T-GRAB: A Synthetic Benchmark for Temporal Graph Reasoning

T-GRAB is a synthetic reasoning benchmark designed for learning on temporal graphs. It was created by Gilestel and the dataset entry was last updated on August 19, 2025. The benchmark is associated with a research paper accessible via the provided URL.

Use Cases

• Benchmarking model performance on temporal graph reasoning tasks based on the synthetic data structure.
• Developing new learning algorithms for temporal graphs based on the described benchmark tasks.
• Studying the generalization capabilities of graph neural networks on synthetic temporal reasoning problems.

Strengths

• Designed specifically as a benchmark for temporal graph reasoning, providing a clear evaluation target.
• Associated with a research paper (arXiv:2507.10183), suggesting a documented methodology.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and dataset size are unknown, which may limit suitability assessment.

Provenance

Source

Gilestel via Hugging Face.

Collection Method

Synthetically generated, as described in the associated research paper.

Freshness

Last updated 2025-08-19 16:48:25; freshness should be verified.