SynLogic

paper dataset
MiniMax
Research on generating massive-scale logical reasoning data for training, with publicly available verifiable reasoning trajectories.
Paper (arXiv)GitHubHuggingFace (Model)HuggingFace (Dataset)

Outputs 2

SynLogic: Synthesizing Verifiable Reasoning Data

paper

Research on generating massive-scale logical reasoning data for training.

Paper (arXiv)GitHubHuggingFace

arXiv: 2505.19641

SynLogic Dataset

dataset

Publicly available set of verifiable reasoning trajectories.

HuggingFaceGitHub
reasoningtraining-dataresearch

Notes

Accepted at NeurIPS 2025.