Bilibili Tech

Oct 17, 2025 · Artificial Intelligence

How Bilibili’s Multimodal Team Won 2nd Place at ICCV MIPI with a Novel SFT+GRPO Strategy

This article details how Bilibili’s multimedia lab leveraged a multimodal training pipeline combining data‑compressed SFT and the GRPO reinforcement‑learning algorithm to achieve a 13.5% metric boost and secure second place in the ICCV MIPI Detailed Image Quality Assessment competition.