SAGE-3D VLN: Vision-Language Navigation Data with Hierarchical Instructions

SAGE-3D VLN Data is a Vision-Language Navigation dataset created by spatialverse, last updated in December 2025. It features 2 million trajectory-instruction pairs across 1,000 indoor scenes. The dataset is designed with hierarchical instructions covering high-level semantic goals to low-level control commands.

Use Cases

• Train vision-language navigation agents based on trajectory-instruction pairs.
• Benchmark agent performance on hierarchical instruction following from goals to commands.
• Research multimodal representation learning for navigation in 1,000 indoor scenes.
• Develop models for the 'VLN + No-goal' task types mentioned in the description.

Strengths

• Contains 2 million trajectory-instruction pairs, indicating a large scale.
• Covers 1,000 indoor scenes, suggesting broad environmental diversity.
• Features hierarchical instruction design, a structured approach to task complexity.

Limitations

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

Provenance

Source

spatialverse

Collection Method

Method of data gathering is not specified in the provided input.

Time Range

Temporal coverage is not specified in the provided input.

Freshness

Last updated 2025-12-17 02:12:32; freshness should be verified.

Geography

Spatial coverage is not specified in the provided input.