The article analyzes SenseTime’s recent SenseNova U1 Pro and SenseNova‑Vision releases, detailing their NEO‑unify architecture, unified multimodal training, extensive SN‑VC‑50M dataset, benchmark breakthroughs, and the broader shift toward a single, long‑term, agentic AI base model.
The paper introduces SICOG, a three‑stage collaborative framework that combines post‑training enhancement, inference optimization, and re‑pretraining with a self‑generated data loop, allowing large multimodal models to continuously improve without massive human‑annotated datasets, and demonstrates consistent gains across dozens of benchmarks.
This issue introduces a Tsinghua research breakthrough for fine‑grained video understanding with multimodal large models, explains how to engineer open‑source models to reliably emit JSON, reviews recent AI‑driven earnings of cloud giants and challenges faced by domestic AI firms, and highlights a new text‑to‑image tool, Recraft.ai.
