T2I Eval Distill GPT-4O: Training Set for Text-to-Image Evaluation

A training set distilled from GPT-4o for the paper 'Automatic Evaluation for Text-to-Image Generation: Fine-grained Framework, Distilled Evaluation Model and Meta-Evaluation Benchmark'. The dataset was created by DataHammer and last updated on December 16, 2024.

Use Cases

• Training evaluation models based on distilled GPT-4o judgments.
• Benchmarking text-to-image generation systems using a fine-grained evaluation framework.
• Conducting meta-evaluation studies on automatic evaluation metrics.
• Developing distilled models for efficient text-to-image quality assessment.

Strengths

• Created by DataHammer, a known contributor in the AI evaluation space.
• Distilled from GPT-4o, a state-of-the-art model, suggesting high-quality training signals.
• Associated with a published academic paper, providing context and validation.

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 is unknown, which may limit suitability assessment.

Provenance

Source

DataHammer via Hugging Face.

Collection Method

Distilled from GPT-4o outputs.

Freshness

Last updated 2024-12-16 06:44:07; freshness should be verified.