Treat Visual Tokens as Text? But Your MLLM Only Needs Fewer Efforts to See

Re-assign community