Skill-3D: Data Splits for Agentic 3D Spatial Reasoning

Skill-3D splits are designed for skill evolution, post-training, and evaluation in agentic 3D spatial reasoning tasks. The dataset, created by author lhy-zju, was last updated on June 14, 2026. Its specific splits likely support training and testing of scene-aware AI agents.

Use Cases

• Evolving scene-aware skills for agents based on the described skill evolution framework.
• Post-training models via supervised fine-tuning (SFT) or GRPO using the provided splits.
• Evaluating agent performance on 3D spatial reasoning benchmarks.
• Training multimodal AI systems that integrate 3D scene understanding.

Strengths

• Dataset is associated with a named research framework (Skill-3D) and a paper.
• Splits are explicitly designed for multiple stages: skill evolution, post-training, and evaluation.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Data freshness should be verified as the last update timestamp is in the future (2026).

Provenance

Source

lhy-zju

Freshness

Last updated 2026-06-14 07:20:13