OPPO’s On-Device Multimodal LLM Engineering Practice at AICon Shenzhen
Dr. Xiyu Yu, multimodal algorithm expert at OPPO, will share OPPO’s full-stack engineering practice for on-device multimodal large language models at AICon Shenzhen 2026. The talk covers quantization-aware training (QAT), sparse compression, long-context cache eviction, decoding acceleration, and automated QALFT pipelines to address inference latency, engineering complexity, and personalization on resource-constrained devices.
OPPO
MediaTek
Qualcomm
InfoQ 中国29.07 · 06:02
