Vision-language models fail at zero-shot detection of climate-specific classes in social media videos, while DINOv2 and ConvNeXt V2 embeddings yield meaningful clusters via minimum-cost multicut.
Climate hoax: The shift from scientific discourse to speculative rhetoric in climate change conversations.Next Research, page 100322
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
ClimateVID -- Social Media Videos Analysis and Challenges Involved
Vision-language models fail at zero-shot detection of climate-specific classes in social media videos, while DINOv2 and ConvNeXt V2 embeddings yield meaningful clusters via minimum-cost multicut.