benchmark · first seen 19 Oct 2026
POPE benchmark
Also: Polling-based Object Probing Evaluation
A benchmark measuring object hallucination in vision-language models via yes/no questions about object presence.
Mentions per week
PapersCodeDiscussion
Earliest sources
- Token-Set Choice Confounds POPE: A Systematic Audit of Yes/No Extraction in VLM Hallucination EvaluationPapers · 19 Oct 2026 · Kesav Kumar Jayakumar, Karthigeyan Thilak
- RAMP: Rethinking Multimodal Evaluation through Answer Readout ProtocolsPapers · 16 Nov 2026 · Jianye Zeng, Qiao Yikang, Duan Huang, Ziying Song
Latest sources
- RAMP: Rethinking Multimodal Evaluation through Answer Readout ProtocolsPapers · 16 Nov 2026 · Jianye Zeng, Qiao Yikang, Duan Huang, Ziying Song
- Token-Set Choice Confounds POPE: A Systematic Audit of Yes/No Extraction in VLM Hallucination EvaluationPapers · 19 Oct 2026 · Kesav Kumar Jayakumar, Karthigeyan Thilak