Search papers, labs, and topics across Lattice.
Affiliation:
2
0
3
Achieving reproducible evaluations across multiple modalities without the overhead of maintaining separate toolchains could revolutionize how foundation models are assessed.
LALMs are falling short in paralinguistic evaluations, missing the mark by 32 percentage points compared to human judges.
