MemBench-InternVL3.5-Eval: Image Editing Method Comparisons

Evaluation dataset for image editing experiments on the ppr10k dataset, comparing four methods under the same selection protocol. Created by author laitifranz and last updated on March 9, 2026, it contains dataset.jsonl files and corresponding edited/source image pairs for reproduction and inspection.

Use Cases

• Benchmarking image editing methods based on the described comparison protocol
• Reproducing experimental results from the associated MemCoach codebase
• Inspecting edited/source image pairs for qualitative analysis of model outputs

Strengths

• Provides a structured comparison of four different image editing methods
• Includes both structured metadata (dataset.jsonl) and raw image pairs for evaluation

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 license information are unknown

Provenance

Source

laitifranz on Hugging Face

Collection Method

Created for evaluation and reproduction of image editing experiments, likely derived from the ppr10k dataset.

Freshness

Last updated 2026-03-09 10:37:57; freshness should be verified