TIC-VLA: Vision-Language-Action Dataset for Robot Navigation

TIC-VLA is a large-scale Vision-Language-Action dataset for robot navigation in dynamic environments. It was created by handsomeYun and last updated on July 3, 2026. The dataset is designed for training and evaluating reasoning-enabled robot navigation models.

Use Cases

• Train robot navigation models based on joint understanding of visual observations and language instructions.
• Evaluate reasoning capabilities of vision-language-action models based on temporal context.
• Benchmark robot navigation performance in dynamic environments based on the dataset's construction from multiple sources.

Strengths

• Designed as a large-scale dataset for a complex multimodal task.
• Constructed from three existing robot navigation datasets: SCAND, GND, and one other.

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, file formats, and license are unknown, which may limit suitability assessment.

Provenance

Source

huggingface

Collection Method

Constructed from three robot navigation datasets: SCAND, GND, and one other.

Freshness

Last updated 2026-07-03 06:06:27; freshness should be verified.