datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
ChessInstruct
ChessInstruct
The ChessInstruct Dataset serves as the foundation for training and fine-tuning Language Models (LLMs) specifically in the realm of chess instruction.
Derived from the laion/strategic_game_chess dataset, this meticulously curated dataset encompasses a wide array of annotated instructional chess content.
Features of the ChessInstruct Dataset:
Rich and Diverse Content: Curated with a broad spectrum of instructional resources including annotated games, strategic analyses… See the full description on the dataset page: https://huggingface.co/datasets/Thytu/ChessInstruct.chess_datasets
Dataset Card for Dataset Name
This dataset card aims to be a base template for new datasets. It has been generated using this raw template.
Dataset Details
Dataset Description
Curated by: [More Information Needed]
Funded by [optional]: [More Information Needed]
Shared by [optional]: [More Information Needed]
Language(s) (NLP): [More Information Needed]
License: [More Information Needed]
Dataset Sources [optional]
Repository: [More… See the full description on the dataset page: https://huggingface.co/datasets/p11-p11/chess_datasets.searcless-chess-10mchess-roberta-basechess_checkmate_in_one_big_bench
Dataset Card for BIG-Bench Checkmate In One Move
This is an adapted version of the BIG-Bench Checkmate in One Move task originally made by Nitish Keskar (nkeskar@salesforce.com).
There also exists a dataset with the image representation of the board positions before the checkmate move.
The original task description is the following:
The goal of this task is to probe the ability of language models to play chess in standard algebraic notation (SAN). The input to the model is a… See the full description on the dataset page: https://huggingface.co/datasets/alonsosilva/chess_checkmate_in_one_big_bench.Chess_openings_dataset
Version 1 of the dataset
Structure of the dataset:
Opening_type:
The title of the opening being played.
Context:
A string representing a list of moves, each move is represented by the previous state of the board, the move that is going to be made, and the effect that the move had on the board.
The board is represented as an 8*8 grid of characters where each character represents a piece or an empty square:
r . . q k b n r
p p p . p . p p
. . n .… See the full description on the dataset page: https://huggingface.co/datasets/nelson2424/Chess_openings_dataset.chess-puzzles-san
Dataset Card for Chess Puzzles (SAN)
4,130,824 chess puzzles from Lichess with moves converted to standard algebric notation.
Puzzles are formatted as standard CSV. The fields are as follows:
PuzzleId,new_FEN,best_Moves,Rating,RatingDeviation
PuzzleId: string, the puzzle's unique identifier. The puzzle would be live at https://lichess.org/training/{PuzzleID}.
new_FEN: string, the FEN string of the position after the opponent has made their move. It is now player's turn to make… See the full description on the dataset page: https://huggingface.co/datasets/bingbangboom/chess-puzzles-san.Chess-FEN-and-NL-Format-30K-Dataset
Dataset Card for Dataset Name
CHESS data with 2 Representation- FEN format and Natural language format.
Dataset Details
Dataset Description
This dataset contains 30000+ Chess data. Chess data can be represented using FEN notation and also in language description format. Each datarow of this dataset contains FEN notation, next best move in UCI format and Natural Language Description of that specific FEN notation.
This dataset is created using Stockfish.… See the full description on the dataset page: https://huggingface.co/datasets/bonna46/Chess-FEN-and-NL-Format-30K-Dataset.titled-tuesday-chess-gamesThis dataset includes chess games played in the Chess.com "Titled Tuesday" events from July 2022 to December 2023. The games in the dataset were played by titled players
who ranked 1-25 in at least one tournament. Since game documents are organized around players and there are duplicate games, the dataset is more suitable for the player-centric analyses.
This GitHub repository contains scripts for scraping the data from Chess.com
website and API as well some demos of statistical analysis of… See the full description on the dataset page: https://huggingface.co/datasets/kirillgoltsman/titled-tuesday-chess-games.chess_spatial_reasoning_10kchess_puzzle_training_datasets_lt-2400
Chess puzzle training datasets: rating below 2400
This is a filtered derivative of
pavelslab-nyu/chess_puzzle_training_datasets.
Every retained row satisfies the exact condition:
Rating < 2400
Rating is the Lichess puzzle rating, not the Elo of either player in the
source game. The original column names, column order, directory layout, and CSV
schemas are preserved. As in the upstream repository, Hugging Face discovers
all three CSVs as one default configuration with one train… See the full description on the dataset page: https://huggingface.co/datasets/violetxi/chess_puzzle_training_datasets_lt-2400.chess_position_evaluationschess_glossary
[!NOTE]
Dataset origin: http://www.shakki.net/kerhot/KemTS/nap-pieces.htm
chesschessgpt-board-probes-dataset
ChessGPT Board Probes Stockfish Dataset
This dataset contains fixed-length PGN strings used to train and test linear probes
on TransformerLens-converted chess models.
Train: stockfish_train.csv (10,000 games)
Test: stockfish_test.csv (10,000 games)
Each PGN string is exactly 365 characters long.
Source: internal subset derived from Stockfish-vs-Stockfish self-play games.
For usage examples, see the jd0g/chessgpt-board-probes model repo.
Uploaded at: 2025-08-29T03:58:31.828571Z
chessChessInstruct
ChessInstruct
The ChessInstruct Dataset serves as the foundation for training and fine-tuning Language Models (LLMs) specifically in the realm of chess instruction.
Derived from the laion/strategic_game_chess dataset, this meticulously curated dataset encompasses a wide array of annotated instructional chess content.
Features of the ChessInstruct Dataset:
Rich and Diverse Content: Curated with a broad spectrum of instructional resources including annotated games, strategic analyses… See the full description on the dataset page: https://huggingface.co/datasets/EnzoDeg40/ChessInstruct.trm-chess-datachess-roberta-pretraining-sansconfigs:
config_name: default
data_files:
split: train
path: train/*.csv
split: eval
path: eval/*.csv
ChessCOT
ChessCOT
The dataset that makes your chess model think like a human before it plays a move.
About
ChessCOT is a dataset designed to train transformers for chess using a Chain of Thought (CoT) approach. The goal is to make the model reason about the position with all possible moves and their consequences in order to predict the best move.
Dataset Statistics
Total Poistions: 4,491,596
Sequence length of sMoves: 128
Sequence length of thought: 128… See the full description on the dataset page: https://huggingface.co/datasets/frosthead/ChessCOT.ChessReasoningDatasetChessDatawithNLAndNonNLFormat
Dataset Card for Dataset Name
This dataset card aims to be a base template for new datasets. It has been generated using this raw template.
Dataset Details
Dataset Description
Curated by: [More Information Needed]
Funded by [optional]: [More Information Needed]
Shared by [optional]: [More Information Needed]
Language(s) (NLP): [More Information Needed]
License: [More Information Needed]
Dataset Sources [optional]
Repository: [More… See the full description on the dataset page: https://huggingface.co/datasets/bonna46/ChessDatawithNLAndNonNLFormat.Chess_datasetThis contains Broken down chess dataset for training purposes
chess
Dataset
Modified version of lichess_elite_2020-06
Created by: https://lichess.org/@/nikonoel
Source: https://database.nikonoel.fr/
Source: https://database.nikonoel.fr/lichess_elite_2020-06.zip
chessDPOThis is an dpo which i created to finetune an chess model.
chess-annotation-datasetChess_openings_dataset
Version 1 of the dataset
Structure of the dataset:
Opening_type:
The title of the opening being played.
Context:
A string representing a list of moves, each move is represented by the previous state of the board, the move that is going to be made, and the effect that the move had on the board.
The board is represented as an 8*8 grid of characters where each character represents a piece or an empty square:
r . . q k b n r
p p p . p . p p
. . n .… See the full description on the dataset page: https://huggingface.co/datasets/coryvegan/Chess_openings_dataset.FIDE_Rated_Chess_Players_By_CountrychessDatachess-400k
Dataset
Modified version of lichess_elite_2020-06
Created by: https://lichess.org/@/nikonoel
Source: https://database.nikonoel.fr/
Source: https://database.nikonoel.fr/lichess_elite_2020-06.zip