Search papers, labs, and topics across Lattice.
2
0
5
Unlock better Automatic Chord Recognition by distilling knowledge from readily available pre-trained models using pseudo-labels on unlabeled data, surpassing the performance of both supervised baselines and the original teacher model.
General-purpose LLM agents stumble badly when faced with the messy reality of diverse, multi-domain tasks, and simply scaling interactions or parallel sampling doesn't fix it.
