Multimodal Multi-Turn Dialogue Safety Evaluations

The dataset, last updated in March 2026, is designed for safeguarding Vision-Language Models (VLMs). It focuses on adversarial robustness and safety alignment for interactive, multi-turn conversations. The dataset was created by author leost233.

Use Cases

• Benchmark VLM safety alignment by analyzing responses to adversarial multi-turn dialogue prompts.
• Evaluate adversarial robustness in multimodal dialogue systems using safety-focused conversation threads.
• Analyze failure modes in Vision-Language Models during interactive, multi-turn conversations.

Strengths

• Dataset is specifically designed for the critical research area of multimodal dialogue safety.
• Targets a contemporary challenge with VLMs capable of multi-turn conversations.
• Associated with tags indicating focus on Adversarial Robustness and Safety Alignment.

Limitations

• Key metadata like row count, column names, and file formats is unavailable, hindering usability assessment.
• The dataset description is truncated, lacking details on data collection methodology and specific contents.
• Sample data is unavailable, making it impossible to preview data structure or quality.

Provenance

Source

huggingface

Collection Method

null

Time Range

null

Freshness

Last updated 2026-03-11.

Geography

null