Beyond Attention or Similarity: Maximizing Conditional Diversity for Token Pruning in MLLMs

Devs

Beyond Attention or Similarity: Maximizing Conditional Diversity for Token Pruning in MLLMs | Read Paper on Bytez