Search papers, labs, and topics across Lattice.
Southeast University
3
0
6
Achieving 134 FPS with under 50 ms latency, ViCoStream redefines the capabilities of VideoLLMs for real-time streaming applications.
Hierarchical alignment of standardized embeddings can significantly enhance zero-shot learning performance by addressing the inherent mismatches between audio-visual and textual modalities.
Forget clunky external modules: MAIL++ injects cross-modal learning directly into the VLM's core, boosting few-shot performance without sacrificing inference speed.
