Search papers, labs, and topics across Lattice.
4
0
4
2
Achieving nearly 4x energy savings and over 4x latency improvements for LLMs on edge devices without sacrificing performance opens new avenues for deploying advanced AI in real-world applications.
Expanding on-chip memory to 1GB can slash energy consumption by 44% during LLM prefill phases, revealing a game-changing approach to LLM efficiency.
p-MEM achieves over 1000 GSa/s/mm虏 throughput, drastically cutting down latency and energy use for probabilistic AI applications.
By enabling draft models to "contemplate the future," ConFu achieves significant speedups in speculative decoding, outperforming EAGLE-3 by 8-11% on Llama-3 models.
