CapRL: Prism Evaluation Files for Vision-Language Models

CapRL Evaluation Files contain 12 Prism evaluation JSON files and a corresponding image folder used by the CapRL Prism evaluation scripts. The dataset, authored by internlm, includes image paths relative to the dataset root, such as 'bench_image_folder/lmm_eval_chartqa/41699051005347.png'. It was last updated on April 17, 2026.

Use Cases

• Benchmarking vision-language models based on the Prism evaluation scripts
• Evaluating multimodal AI performance using structured JSON files
• Testing model capabilities on chart-based visual question answering tasks based on the example image path

Strengths

• Includes 12 distinct Prism evaluation JSON files
• Contains a dedicated folder of benchmark images referenced by the JSON files

Limitations

• Column-level documentation is absent; field semantics must be inferred after download
• Row count is unknown, which may limit suitability assessment

Provenance

Source

internlm

Freshness

Last updated 2026-04-17 11:26:26