TAM generates clearer token-level visual explanations for multimodal LLMs by subtracting context-token interference and denoising activation maps.
Text2live: Text-driven layered image and video editing
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Token Activation Map to Visually Explain Multimodal LLMs
TAM generates clearer token-level visual explanations for multimodal LLMs by subtracting context-token interference and denoising activation maps.