IRA is a stochastic attention mechanism that regulates visual information injection in VLMs to yield smoother embedding trajectories and reduced attention sinks.
Attention guided alignment in efficient vision-language models.arXiv preprint arXiv:2511.17793,
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
-
Information-Regularized Attention for Visual-Centric Reasoning
IRA is a stochastic attention mechanism that regulates visual information injection in VLMs to yield smoother embedding trajectories and reduced attention sinks.