The survey organizes hallucinations in Vid-LLMs into dynamic distortion and content fabrication, reviews evaluation benchmarks and mitigation methods, and traces root causes to weak temporal modeling and visual grounding.
Luowei Zhou, Chenliang Xu, and Jason J
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
-
Distorted or Fabricated? A Survey on Hallucination in Video LLMs
The survey organizes hallucinations in Vid-LLMs into dynamic distortion and content fabrication, reviews evaluation benchmarks and mitigation methods, and traces root causes to weak temporal modeling and visual grounding.