Search papers, labs, and topics across Lattice.
5
0
9
5
WeSep reveals that decoupling cue modules from separator architectures can significantly enhance the adaptability and performance of Target Speaker Extraction systems in real-world environments.
Real-world conversational dynamics significantly challenge target speaker extraction, revealing that even advanced systems struggle with natural overlap and noise.
PhysGraph achieves state-of-the-art results in multi-object mass estimation and articulation prediction by seamlessly integrating physical properties into 3D scene representations.
LLM defenses can achieve a 79% reduction in attack success rate against evolving multi-round attacks by using a stateful, multi-agent cooperative framework.
G-STAR tackles long-form, multi-speaker ASR by giving Speech-LLMs time-aware speaker tracking, enabling robust identity linking across chunks.
