1 paper · 1 filter
Yike Wu, Keyan Chen, Qixiong Wang +6
Test-time agentic computation can improve multimodal large language models (MLLMs) through planning, tool use, iterative perception, and reflection, but indiscriminate computation…