HRIBench is a new 1,000-question VQA benchmark for five HRI perception domains; state-of-the-art vision-language models are neither accurate enough nor fast enough for real-time human-robot interaction.
Discourse Processes52(4), 255–289 (2015)
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.RO 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
HRIBench: Benchmarking Vision-Language Models for Real-Time Human Perception in Human-Robot Interaction
HRIBench is a new 1,000-question VQA benchmark for five HRI perception domains; state-of-the-art vision-language models are neither accurate enough nor fast enough for real-time human-robot interaction.