Search papers, labs, and topics across Lattice.
Instituto de Investigaci贸n en Visi贸n Artificial
3
0
5
Compact prompt routing and lightweight temporal adaptation enable superior long video extrapolation without the overhead of full model fine-tuning.
FreqDepthKV achieves a remarkable 3.9x effective compression ratio while preserving high accuracy in long-context LLM tasks, revolutionizing cache management.
Achieving an 8.3x reduction in memory usage without sacrificing retrieval quality could redefine efficiency standards for long-context language models.
