FG-CLIP Wins Silver in the 2026 CSIG Innovation Technology List
Recognition for fine-grained multimodal understanding
FG-CLIP has received a silver award in the 2026 China Society of Image and Graphics Innovation Technology List. The recognition reflects both the project’s research contribution and its progress toward practical deployment.
Conventional image–text models often align only broad concepts. FG-CLIP is designed to distinguish attributes, local regions, long descriptions, and subtle differences between otherwise similar candidates. FG-CLIP 2 extends this capability with stronger bilingual data and evaluation across eight categories and 29 tasks.
From benchmarks to production
The project retains an efficient dual-encoder architecture, making fine-grained embeddings practical for retrieval, recommendation, document intelligence, and semantic video monitoring. Engineering work has also focused on throughput and service integration so that the model can be used beyond an experimental benchmark.
FG-CLIP capabilities are now available through 360 AI Research’s online products and APIs, giving teams a direct way to evaluate detailed cross-modal retrieval in their own workflows.