ModelNet40

anchor

Findings

• IC-1496A single frozen transformer block from LLaMA-7B consistently improves performance across diverse visual tasks when appended to existing visual encoders [eval]