ReVSI: A Benchmark for Visual Spatial Intelligence Evaluation of Vision-Language Models

The ReVSI benchmark and dataset was introduced in the paper 'ReVSI: Rebuilding Visual Spatial Intelligence Evaluation for Accurate Assessment of VLM 3D Reasoning'. It was created by researchers from Simon Fraser University, Hong Kong University of Science and Technology, University of Waterloo, and the Alberta Machine Intelligence Institute (Amii). The dataset page was last updated on 2026-04-27.

Use Cases

• Benchmarking the 3D spatial reasoning performance of Vision-Language Models based on the described evaluation tasks.
• Training or fine-tuning models to improve visual spatial intelligence based on the benchmark's structure.
• Conducting ablation studies to understand which model components contribute to spatial reasoning based on the evaluation framework.

Strengths

• Created by a multi-institutional research team from four academic organizations.
• Specifically designed for the accurate assessment of 3D reasoning in Vision-Language Models.

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 sample data are unknown, which may limit suitability assessment.

Provenance

Source

3dlg-hcvc on Hugging Face, associated with the academic paper 'ReVSI: Rebuilding Visual Spatial Intelligence Evaluation for Accurate Assessment of VLM 3D Reasoning'.

Collection Method

Likely constructed for research and benchmarking purposes, as described in the associated paper.

Time Range

null

Freshness

Last updated 2026-04-27 07:39:07; freshness should be verified.

Geography

null