1M Sudoku Puzzles (9x9)
A dataset of 1,000,000 unique 9x9 Sudoku puzzles and their corresponding solutions. Each puzzle has between 40 and 60 blank cells, spanning various difficulty levels.
Dataset Structure
The dataset is provided in a single CSV file with the following columns:
puzzle: 81-character string representing the initial board (0 = blank).
solution: 81-character string representing the completed board.
Example Row
puzzle,solution… See the full description on the dataset page: https://huggingface.co/datasets/certen/Sudoku-1M-Puzzles.GAMBIT-lichess-puzzle-positions
GAMBIT: Generalization or Memorization? Brittleness Testing for Chess-Trained Language Models
Overview
Training/validation split of puzzles used in GAMBIT: Generalization or Memorization? Brittleness Testing for Chess-Trained Language Models
Parsed in text format from the Lichess Puzzle Database
sampling_log.txt — Displays statistics about samples for each theme (74 unique themes)
training-puzzle-positions.txt — Large set of puzzle positions as position + best… See the full description on the dataset page: https://huggingface.co/datasets/ethanjtang/GAMBIT-lichess-puzzle-positions.pencil-puzzle-bench
Pencil Puzzle Bench Dataset
This repository contains the puzzle datasets and benchmark results for Pencil Puzzle Bench.
Read the Paper | Website & Leaderboard
62,231 puzzles across 94 puzzle types with verified unique solutions.
Files
Puzzle Datasets
full_dataset.jsonl - Full dataset (62,231 puzzles)
golden_300.jsonl - 300 puzzles (20 types × 15 each) for standard evaluation
golden_30.jsonl - 30-puzzle subset for expensive/agentic strategies… See the full description on the dataset page: https://huggingface.co/datasets/bluecoconut/pencil-puzzle-bench.puzzles-for-vision-llmGLM-5.2-Logic-Puzzles
GLM-5.2 · Logical Puzzles
6000x traces distilled from GLM-5.2 on High reasoning
Token Count: 5M~?
Distribution:
Puzzles:
•Tokenization blindless ex: counting the r's in strawberry
•Goal reasoning ex: the car wash test (theres no car wash question exactly just prompts like it so its not just benchmaxxing)
•Reading comprehension traps
•Temporal reasoning
•Many other categories not worth mentioning
Prompts… See the full description on the dataset page: https://huggingface.co/datasets/ianncity/GLM-5.2-Logic-Puzzles.chess-puzzles-images-large
Dataset Card for Chess Puzzles Images (large)
This dataset contains 1,249,999 chess board positions in JPG format, derived from Lichess puzzles. Each image is accompanied by a shortened FEN string, indication for the color to play as, castling and en passant availability, and best moves in standard algebraic notation.
The fields are as follows:
image: image, A visual representation of the chess board showing the current piece arrangement.
board_state: string, A shortened FEN… See the full description on the dataset page: https://huggingface.co/datasets/bingbangboom/chess-puzzles-images-large.so101-puzzle-v9This dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v3.0",
"robot_type": "so_follower",
"total_episodes": 53,
"total_frames": 47059,
"total_tasks": 1,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 200,
"fps": 30,
"splits": {
"train": "0:53"
},
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
"video_path":… See the full description on the dataset page: https://huggingface.co/datasets/aivanni/so101-puzzle-v9.PuzzleVQAPuzzleVQA: Diagnosing Multimodal Reasoning Challenges of Language Models with Abstract Visual Patterns
dcagent2-terminal-bench-2-dcagent-all-puzzles-sandboxes-traces-terminus-2-with-98076364puzzlemate-pick-flip
puzzlemate-pick-flip
遵循 LeRobot v3.0 规范。
可视化链接: https://huggingface.co/spaces/lerobot/visualize_dataset?path=ni25y/puzzlemate-picknflip
rebus-puzzles
Rebus Dataset
The Rebus Dataset is a collection of 221 rebus puzzle images, each annotated with corresponding textual solutions and metadata.It was introduced as part of the paper:
Reasoning Riddles: How Explainability Reveals Cognitive Limits in Vision-Language ModelsPrahitha Movva, 2025arXiv:2510.02780
The dataset is designed to support research in visual reasoning, multimodal interpretability, and cognitive evaluation of vision–language models.
Dataset Structure… See the full description on the dataset page: https://huggingface.co/datasets/pmovva/rebus-puzzles.Word-Puzzles-ARC-Unique-50000
Word-Puzzles-ARC-Unique-46000
This dataset is a synthetic 46,000-row word-puzzle corpus focused on answerable reasoning tasks with explicit gold answers.
Version
This upload corresponds to the harder v2 build.
Bucket mix
15,000 formal deduction
12,500 constraint-based lexical deduction
10,000 symbolic substitution
7,500 semantic association
1,000 riddles
Hardening changes in v2
formal puzzles use 6 entities instead of 5
lexical… See the full description on the dataset page: https://huggingface.co/datasets/Crownelius/Word-Puzzles-ARC-Unique-50000.Puzzle_Perception
This dataset has moved
It is now maintained at:
PuzzleComm/Puzzle_Perception
Browse all our datasets at PuzzleComm
Jigsaw-Puzzles
Jigsaw-Puzzles Dataset
Jigsaw-Puzzles is a novel benchmark consisting of 1,100 carefully curated real-world images with high spatial complexity, designed to rigorously evaluate Vision-Language Models' (VLMs) spatial perception, structural understanding, and reasoning capabilities. The dataset minimizes reliance on domain-specific knowledge to better isolate and assess general spatial reasoning, positioning itself as a challenging and diagnostic benchmark for advancing spatial… See the full description on the dataset page: https://huggingface.co/datasets/zesen01/Jigsaw-Puzzles.