sensenova/SenseNova-SI-1.1-Qwen2.5-VL-7B
SenseNova-SI-1.1-Qwen2.5-VL-7B is a 7 billion parameter multimodal foundation model developed by SenseNova, built upon the Qwen2.5-VL architecture. It is specifically designed to enhance spatial intelligence, demonstrating strong performance across various spatial reasoning benchmarks. This model excels at tasks requiring visual understanding and spatial cognition, making it suitable for applications demanding advanced scene comprehension.
Loading preview...
SenseNova-SI-1.1-Qwen2.5-VL-7B: Enhanced Spatial Intelligence
SenseNova-SI-1.1-Qwen2.5-VL-7B is a 7 billion parameter multimodal foundation model from the SenseNova-SI family, specifically engineered to address deficiencies in spatial intelligence within existing multimodal models. Built upon the Qwen2.5-VL architecture, this model leverages a systematically curated dataset, SenseNova-SI-8M, comprising eight million diverse data samples focused on spatial capabilities.
Key Capabilities
- Advanced Spatial Reasoning: Demonstrates strong performance across a broad range of spatial intelligence benchmarks, including VSI-Bench (58.1%), MMSI (32.8%), MindCube (54.7%), ViewSpatial (45.5%), and SITE (43.9%).
- Multimodal Understanding: Maintains robust general multimodal understanding, achieving 84.9% on MMBench-En.
- Data-Driven Improvement: Benefits from extensive training on a diverse dataset designed to cultivate and scale spatial intelligence.
When to Use This Model
- Applications Requiring Spatial Cognition: Ideal for tasks where understanding spatial relationships, object locations, and scene geometry is critical.
- Research in Spatial AI: Useful for researchers exploring the scaling of spatial intelligence in multimodal foundation models and analyzing the impact of diverse data training.
- Benchmarking Spatial Tasks: Provides a strong baseline for evaluating performance on various spatial intelligence benchmarks, as detailed on the EASI leaderboard.