Search papers, labs, and topics across Lattice.
7
0
12
8
HarnessBridge can outperform specialized harnesses while cutting token usage and trajectory length, revolutionizing LLM agent interactions.
Multi-turn RL agents can learn far more effectively by explicitly monitoring and controlling uncertainty at both the token and turn levels, leading to more stable training and higher performance.
Achieve 75% input length reduction in LLMs with minimal performance loss by compressing token embeddings directly in the latent space.
LLM inference spends up to 97% of its time just *preparing* memory, but offloading that work to an FPGA can more than double inference speed.
Forget scaling compute – the future of AI hinges on a 1000x leap in energy efficiency via tight AI+Hardware co-design over the next decade.
ARLArena reveals the hidden instability of agentic RL, offering a path to more reliable LLM-based agents via a novel stable policy optimization method (SAMPO).
Democratizing hardware design and enabling next-generation hardware systems requires strategic NSF investment in AI/EDA collaboration, foundational AI, data infrastructure, and workforce development.
