Yan Song

This work augments the LALM with a dedicated frame-level grounding model while leveraging its semantic modeling capability to represent the event query, and shows that the grounding model can provide temporal evidence to support downstream reasoning.

Traditional trajectory-based training collapses policy diversity, but DART-SD's topology-aware approach ensures agents can explore optimal paths without penalty.

The optimal spectrogram configuration for audio and speech analysis hinges on a nuanced interplay between front-end feature representation and back-end classifier architecture, varying significantly across tasks.