datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
boardgamebench-answer-only
BoardGameBench Answer-Only Reasoning Dataset
This dataset contains 1,282,766 board-game reasoning examples generated from BoardGameBench, a benchmark and data-generation project for evaluating language models on structured board-game decision making.
Each row asks a model to inspect a legal board position and return the best move. The format is intentionally simple:
id,prompt,answer
The answer field is the target move label, such as C4, f6, 11,7, or e2-d3. This makes the dataset… See the full description on the dataset page: https://huggingface.co/datasets/homerquan/boardgamebench-answer-only.boardgamebench-answer-sft
BoardGameBench Answer SFT Dataset
This dataset contains 1,282,766 answer-only board-game reasoning examples generated from BoardGameBench, a benchmark and data-generation project for evaluating language models on structured board-game decision making.
This is the supervised fine-tuning corpus used before the later DPO and GRPO stages for the nemotron-boardgame-answer-lora-b4-safe-final adapter.
Each row asks a model to inspect a legal board position and return the best move. The… See the full description on the dataset page: https://huggingface.co/datasets/homerquan/boardgamebench-answer-sft.BoardgameQA-trajectory
BoardGameQA Trajectory Dataset
Model-generated solution trajectories for BoardGameQA, a natural-language deductive reasoning benchmark built from board-game-style rule sets with conflicting/preferential rules. Each question asks whether a goal statement is yes (proved), no (disproved), or unknown. Each row is one model response to a single problem, including the hidden chain-of-thought (when available) and the final response.
Dataset Summary
Split
Rows… See the full description on the dataset page: https://huggingface.co/datasets/sxiong/BoardgameQA-trajectory.boardgamebench-answer-grpo
BoardGameBench Answer GRPO Dataset
This dataset contains 10,000 BoardGameBench prompt/reward examples generated for GRPO-style reinforcement learning on board-game move selection.
The final nemotron-boardgame-answer-lora-b4-safe-final adapter used this reviewed GRPO corpus after SFT and DPO. For that final pilot run, training used the first 512 examples from grpo_train.jsonl; the full 10k reviewed set is published here for reproducibility and follow-up training.
Format… See the full description on the dataset page: https://huggingface.co/datasets/homerquan/boardgamebench-answer-grpo.boardgamebench-answer-dpo
BoardGameBench Answer DPO Dataset
This dataset contains the reviewed preference examples used for the DPO stage of the nemotron-boardgame-answer-lora-b4-safe-final adapter.
It is a compact pilot set of 10 BoardGameBench preference rows. Each row presents the same board-game decision prompt with a preferred answer and a plausible rejected answer. The preferred answer is selected from engine-guided move comparisons and includes the exact move label.
Format
The main… See the full description on the dataset page: https://huggingface.co/datasets/homerquan/boardgamebench-answer-dpo.