LLaMA-CLIP fusion improves multimodal digital-human interaction benchmarks
A LLaMA-CLIP cross-attention system improved multimodal matching, intention recognition and personalization in benchmark tests while recording batch-size-one inference latency of ...
