Search papers, labs, and topics across Lattice.
2
0
5
6
Self-supervised speech models don't learn all linguistic features at once: the *order* in which they learn depends heavily on the pre-training objective and the level of abstraction of the linguistic feature.
Instruction tuning can reduce masculine bias in decoder-only MT models, but these models still don't consistently outperform encoder-decoder architectures on gender-specific translation tasks.
