Search papers, labs, and topics across Lattice.
Sino-German Joint Software Institute, Beihang University
3
0
6
RATrain achieves a remarkable 1.35x end-to-end speedup for LLM training on bandwidth-limited supercomputers, challenging the notion that high-bandwidth environments are necessary for efficiency.
LLM inference on supercomputers doesn't have to be a bottleneck: THInfer achieves up to 84% higher throughput than A800 GPUs by co-designing hardware-aware kernels and a communication-optimized pipeline.
LLMs still can't reliably reverse engineer stripped binaries, and REBench offers a standardized, fair-by-construction benchmark to finally measure progress.
