Official implementation of paper "MiniGPT-5: Interleaved Vision-and-Language Generation via Generative Vokens"
Terminal score: 0–1000 raw, weighted across 4 dimensions. Public score: 0–10 normalized (shown in the 30-day stars chart above).