LG Uplus to present smartphone AI token optimization at NeurIPS 2026

LG Uplus said two of its papers on AI token optimization were accepted to the On-Device Intelligence workshop at NeurIPS 2026. The company built a small language model for call services that processes tokens about 4.5x faster and cuts battery drain by 77%, while on-device training matched server-trained accuracy.

• Two LG Uplus papers accepted to the NeurIPS 2026 workshop, held Dec 6-12 in Sydney
• Token processing about 4.5x faster while keeping cloud-level quality
• Battery consumption cut by 77% through on-device optimization
• On-device training without sending data to servers matched server accuracy