Search papers, labs, and topics across Lattice.
3
0
5
1
GPU kernel optimization agents can regress by up to 19% when transferring distilled knowledge across DSLs, revealing that source attempt quality, not just success, dictates transfer efficacy.
Current OmniLLMs stumble when processing real-world, long-form audio-visual content, achieving only ~35-65% accuracy on a new benchmark designed to test long-term memory and fine-grained understanding.
LLMs can now generate mobile kernels that outperform native libraries 27.4% of the time, thanks to a new multi-agent system that tackles compilation failures and hallucinations.
