Introduces VIG metric to measure visual contribution via perplexity reduction and applies it for selective training of LVLMs on high-VIG samples and tokens to improve grounding with reduced supervision.
An image is worth 1/2 tokens after layer 2: Plug-and-play inference acceleration for large vision-language models
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
-
Focusing Where Vision Matters: Selective Training for Large Vision Language Models via Visual Information Gain
Introduces VIG metric to measure visual contribution via perplexity reduction and applies it for selective training of LVLMs on high-VIG samples and tokens to improve grounding with reduced supervision.