POPE Benchmark

What is POPE?

POPE is an improved evaluation method for LVLMs' object hallucination by proposing a polling-based query method. It offers a more stable and flexible solution.

Released

2023-05-17

Modality

multimodal

Evaluates

多模态, Multimodal, 理解, Understanding, 多模态模型, VLM, 图像理解, Image Understanding, 事实可靠性, Factual Reliability, 不支持, Unsupported

Openness

restricted

Importer

OpenCompass Hub

Reported scores

0

Source provenance

• Original evidence https://hub.opencompass.org.cn/dataset-detail/POPE

POPE paper, code and dataset

• Paper https://arxiv.org/abs/2305.10355
• Code https://github.com/AoiDragon/POPE
• Dataset https://huggingface.co/datasets/lmms-lab/POPE

Which sources cite POPE?

• POPE

Related 多模态, Multimodal, 理解, Understanding, 多模态模型, VLM, 图像理解, Image Understanding, 事实可靠性, Factual Reliability, 不支持, Unsupported benchmarks

• 3DCodeBench
• A-Bench
• A-OKVQA
• AbsPyramid
• ActionAtlas
• ActiveFly-Bench
• Ada-LEval
• AECBench