absorb.md

Yulun Wang

Chronological feed of everything captured from Yulun Wang.

Error-Suppressed Quantum Pipeline Solves Nontrivial Binary Optimization at 156-Qubit Scale

A hybrid quantum-classical pipeline integrates custom ansatz, dual variational updates, parametric compilation, hardware error suppression, and O(n) post-processing to solve unconstrained binary combinatorial optimization on gate-model quantum hardware. Without these components, outputs match random sampling, underscoring their necessity. On IBM devices, it achieves 100% approximation ratios for Max-Cut on 3-regular graphs up to 156 qubits and ≥99.5% for 156-qubit spin-glass ground states, outperforming classical local solvers.

NCCLX: Scaling Collective Communication for Large Language Models

The NCCLX framework addresses the communication bottlenecks for LLM training and inference on GPU clusters exceeding 100,000 GPUs. It optimizes for both high-throughput synchronous training and low-latency inference demands. This solution facilitates operation of next-generation LLMs at unprecedented scales.