Search papers, labs, and topics across Lattice.
Tencent Inc
1
0
4
Current video understanding models struggle with long-horizon robustness and non-speech audio, as revealed by the new OmniPro benchmark designed for comprehensive omni-modal proactive evaluation.