Weakly Supervised From survey PDF — verify links
CALLM
CALLM — LLM Knowledge Enhancement
IPTA · 2024Human-reviewed summary status
Survey summary
CALLM cascades a 3D deep autoencoder with a large video-language model: the autoencoder's binary classification flags abnormality and triggers a follow-up query to the LVLM for detailed explanation, using weakly supervised caption extraction and pseudo-instruction injection to improve performance.
Taxonomy placement
Weakly Supervised
Weakly SupervisedLLM Knowledge Enhancement
Method characteristics
Weakly Supervised3D autoencoderVideo-LLaMAPseudo-instruction