arXiv:2509.25773v3 Announce Type: replace-cross Abstract: AI models capable of comprehending humor hold real-world promise -- for example, enhancing engagement in human-machine interactions. To gauge and diagnose the capacity of multimodal large language models (MLLMs) for humor understanding, we introduce v-HUB, a novel video humor understanding benchmark. v-HUB comprises a curated collection of non-verbal short videos, reflecting real-world scenarios where humor can be appreciated purely through visual cues. We pair each video clip with rich annotations to support a variety of evaluation tas
The proliferation of multimodal large language models necessitates robust benchmarks to accurately assess and improve their complex cognitive capabilities, such as humor understanding, which is crucial for advanced human-AI interaction.
Understanding non-verbal humor is a significant step towards more sophisticated and human-like AI, unlocking new applications in entertainment, education, and personalized digital experiences.
This benchmark provides a standardized method to evaluate and drive progress in AI's ability to comprehend subtle, non-verbal social cues, refining model development previously limited by subjective or narrow evaluations.
- · AI researchers
- · MLLM developers
- · Entertainment industry
- · AI ethics and safety organizations
- · AI models with weak multimodal integration
- · Companies relying on simplistic AI interaction
- · Developers neglecting emotional AI capabilities
AI models will become more adept at identifying and generating contextually appropriate humor in video and audio.
This improved humor understanding will enhance the naturalness and engagement of AI companions, virtual assistants, and creative AI applications.
The ability of AI to grasp humor might lead to new ethical considerations regarding manipulative AI or the generation of offensive content, requiring advanced regulatory frameworks.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.CL